Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

This is WILD ! 2 x 3090s 🚀 qwen3.6-35b-a3b-autoround on 3090 35B 32 parallel coding agents • Total: 25,943 tokens in 17s • Aggregate: 252275.4 tok/s peak · 1548.1 tok/s sustained • Per-stream: 20000.0 high / 1666.7 low / 9567.7 avg tok/s • TTFT 0.43s avg · E2E 17s avg

Tech2Wild

1,499 subscribers

14,587 просмотров • 9 дней назад •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

How Fast is Gemma 4 on a MacBook Pro M4? Benchmarking Google's new MoE (26B-A4B) > Model size: 26.1 GiB > Load time: ~4.2s Comparing single request VS > concurrent requests performance > 32k total context, 4 parallel slots single request behavior > TTFT: 5.68s > prompt: 3,701 tokens @ 652 tok/s > decode: 40.08 tok/s sequential (1 request at a time): > avg duration: 20.5s > p99: 22.1s > throughput: 40.11 tok/s > clean finishes: 100% concurrent (4 parallel requests): > aggregate throughput: 47.25 tok/s > total system throughput: 262.27 tok/s > avg duration: 65.1s > p95 latency: 68.8s > req/sec: 0.058 Head-to-Head: Sequential vs Concurrent throughput: > 40.11 tok/s → 47.25 tok/s (+17.8%) > small gain despite 4x parallelism latency per request: > 20.5s → 65.1s (~3.2x slower) > you pay heavily for concurrency system throughput (true utilization): > ~40 tok/s → 262 tok/s (~6.5x total output) > this is where concurrency wins tokens per second (decode ceiling): > ~40 tok/s steady in both modes > hardware-bound, not scheduler-bound TTFT impact: > ~5.7s baseline → buried under queueing in concurrent > “headers waittime” becomes the bottleneck What this actually means? - You don’t get linear scaling from parallel slots - You trade latency for total output - Mac Unified Memory setup is clearly saturating - Bandwidth + Scheduling overhead show up immediately This is exactly why GPUs dominate here Concurrency without killing latency

How Fast is Gemma 4 on a MacBook Pro M4? Benchmarking Google's new MoE (26B-A4B) > Model size: 26.1 GiB > Load time: ~4.2s Comparing single request VS > concurrent requests performance > 32k total context, 4 parallel slots single request behavior > TTFT: 5.68s > prompt: 3,701 tokens @ 652 tok/s > decode: 40.08 tok/s sequential (1 request at a time): > avg duration: 20.5s > p99: 22.1s > throughput: 40.11 tok/s > clean finishes: 100% concurrent (4 parallel requests): > aggregate throughput: 47.25 tok/s > total system throughput: 262.27 tok/s > avg duration: 65.1s > p95 latency: 68.8s > req/sec: 0.058 Head-to-Head: Sequential vs Concurrent throughput: > 40.11 tok/s → 47.25 tok/s (+17.8%) > small gain despite 4x parallelism latency per request: > 20.5s → 65.1s (~3.2x slower) > you pay heavily for concurrency system throughput (true utilization): > ~40 tok/s → 262 tok/s (~6.5x total output) > this is where concurrency wins tokens per second (decode ceiling): > ~40 tok/s steady in both modes > hardware-bound, not scheduler-bound TTFT impact: > ~5.7s baseline → buried under queueing in concurrent > “headers waittime” becomes the bottleneck What this actually means? - You don’t get linear scaling from parallel slots - You trade latency for total output - Mac Unified Memory setup is clearly saturating - Bandwidth + Scheduling overhead show up immediately This is exactly why GPUs dominate here Concurrency without killing latency

Ahmad

88,866 просмотров • 3 месяцев назад

Everyone's comparing the DGX Spark to a 5090 and calling it slow. I think that's the wrong comparison. I ran Qwen3.6 35B-A3B FP8 with the full 262K context window enabled (~96GB RAM) — something gaming GPUs can't really do. Results: 🟢No context: 51.3 tok/s, TTFT 110ms 🟣200K prefill: 34.6 tok/s, TTFT 85s (~2,341 tok/s prefill) Prefill is way faster than a Mac. And 35 tok/s deep into 200K context, on a model this strong, is genuinely usable. The Spark plays a different game.

Everyone's comparing the DGX Spark to a 5090 and calling it slow. I think that's the wrong comparison. I ran Qwen3.6 35B-A3B FP8 with the full 262K context window enabled (~96GB RAM) — something gaming GPUs can't really do. Results: 🟢No context: 51.3 tok/s, TTFT 110ms 🟣200K prefill: 34.6 tok/s, TTFT 85s (~2,341 tok/s prefill) Prefill is way faster than a Mac. And 35 tok/s deep into 200K context, on a model this strong, is genuinely usable. The Spark plays a different game.

stevibe

33,301 просмотров • 2 месяцев назад

DFlash v0.1.4 : custom Metal verify kernels for quantized Qwen3 hybrid models, plus significant peak memory reduction at long context. M5 Max 40-core GPU, 64GB, stock mlx_lm baseline: Qwen3.6-35B-A3B-4bit: ► @ 1024 · 138.3 → 300.3 tok/s (2.20x) ► @ 2048 · 135.6 → 246.4 tok/s (1.81x) ► @ 4096 · 134.5 → 208.4 tok/s (1.56x) ► @ 8192 · 133.2 → 177.4 tok/s (1.33x) Qwen3.5-27B-4bit: ► @ 1024 · 33.5 → 79.0 tok/s (2.37x) ► @ 2048 · 33.1 → 70.2 tok/s (2.12x) ► @ 4096 · 31.5 → 55.7 tok/s (1.77x) ► @ 8192 · 33.9 → 45.3 tok/s (1.34x) Working on making this usable for agentic workloads goal is to never drop below baseline at any context depth. LLM decode is memory-bandwidth bound. M5 Max runs at 614 GB/s, that's 1.5x more than M1-M4 Max (400-410 GB/s). Results will vary on lower bandwidth chips.

DFlash v0.1.4 : custom Metal verify kernels for quantized Qwen3 hybrid models, plus significant peak memory reduction at long context. M5 Max 40-core GPU, 64GB, stock mlx_lm baseline: Qwen3.6-35B-A3B-4bit: ► @ 1024 · 138.3 → 300.3 tok/s (2.20x) ► @ 2048 · 135.6 → 246.4 tok/s (1.81x) ► @ 4096 · 134.5 → 208.4 tok/s (1.56x) ► @ 8192 · 133.2 → 177.4 tok/s (1.33x) Qwen3.5-27B-4bit: ► @ 1024 · 33.5 → 79.0 tok/s (2.37x) ► @ 2048 · 33.1 → 70.2 tok/s (2.12x) ► @ 4096 · 31.5 → 55.7 tok/s (1.77x) ► @ 8192 · 33.9 → 45.3 tok/s (1.34x) Working on making this usable for agentic workloads goal is to never drop below baseline at any context depth. LLM decode is memory-bandwidth bound. M5 Max runs at 614 GB/s, that's 1.5x more than M1-M4 Max (400-410 GB/s). Results will vary on lower bandwidth chips.

bstn 👁️

23,120 просмотров • 2 месяцев назад

Qwen3.6 35B-A3B dropped yesterday, so I ran it on 4 GPUs to see how it performs: 🟣 RTX 3090 — 49.78 tok/s, TTFT 852ms 🟡 RTX 4090 — 118.93 tok/s, TTFT 686ms 🟢 RTX 5090 — 160.37 tok/s, TTFT 409ms 🔵 DGX Spark — 59.98 tok/s, TTFT 228ms I went with ollama as the backend because honestly, it's the easiest way for most people to get started. One command, model pulled, done. I used Q4_K_M (24GB) across all four cards. The reason is the 3090 and 4090 don't support NVFP4 (only the 5090 and DGX Spark could use it). Keeping the same quant everywhere felt like the fairest way to compare. And yes, you can absolutely squeeze more performance out of every card with vLLM, SGLang, or TensorRT-LLM. But that's not what this test is about. This is just the out-of-the-box experience for folks who own a GPU and want to try the new model tonight.

Qwen3.6 35B-A3B dropped yesterday, so I ran it on 4 GPUs to see how it performs: 🟣 RTX 3090 — 49.78 tok/s, TTFT 852ms 🟡 RTX 4090 — 118.93 tok/s, TTFT 686ms 🟢 RTX 5090 — 160.37 tok/s, TTFT 409ms 🔵 DGX Spark — 59.98 tok/s, TTFT 228ms I went with ollama as the backend because honestly, it's the easiest way for most people to get started. One command, model pulled, done. I used Q4_K_M (24GB) across all four cards. The reason is the 3090 and 4090 don't support NVFP4 (only the 5090 and DGX Spark could use it). Keeping the same quant everywhere felt like the fairest way to compare. And yes, you can absolutely squeeze more performance out of every card with vLLM, SGLang, or TensorRT-LLM. But that's not what this test is about. This is just the out-of-the-box experience for folks who own a GPU and want to try the new model tonight.

stevibe

398,628 просмотров • 2 месяцев назад

testing Qwen3.5-35B-A3B latest optimized version by UnslothAI on a single RTX 3090. one detailed prompt. zero handholding. watch a 3B model scaffold an entire multifile game project autonomously. the setup: > model: Qwen3.5-35B-A3B (80B total, only 3B active per token) > quant: UD-Q4_K_XL by Unsloth (MXFP4 layers removed in latest update) > speed: 112 tok/s generation, ~130 tok/s prefill > context: 262K tokens > flags: -ngl 99 -c 262144 -np 1 --cache-type-k q8_0 --cache-type-v q8_0 > engine: llama.cpp > agent: Claude Code talk to localhost:8080 (llama.cpp now has native Anthropic API endpoint. no LiteLLM needed) q8_0 KV cache cuts VRAM usage in half vs f16 at 262K. -np 1 is default but worth noting. parallel slots multiply KV cache and at 262K that's an instant OOM. the prompt was more detailed than this but you get the idea: build a space shooter with parallax backgrounds, particle systems, procedural audio, 4 enemy types, boss fights, power-up system, and ship upgrades. 8 JavaScript modules. no libraries. game's called Octopus Invaders. gameplay footage dropping next.

testing Qwen3.5-35B-A3B latest optimized version by UnslothAI on a single RTX 3090. one detailed prompt. zero handholding. watch a 3B model scaffold an entire multifile game project autonomously. the setup: > model: Qwen3.5-35B-A3B (80B total, only 3B active per token) > quant: UD-Q4_K_XL by Unsloth (MXFP4 layers removed in latest update) > speed: 112 tok/s generation, ~130 tok/s prefill > context: 262K tokens > flags: -ngl 99 -c 262144 -np 1 --cache-type-k q8_0 --cache-type-v q8_0 > engine: llama.cpp > agent: Claude Code talk to localhost:8080 (llama.cpp now has native Anthropic API endpoint. no LiteLLM needed) q8_0 KV cache cuts VRAM usage in half vs f16 at 262K. -np 1 is default but worth noting. parallel slots multiply KV cache and at 262K that's an instant OOM. the prompt was more detailed than this but you get the idea: build a space shooter with parallax backgrounds, particle systems, procedural audio, 4 enemy types, boss fights, power-up system, and ship upgrades. 8 JavaScript modules. no libraries. game's called Octopus Invaders. gameplay footage dropping next.

Sudo su

167,035 просмотров • 4 месяцев назад

dflash-mlx v0.1.7 is out. Big adaptive-runtime update, still focused mostly on Qwen3.6 27B 4-bit. @ 2048 tokens, M5 Max, stock mlx_lm baseline: ► 1024: 33.26 → 98.05 tok/s (x2.95) ► 2048: 32.34 → 90.67 tok/s (x2.81) ► 4096: 30.58 → 93.55 tok/s (x3.06) ► 8192: 26.03 → 79.12 tok/s (x3.04) ► 16384: 21.50 → 60.77 tok/s (x2.78) Main change: adaptive verify got a lot smarter. Instead of blindly trying to verify large 16-token blocks all the time, DFlash now watches acceptance + tokens/cycle + real cycle cost. When the draft gets weaker, it drops to smaller 4-token blocks, then probes back up only when the recent cycles make sense. In practice: less wasted verify work, better long-context behavior, and much more useful metrics to understand what is happening. ► retuned adaptive verify for long-context / agentic decode ► richer metrics: tokens/cycle, adaptive block state, CopySpec counters ► /metrics now has real decode avg + logical/real/restored prefill rates ► AIME25 benchmark suite with exact integer scoring ► Qwen thinking default now follows tokenizer/request behavior ► GDN recurrent exactness fixes I also started running AIME25-style long generations. Even around 45k generated tokens, I was still seeing ~40 tok/s on 27B 4-bit. Over the next few days I’ll share more demos: AIME runs, real OpenCode game/project sessions, and full metrics along the way. Still optimizing hard for 27B 4-bit first, while working on custom kernels per Apple GPU generation so more machines can benefit.

dflash-mlx v0.1.7 is out. Big adaptive-runtime update, still focused mostly on Qwen3.6 27B 4-bit. @ 2048 tokens, M5 Max, stock mlx_lm baseline: ► 1024: 33.26 → 98.05 tok/s (x2.95) ► 2048: 32.34 → 90.67 tok/s (x2.81) ► 4096: 30.58 → 93.55 tok/s (x3.06) ► 8192: 26.03 → 79.12 tok/s (x3.04) ► 16384: 21.50 → 60.77 tok/s (x2.78) Main change: adaptive verify got a lot smarter. Instead of blindly trying to verify large 16-token blocks all the time, DFlash now watches acceptance + tokens/cycle + real cycle cost. When the draft gets weaker, it drops to smaller 4-token blocks, then probes back up only when the recent cycles make sense. In practice: less wasted verify work, better long-context behavior, and much more useful metrics to understand what is happening. ► retuned adaptive verify for long-context / agentic decode ► richer metrics: tokens/cycle, adaptive block state, CopySpec counters ► /metrics now has real decode avg + logical/real/restored prefill rates ► AIME25 benchmark suite with exact integer scoring ► Qwen thinking default now follows tokenizer/request behavior ► GDN recurrent exactness fixes I also started running AIME25-style long generations. Even around 45k generated tokens, I was still seeing ~40 tok/s on 27B 4-bit. Over the next few days I’ll share more demos: AIME runs, real OpenCode game/project sessions, and full metrics along the way. Still optimizing hard for 27B 4-bit first, while working on custom kernels per Apple GPU generation so more machines can benefit.

bstn 👁️

16,334 просмотров • 1 месяц назад

oMLX last version seems having some issues in decoding. 🧐 Here a test with Qwen3.6-35B-A3B-MLX-6bit on M5 Max using. 🥇 LM Studio MLX 1.8.5 → 100.9 toks/s 🥈 mlx-vlm 0.6.2 → 100.1 toks/s 🥉 oMLX 0.4.2 dev3 → 58.7 toks/s 👀 Avg Gen TPS: oMLX 58.7 → LM Studio 100.9 (+71.8%) I have to thank pymike00 that raised this oMLX issue after seeing my video on using it with Codex. I bet there is a bug in oMLX chat and server at the moment, because internal benchmarks are ok, video attached. I bet Jun Kim will fix this soon 💪

oMLX last version seems having some issues in decoding. 🧐 Here a test with Qwen3.6-35B-A3B-MLX-6bit on M5 Max using. 🥇 LM Studio MLX 1.8.5 → 100.9 toks/s 🥈 mlx-vlm 0.6.2 → 100.1 toks/s 🥉 oMLX 0.4.2 dev3 → 58.7 toks/s 👀 Avg Gen TPS: oMLX 58.7 → LM Studio 100.9 (+71.8%) I have to thank pymike00 that raised this oMLX issue after seeing my video on using it with Codex. I bet there is a bug in oMLX chat and server at the moment, because internal benchmarks are ok, video attached. I bet Jun Kim will fix this soon 💪

Ivan Fioravanti ᯅ

19,931 просмотров • 1 месяц назад

nvidia's 3B mamba destroyed alibaba's 3B deltanet on the same RTX 3090. only 24 days between releases. same active parameters, same VRAM tier, completely different architectures. nemotron cascade 2: 187 tok/s. flat from 4K to 625K context. zero speed loss. flags: -ngl 99 -np 1. that's it. no context flags, no KV cache tricks. auto-allocates 625K. qwen 3.5 35B-A3B: 112 tok/s. flat from 4K to 262K context. zero speed loss. flags: -ngl 99 -np 1 -c 262144 --cache-type-k q8_0 --cache-type-v q8_0. needed KV cache quantization to fit 262K. both models held a flat line across every context level. both architectures are context-independent. but nvidia's mamba2 is 67% faster at generating tokens on the exact same hardware and needs fewer flags to get there. same node, same GPU, same everything. the only variable is the model. gold medal math olympiad winner running at 187 tokens per second on single RTX 3090 a card from 6 years ago. nvidia cooked.

nvidia's 3B mamba destroyed alibaba's 3B deltanet on the same RTX 3090. only 24 days between releases. same active parameters, same VRAM tier, completely different architectures. nemotron cascade 2: 187 tok/s. flat from 4K to 625K context. zero speed loss. flags: -ngl 99 -np 1. that's it. no context flags, no KV cache tricks. auto-allocates 625K. qwen 3.5 35B-A3B: 112 tok/s. flat from 4K to 262K context. zero speed loss. flags: -ngl 99 -np 1 -c 262144 --cache-type-k q8_0 --cache-type-v q8_0. needed KV cache quantization to fit 262K. both models held a flat line across every context level. both architectures are context-independent. but nvidia's mamba2 is 67% faster at generating tokens on the exact same hardware and needs fewer flags to get there. same node, same GPU, same everything. the only variable is the model. gold medal math olympiad winner running at 187 tokens per second on single RTX 3090 a card from 6 years ago. nvidia cooked.

Sudo su

186,642 просмотров • 3 месяцев назад

Anyone with 8GB or 12GB VRAM setups needs to understand that "-ncmoe" is the key flag to boost performance on llama.cpp Here are my results for Qwen3.6 35B A3B, with 64k q8_0 context on a 8GB RTX 3070Ti: ⚪️ no flag → 8.7 tok/s RAM: 13.6GB & VRAM: 7.8GB 🔴 -ncmoe 35 → 27.5 tok/s RAM: 12.1GB & VRAM: 4.3GB 🟢 -ncmoe 30 → 32.5 tok/s RAM: 12GB & VRAM: 5.6GB 🔵 -ncmoe 25 → 40.9 tok/s RAM: 12GB & VRAM: 6.9GB Please note the ram and vram usage you see are total usage of a windows pc, with the model running. My friend's setup: 8GB VRAM and 16GB RAM. You can boost performance by switching to Linux, just something to keep in mind. Basically, this flag keeps the MoE experts in the first X layers on your CPU + RAM, instead of eating all your VRAM straight away. This is a smart hybrid offload way that lets you run bigger models without OOM while keeping the rest on your GPU for speed. As we can see on the data, there's a sweet spot. When we lower it from 35 to 25, speed bumps +50% because there are more layers on your GPU (look at the VRAM usage). The key here is to play around with the number and fit as much as possible on your VRAM, goal is to have 1GB/800MB headroom to avoid stress. ↓ server flags below

Anyone with 8GB or 12GB VRAM setups needs to understand that "-ncmoe" is the key flag to boost performance on llama.cpp Here are my results for Qwen3.6 35B A3B, with 64k q8_0 context on a 8GB RTX 3070Ti: ⚪️ no flag → 8.7 tok/s RAM: 13.6GB & VRAM: 7.8GB 🔴 -ncmoe 35 → 27.5 tok/s RAM: 12.1GB & VRAM: 4.3GB 🟢 -ncmoe 30 → 32.5 tok/s RAM: 12GB & VRAM: 5.6GB 🔵 -ncmoe 25 → 40.9 tok/s RAM: 12GB & VRAM: 6.9GB Please note the ram and vram usage you see are total usage of a windows pc, with the model running. My friend's setup: 8GB VRAM and 16GB RAM. You can boost performance by switching to Linux, just something to keep in mind. Basically, this flag keeps the MoE experts in the first X layers on your CPU + RAM, instead of eating all your VRAM straight away. This is a smart hybrid offload way that lets you run bigger models without OOM while keeping the rest on your GPU for speed. As we can see on the data, there's a sweet spot. When we lower it from 35 to 25, speed bumps +50% because there are more layers on your GPU (look at the VRAM usage). The key here is to play around with the number and fit as much as possible on your VRAM, goal is to have 1GB/800MB headroom to avoid stress. ↓ server flags below

left curve dev

166,063 просмотров • 2 месяцев назад

Hot off the press 🔥 AEON vLLM Ultimate → v0.24.0. Link below ⤵️ The ultimate container for AI on NVIDIA DGX Spark — vLLM compiled from source for GB10/sm_121a. 🔓 UNLOCKED (impossible until this release): ▸ DFlash + FP8 KV cache — 2× KV capacity, draft acceptance intact, measured on-device ▸ FP8 KV on vision models — the KV-bandwidth lever is back ▸ DFlash on FlashInfer & Triton backends (was FlashAttention-only) 🛠 Carried ahead of upstream (still-open PRs, merged + validated): ▸ Triton NVFP4 KV cache → up to 3× KV capacity ▸ DFlash sliding-window drafters + prefix-cache stability for DFlash ▸ High-concurrency DFlash — clean through c=64 🧰 Fixes stock v0.24.0 doesn't have: ▸ Gemma-4 tied embeddings (ModelOpt) — stock crashes at load, ours doesn't ▸ Tool calling out of the box (stock --no-deps builds 500) ▸ Unified-memory hardened — no silent KV overcommit on the 121 GB pool 📊 One image, the whole fleet: Gemma-4-26B-A4B · Qwen3.6-27B · Qwen3.6-35B-A3B (hybrid GDN) 536 tok/s @ c16 · 472 tok/s @ c12 · 262K context · image+video · tools + reasoning parsers · full CUDA-graph capture 🐳 docker pull

Hot off the press 🔥 AEON vLLM Ultimate → v0.24.0. Link below ⤵️ The ultimate container for AI on NVIDIA DGX Spark — vLLM compiled from source for GB10/sm_121a. 🔓 UNLOCKED (impossible until this release): ▸ DFlash + FP8 KV cache — 2× KV capacity, draft acceptance intact, measured on-device ▸ FP8 KV on vision models — the KV-bandwidth lever is back ▸ DFlash on FlashInfer & Triton backends (was FlashAttention-only) 🛠 Carried ahead of upstream (still-open PRs, merged + validated): ▸ Triton NVFP4 KV cache → up to 3× KV capacity ▸ DFlash sliding-window drafters + prefix-cache stability for DFlash ▸ High-concurrency DFlash — clean through c=64 🧰 Fixes stock v0.24.0 doesn't have: ▸ Gemma-4 tied embeddings (ModelOpt) — stock crashes at load, ours doesn't ▸ Tool calling out of the box (stock --no-deps builds 500) ▸ Unified-memory hardened — no silent KV overcommit on the 121 GB pool 📊 One image, the whole fleet: Gemma-4-26B-A4B · Qwen3.6-27B · Qwen3.6-35B-A3B (hybrid GDN) 536 tok/s @ c16 · 472 tok/s @ c12 · 262K context · image+video · tools + reasoning parsers · full CUDA-graph capture 🐳 docker pull

ÆON FORGE ✨

24,478 просмотров • 13 дней назад

$A $351 MINI PC IS RUNNING 26-BILLION-PARAMETER AI MODELS AT 20 TOKENS/SEC AND HERMES AGENT ON TOP OF IT This is the Minisforum UM790 Pro. AMD Ryzen 9 7940HS, Radeon 780M iGPU, 48GB DDR5. The BIOS reports the GPU has 4GB of VRAM Here's the part people get wrong. The 780M has no dedicated VRAM at all it borrows from system RAM. Vulkan ignores the BIOS number and reads the full 48GB pool directly That's the whole trick. 21+ GB allocated to model weights on a "4GB" GPU. Because the models are MoE, only 4 billion of the parameters activate per token a fraction of the reads a dense model needs. That's why 20 tok/s works here Gemma 4 26B MoE holds 19.5 tok/s with 196K context. Qwen3.5-35B-A3B holds 20.8. Nemotron Cascade 2 clears 24.8. A dense 31B, by contrast, drops to 4 tok/s it reads the entire model every token, no way around it On top: Hermes Agent, full agentic workflows terminal, file ops, web, 40+ tools against local models only. No API keys. The wall between you and a usable local agent used to be a GPU you couldn't afford. Now it's a BIOS setting most people never check Bookmark this & Try it yourself ↓$

A $351 MINI PC IS RUNNING 26-BILLION-PARAMETER AI MODELS AT 20 TOKENS/SEC AND HERMES AGENT ON TOP OF IT This is the Minisforum UM790 Pro. AMD Ryzen 9 7940HS, Radeon 780M iGPU, 48GB DDR5. The BIOS reports the GPU has 4GB of VRAM Here's the part people get wrong. The 780M has no dedicated VRAM at all it borrows from system RAM. Vulkan ignores the BIOS number and reads the full 48GB pool directly That's the whole trick. 21+ GB allocated to model weights on a "4GB" GPU. Because the models are MoE, only 4 billion of the parameters activate per token a fraction of the reads a dense model needs. That's why 20 tok/s works here Gemma 4 26B MoE holds 19.5 tok/s with 196K context. Qwen3.5-35B-A3B holds 20.8. Nemotron Cascade 2 clears 24.8. A dense 31B, by contrast, drops to 4 tok/s it reads the entire model every token, no way around it On top: Hermes Agent, full agentic workflows terminal, file ops, web, 40+ tools against local models only. No API keys. The wall between you and a usable local agent used to be a GPU you couldn't afford. Now it's a BIOS setting most people never check Bookmark this & Try it yourself ↓

slash1s

35,742 просмотров • 26 дней назад

the tiebreaker is done. qwen 3.5 27B dense. single RTX 3090. one prompt. zero steering. zero human edits. 1,827 lines across 10 files. 13 minutes. full thinking mode. runs on first load. hermes 4.3 got the same prompt with 2x 3090s and 5x the context it needed. wrote 1,249 lines, left empty files, needed 3 interventions, game was broken on load. same architecture class. same quant. hermes got double the hardware. completely different result. dense wasn't the problem. hermes was. but here's what got me. this model thinks at 27 tok/s. every single token carries 27 billion parameters of reasoning. MoE hit 112 tok/s but only 3B active per token. the dense model is slower and it doesn't matter. watch 13 minutes of autonomous coding on a consumer GPU with zero intervention and tell me speed is what matters. a year ago this wasn't possible. now it runs on hardware you can buy used for $900. no API. no subscription. no cloud. just a 3090 doing what data centers did 18 months ago. full unedited session in the video. every token, every file, every thinking chain. 16 minutes. hit play.

the tiebreaker is done. qwen 3.5 27B dense. single RTX 3090. one prompt. zero steering. zero human edits. 1,827 lines across 10 files. 13 minutes. full thinking mode. runs on first load. hermes 4.3 got the same prompt with 2x 3090s and 5x the context it needed. wrote 1,249 lines, left empty files, needed 3 interventions, game was broken on load. same architecture class. same quant. hermes got double the hardware. completely different result. dense wasn't the problem. hermes was. but here's what got me. this model thinks at 27 tok/s. every single token carries 27 billion parameters of reasoning. MoE hit 112 tok/s but only 3B active per token. the dense model is slower and it doesn't matter. watch 13 minutes of autonomous coding on a consumer GPU with zero intervention and tell me speed is what matters. a year ago this wasn't possible. now it runs on hardware you can buy used for $900. no API. no subscription. no cloud. just a 3090 doing what data centers did 18 months ago. full unedited session in the video. every token, every file, every thinking chain. 16 minutes. hit play.

Sudo su

91,135 просмотров • 4 месяцев назад

first impressions of qwen 3.5 27B dense on a single RTX 3090. 35 tok/s. from 4K all the way to 300K+ context. no speed drop. hermes 4.3 started at 35 and degraded to 15 as context filled. qwen dense holds. MoE held 112 flat. 3x faster but only 3B of 35B active per token. architecture tradeoff. Q4_K_M on 16.7GB. native context 262K. pushed past training limit to 376K before VRAM ceiling on 24GB. tried q8 KV cache at 262K, speed collapsed to 11 tok/s. q4_0 KV is the sweet spot. flash attention mandatory. built in reasoning mode. the model thinks step by step before it answers. full chain of thought surviving Q4 quant. 1,799+ token thinking chains with self correction loops. on a single consumer GPU. gave it one prompt: "build a realtime particle galaxy simulation in one HTML file." 3,340 tokens. 95 seconds. one shot. ran on first load. full reasoning and coding in the video below. optimal config if you want to skip the hours of testing: llama-server -ngl 99 -c 262144 -fa on --cache-type-k q4_0 --cache-type-v q4_0 this is just the warmup. octopus invaders is next: 10 files, 3,400+ lines, zero steering. the prompt hermes quit at 22%. already more impressed than expected. full results coming soon.

first impressions of qwen 3.5 27B dense on a single RTX 3090. 35 tok/s. from 4K all the way to 300K+ context. no speed drop. hermes 4.3 started at 35 and degraded to 15 as context filled. qwen dense holds. MoE held 112 flat. 3x faster but only 3B of 35B active per token. architecture tradeoff. Q4_K_M on 16.7GB. native context 262K. pushed past training limit to 376K before VRAM ceiling on 24GB. tried q8 KV cache at 262K, speed collapsed to 11 tok/s. q4_0 KV is the sweet spot. flash attention mandatory. built in reasoning mode. the model thinks step by step before it answers. full chain of thought surviving Q4 quant. 1,799+ token thinking chains with self correction loops. on a single consumer GPU. gave it one prompt: "build a realtime particle galaxy simulation in one HTML file." 3,340 tokens. 95 seconds. one shot. ran on first load. full reasoning and coding in the video below. optimal config if you want to skip the hours of testing: llama-server -ngl 99 -c 262144 -fa on --cache-type-k q4_0 --cache-type-v q4_0 this is just the warmup. octopus invaders is next: 10 files, 3,400+ lines, zero steering. the prompt hermes quit at 22%. already more impressed than expected. full results coming soon.

Sudo su

120,098 просмотров • 4 месяцев назад

no prompt engineering, no agentic harness, no tool calls. just me being lazy in llama.cpp's web ui and gemma 4 31b dense taking the task seriously. i typed "create gpu marketplace cards with hardware specs and prices per hour" and the model went and coded this ui, one shot, navy bg, glassmorphism cards, neon accent buttons, realistic pricing tiers per architecture. it even wrote a "why this looks premium" explanation under the code. for context this is a q4 quant of google's 31b dense thinking model, running on a rtx 5090 mobile 24gb in the rog scar 18 at around 15 tok/s sustained, same vram tier as a 3090 or 4090 desktop, so whatever you see here translates directly to your card at home. the whole interaction was me not trying and the model reasoning harder than the prompt deserved. that tells me more about where local ai is at in april 2026 than any leaderboard score. next test drops gemma 4 into hermes agent, autonomous tool calling, multi step reasoning, real agentic loop instead of a chat window. let's see what the same model does when it gets the right environment. more experiments coming anon. octopus invaders queued. same hardware, different tasks, all published here on x and all translatable to your 24gb card. for now the video below shows it coding live, gpu going brrr.

no prompt engineering, no agentic harness, no tool calls. just me being lazy in llama.cpp's web ui and gemma 4 31b dense taking the task seriously. i typed "create gpu marketplace cards with hardware specs and prices per hour" and the model went and coded this ui, one shot, navy bg, glassmorphism cards, neon accent buttons, realistic pricing tiers per architecture. it even wrote a "why this looks premium" explanation under the code. for context this is a q4 quant of google's 31b dense thinking model, running on a rtx 5090 mobile 24gb in the rog scar 18 at around 15 tok/s sustained, same vram tier as a 3090 or 4090 desktop, so whatever you see here translates directly to your card at home. the whole interaction was me not trying and the model reasoning harder than the prompt deserved. that tells me more about where local ai is at in april 2026 than any leaderboard score. next test drops gemma 4 into hermes agent, autonomous tool calling, multi step reasoning, real agentic loop instead of a chat window. let's see what the same model does when it gets the right environment. more experiments coming anon. octopus invaders queued. same hardware, different tasks, all published here on x and all translatable to your 24gb card. for now the video below shows it coding live, gpu going brrr.

Sudo su

28,939 просмотров • 2 месяцев назад

this is what a 24gb VRAM builds in 2026. one prompt. ten files. 3,483 lines of code. zero handholding. i gave Qwen3.5-35B-A3B a single detailed spec describing the full game architecture and hit enter. enemy types, particle systems, procedural audio, powerups, boss fights, ship upgrades, parallax backgrounds, everything in one message. the model planned the file structure itself, wrote every module in dependency order, wired all the imports, and served the game on port 3001. it ran on first load. when it hit a bug in collision detection it read its own error output, found the issue, fixed it, and kept building. this is pure agent loop running on local hardware. what you're looking at is pixelated octopus aliens with tentacle animations, 4 layer parallax space background with planets at different depths, a full particle system handling explosions and ink splatter and engine trails and bullet impacts, procedural audio through Web Audio API with zero sound files loaded, unleash mode with combo multiplier, boss fights every 5 levels, ship upgrades that unlock as you progress. no libraries. no frameworks. vanilla JS and Canvas. 3B active parameters. single RTX 3090. llama.cpp with q8_0 KV cache at 262K context. Claude Code pointed at localhost:8080 through the native Anthropic endpoint. no API costs. 112 tok/s. a GPU you can buy used for $800. game is called Octopus Invaders and i actually like playing it.

this is what a 24gb VRAM builds in 2026. one prompt. ten files. 3,483 lines of code. zero handholding. i gave Qwen3.5-35B-A3B a single detailed spec describing the full game architecture and hit enter. enemy types, particle systems, procedural audio, powerups, boss fights, ship upgrades, parallax backgrounds, everything in one message. the model planned the file structure itself, wrote every module in dependency order, wired all the imports, and served the game on port 3001. it ran on first load. when it hit a bug in collision detection it read its own error output, found the issue, fixed it, and kept building. this is pure agent loop running on local hardware. what you're looking at is pixelated octopus aliens with tentacle animations, 4 layer parallax space background with planets at different depths, a full particle system handling explosions and ink splatter and engine trails and bullet impacts, procedural audio through Web Audio API with zero sound files loaded, unleash mode with combo multiplier, boss fights every 5 levels, ship upgrades that unlock as you progress. no libraries. no frameworks. vanilla JS and Canvas. 3B active parameters. single RTX 3090. llama.cpp with q8_0 KV cache at 262K context. Claude Code pointed at localhost:8080 through the native Anthropic endpoint. no API costs. 112 tok/s. a GPU you can buy used for $800. game is called Octopus Invaders and i actually like playing it.

Sudo su

153,735 просмотров • 4 месяцев назад

hey if you're thinking about running qwopus (the claude opus distilled qwen 3.5 27B) as a coding agent, this might save you a few hours. i tested both the base and the distilled version on the same hardware. single RTX 3090. same prompt. same context. same everything. the only variable was the model weights. base qwen 3.5 27B built octopus invaders in 13 minutes. 1,827 lines across 11 files. zero steering. one scope bug that took 2 lines to fix. game ran. qwopus couldn't finish the same task. enemies overlapping on screen. bullets not firing. controls worked but the game was broken. i had to steer it multiple times and it still didn't produce a playable result. both run at 35 tok/s. both use thinking mode. the distilled version actually has better jinja compatibility and doesn't stall midtask like base does on claude code. for conversation and reasoning it feels sharper. but for multifile autonomous coding where the model needs to coordinate 10+ files without losing track, base wins and it's not close. distillation compresses reasoning patterns but seems to lose precision on complex coordination. the model "thinks" well but can't hold the full picture across files the way base can. tested on opencode (base) and claude code (both). next up is hermes agent framework on base. same hardware. same prompt. comparing agents now, not just models. video below. first half is the distilled model's broken game. second half is what base built on the same 3090. judge for yourself.

hey if you're thinking about running qwopus (the claude opus distilled qwen 3.5 27B) as a coding agent, this might save you a few hours. i tested both the base and the distilled version on the same hardware. single RTX 3090. same prompt. same context. same everything. the only variable was the model weights. base qwen 3.5 27B built octopus invaders in 13 minutes. 1,827 lines across 11 files. zero steering. one scope bug that took 2 lines to fix. game ran. qwopus couldn't finish the same task. enemies overlapping on screen. bullets not firing. controls worked but the game was broken. i had to steer it multiple times and it still didn't produce a playable result. both run at 35 tok/s. both use thinking mode. the distilled version actually has better jinja compatibility and doesn't stall midtask like base does on claude code. for conversation and reasoning it feels sharper. but for multifile autonomous coding where the model needs to coordinate 10+ files without losing track, base wins and it's not close. distillation compresses reasoning patterns but seems to lose precision on complex coordination. the model "thinks" well but can't hold the full picture across files the way base can. tested on opencode (base) and claude code (both). next up is hermes agent framework on base. same hardware. same prompt. comparing agents now, not just models. video below. first half is the distilled model's broken game. second half is what base built on the same 3090. judge for yourself.

Sudo su

43,806 просмотров • 4 месяцев назад

$AMD| The FOMO to buy AMD Chips is NOW 🧵 Not Financial Advice! DYOR! Research Purpose Only! The Inference Queen is the biggest winner in Agentic AI where all other CPUs are struggling to compete with a 2yr old EPYC Turin and EPYC Venice is in mass production phase. AMD stresses deployability today on standard x86 platforms (no proprietary architectures required), full software compatibility, and open standards. This positions Venice + Helios as a practical, high-density alternative to competing solutions while underscoring that agentic AI shifts the balance toward CPU-rich racks alongside GPUs, and most importantly, lowering the cost of token to accelerate adoption and innovation. Context: The Wall Street Journal yesterday came out with an article that OpenAI is condiering drasstically lowering the token prices to win more customers from Anthropic. The narrative "they" are trying to exacerbate the current AI selloff won't last long. This is a fundamental misunderstanding of what is going on, or what I already discussed for months and years. Followers and Subscribers already knew this for years, that this day would come, where token cost will bcome the central discussion among enterprises as there is no such thing as unlimited budget or Tokenmaxxing when they use $NVDA chips or In-house Hyperscalers chips. I will link various threads if you are interested in understanding the full picture from supply chain to recent TSMC Rapid 2nm expansion up to 12 Fabs total by 2027/2028. Hyperscalers and AI natives effectively have no choice but to buy more AMD system for Agentic AI as leadership in economical, power-aware, high-volume internal + agentic use. However, due to supply constraints where Supply is far behind Demand, this makes multi-vendor reality along with in-house chips drive faster industry progress, lower overall costs, and better sustainability. NVIDIA’s Vera Rubin cannot compete with a 2 years old EPYC Turin, but AMD under Dr. Lisa Su has engineered the lowest cost-per-million-tokens, highly competitive energy-efficient solutions, and superior CPU orchestration for agentic AI at scale with Helios. Dr. Su has championed this shift since at least 2023, foreseeing the rise of agentic workflows that demand far more orchestration, parallel agents, and balanced compute well before the industry fully embraced it. Her long-term vision of AI moving from simple prompts to always on, multi-agent systems has driven AMD’s investments in high-core EPYC CPUs and integrated rack-scale solutions, perfectly positioning the company for today’s realities. The OpenAI-AMD 1GW Helios deployment (starting H2 2026) represents a pivotal vertical integration move that directly supercharges the inference economics. This isn't incremental; it's a structural shift toward ownership of massive, optimized rack-scale capacity, enabling the lowest token costs and triggering the enterprise adoption flywheel. We need to be honest, $AMD is the only company that made a big bet on Inference since the day Chatgpt became sensational where $NVDA and others were betting big on Training. At the end of the day, Token bill from Anthropic has to obey economics. Meaning the bills rise, companies have to get more out of it to justify the cost. It cannot be an unlimited inference budget, and it has to show up on efficiency, profitability and operating leverage. 1. Tokenomics After you understand this, you will understand why Citi cited Anthropic is likely to sign a deal with $AMD along with Hyperscalers, AI Labs, Sovereign AI like Softbank 5GW in France and many other countries. However, OpenAI and $META are now wanting faster deployment, and they are AMD shareholders now, they have prioritized allocation. Anthropic and Hyperscalers just cannot compete when Helios Rack lower token cost to$0.0003–$0.0005 per million tokens at GW scale. Cost to build 1GW data center 1GW Helios Rack full build is estimated $30-$35B 1GW Rubin Rack full build is estimated $45-$55B Inference (Cost per Million Tokens) ~$NVDA B200 / HGX: ~$0.02–$0.08 on optimized workloads (FP4/MXFP4, speculative decoding). Significant improvement over Hopper but still premium-priced. GB200 NVL72 rack-scale: $0.05–$0.25+ ~$AMD Helios Racks: $0.0003-$0.0005 per M tokens, dramatically lower than NVIDIA equivalents in owned infra. MI355X node-level: Up to 40% more tokens per dollar vs. competing solutions ( B200), driven by higher memory capacity (up to 288GB+ HBM), strong bandwidth, and lower acquisition costs. Training ~$NVDA Rubin Rack is estimated $0.7-$1.2/M Tokens ~$AMD Helios Rack is estimated $0.65-$1.0/M Tokens Now, OpenAI, META and Hyperscalers can lower Inference cost even further with $AMD EPYC Venice "dense rack" or Agentic AI Rack. AMD published a detailed technical blog emphasizing that the future of agentic AI autonomous, multi-step AI systems requiring heavy orchestration, databases, caching, APIs, and control planes demands massive CPU-dense rack-scale infrastructure, not just GPUs. The catalyst prominently positions their upcoming 6th Gen EPYC "Venice" processors as the key enabler for next-generation dense racks, delivering leadership throughput under real-world power, cooling, and density constraints. ~EPYC Venice (Zen 6 architecture, up to 256 cores / 512 threads per socket) is projected to deliver exceptional rack-level performance. In AMD’s modeled 100 kW rack comparisons, Venice-powered systems are expected to achieve ~3.30x the throughput of NVIDIA’s Vera (88-core Olympus) baseline across a broad mix of agentic-supporting workloads. ~This builds on current-generation 5th Gen EPYC "Turin" (up to 192 cores), which already delivers ~2.37x rack throughput vs. Vera and ~1.6x vs. Intel’s Xeon 6980P (128 cores). ~ Liquid-cooled Turin deployments already support >27,000 CPU cores per rack today. Venice is architected to push this beyond 36,000 cores in the same rack class, dramatically increasing concurrent agent capacity and overall infrastructure efficiency. 2. Ownership vs renting compute from Hyperscalers matter to OpenAI and only owning $AMD chips can meaningfully lower token cost for enterprises. ~Eliminates cloud overhead: No provider margins, utilization buffers, or egress fees. Direct control over power contracts, cooling, scheduling, and orchestration at dedicated facilities. ~Helios optimizations at GW scale: Rack-level density (1.4+ exaFLOPS FP8 per rack), high HBM4 bandwidth, EPYC orchestration for agentic workloads, and superior TCO/TDP. AMD's long-standing focus on tokens per dollar/watt shines here 20-40%+ efficiency edges in inference-heavy scenarios. ~At 1GW+ optimized deployment, inference hits $0.0003–$0.0005 per million tokens (community/analyst models tied to Helios metrics). This is dramatically lower than typical rented/cloud equivalents, especially for high-volume output tokens in agentic flows. High token bills today, enterprises running heavy agentic/coding/analysis workloads can face $50-100M+/month at current API rates (flagship models $5-30+/M output, scaled to massive volumes). Post-Helios compression, same volume will drop to $10-15M/month (or better) via lower underlying costs passed through as pricing flexibility, volume tiers, caching, or batch discounts. ROI thresholds collapse. More companies greenlight pilots → production → massive scaling. Agentic AI (autonomous workflows) multiplies token demand exponentially, but affordability removes the friction. OpenAI gains flexibility, Unlike more cloud-dependent rivals (Anthropic), they can lower effective pricing, offer aggressive enterprise bundles, or absorb volume without margin destruction directly tackling "high token bill" complaints while maintaining profitability as usage explodes. 3. Agentic AI Models shifted CPU:GPU Ratio to 1:1 toward 3-5:1 with Explosively Token-Hungry Workloads Agentic AI (autonomous, multi-step agents with planning, tool use, iteration, and self-correction) is fundamentally more compute and token intensive than conversational or single-turn generative AI. Agentic AI. autonomous, multi-step workflows with orchestration, tool use, parallel agents, data movement, and enterprise integration has dramatically increased the importance of strong host CPUs alongside GPUs. This shifts the CPU-to-GPU ratio higher and makes balanced systems critical toward 1:1 to 5:1 as enterprises testing more than 5-10 agents. AMD EPYC Venice excels ~Leadership core density (up to 256 Zen 6 cores per socket) for running many agents in parallel, orchestration layers, and high-throughput control-plane tasks. ~Superior performance-per-core and power efficiency ( up to 2.1x higher perf/core and 2.26x better SPECpower vs. NVIDIA Grace in benchmarks). ~Tight integration in Helios: One Venice CPU + multiple MI450 GPUs per node, enabling efficient data feeding to GPUs ("zero-copy"), parallel execution, and full rack utilization for complex agentic loops. Hyperscalers (Meta, Microsoft, Amazon, Google, Softbank) and AI natives (OpenAI, Anthropic...) are adopting high-core EPYC at scale specifically for these agentic demands, as CPUs now handle a larger share of non-model work (orchestration, policy enforcement, tool calls). This complements AMD’s lower-cost GPUs for overall TCO wins. ~Agents often generate 10–100x+ more tokens per task due to iterative reasoning chains, multiple tool calls, verification loops, and long-context orchestration. ~Goldman Sachs forecasts token consumption multiplying 24x by 2030 (to 120 quadrillion tokens/month) largely driven by agentic adoption in consumer and enterprise. ~Enterprise data shows agent-pattern workloads growing at 680% annualized rates, projected to surpass conversational AI in token volume by Q3 2026. ~Daily enterprise agent token consumption is already in the billions, with complex workflows (coding, workflows, analysis) amplifying this dramatically. 4. Competitive Edge: Winning Customers from Anthropic Anthropic’s Claude models (especially Opus/Sonnet) excel in complex reasoning and agentic coding, commanding premium positioning. However, their higher underlying costs (heavier reliance on third-party cloud with margins) limit pricing flexibility compared to OpenAI’s owned Helios capacity. Anthropic is on track to generate $10.9 billion in Q2 revenue. The company expects to achieve its first-ever quarterly adjusted operating profit of $559 million. However, sustaining full-year profitability remains challenging due to immense computing and model training costs The truth is, Anthropic has no choice but to buy as much $AMD chips as possible if they want to compete with OpenAI or get investors attention. This 5% adjusted operating profit to revenue ratio is just pathetic. Current pricing dynamics (2026): OpenAI already undercuts on many tiers ( flagship output tokens significantly cheaper than equivalent Claude Opus). Nano/mini models offer 5–10x advantages for volume work. Anthropic holds edges in long-context flat pricing and certain reasoning quality. OpenAI after Helios Rack Ownership, At $0.0003–$0.0005/M effective costs, OpenAI gains massive headroom to: ~Aggressively discount high-volume agentic tiers or bundles. ~Offer “unlimited” enterprise plans or usage-based models that Anthropic struggles to match without margin erosion. ~Target cost-sensitive, high-throughput agent deployments (dev tools, automation platforms) where token bills explode. Enterprises facing $ millions in monthly agentic bills will migrate to the provider delivering better economics at scale. OpenAI’s combination of strong models (o-series reasoning) + lowest TCO positions it to erode Anthropic’s enterprise share, especially as agentic becomes the dominant token consumer. Cheaper tokens expand the total addressable market dramatically. This feeds the data/model improvement loop, justifying further capex. AMD benefits from proven scale pulling in more customers (Meta, Oracle, Microsfot, Amazon, Softbank, TensorWave, LumaAI ... already aligned on Helios). Conclusion: Dr. Lisa Su has been laser focused on inference economics since at least 2022–2023, repeatedly emphasizing that the real battleground for AI scalability would be TCO, power efficiency (TDP), and ultimately tokens per dollar and per watt not just raw training FLOPS. While many viewed inference as a secondary, commoditized workload, Dr. Su architected AMD’s roadmap around rack-scale systems optimized for high-volume, sustained inference that would dominate as models matured and usage exploded. Helios represents the culmination of that multi-year bet: a fully integrated, open platform designed precisely for the economics of massive token throughput. This deep, strategic partnership with OpenAI starting with the 1GW Helios deployment in H2 2026 and scaling to 6GW, is the embodiment of that shared vision. Both companies foresaw a future where agentic AI models evolve to become extraordinarily token-hungry: autonomous agents executing complex, iterative workflows with planning, tool use, verification loops, and long-context reasoning. These workloads can consume 100x+ more tokens per task than traditional chat or single-turn generation, driving exponential demand as capabilities improve and enterprises deploy them at scale. By owning and optimizing this massive Helios capacity at GW scale, OpenAI achieves inference costs as low as $0.0003–$0.0005 per million tokens. This structural cost advantage allows OpenAI to absorb the coming token explosion profitably, dramatically lower effective pricing for enterprises, and win high-volume agentic workloads from higher-cost competitors like Anthropic. What was once a prohibitive monthly token bill becomes an affordable accelerator for productivity and innovation. The OpenAI-AMD alliance validates Dr. Su’s prescient strategy and turns the Agentic flywheel into reality: Collapsing inference costs → explosive token consumption → richer data and better models → accelerate greater demand. This partnership doesn’t just address today’s economics, it positions both leaders at the center of the infrastructure buildout that will power AI’s next decade. By delivering the lowest inference economics at scale, OpenAI not only solves enterprise bill pain but gains a decisive weapon to win share from higher-cost rivals like Anthropic. And that is why OpenAI and $META will deploy EPYC Dense Rack Not Financial Advice! DYOR! Research Purpose Only!

$AMD| The FOMO to buy AMD Chips is NOW 🧵 Not Financial Advice! DYOR! Research Purpose Only! The Inference Queen is the biggest winner in Agentic AI where all other CPUs are struggling to compete with a 2yr old EPYC Turin and EPYC Venice is in mass production phase. AMD stresses deployability today on standard x86 platforms (no proprietary architectures required), full software compatibility, and open standards. This positions Venice + Helios as a practical, high-density alternative to competing solutions while underscoring that agentic AI shifts the balance toward CPU-rich racks alongside GPUs, and most importantly, lowering the cost of token to accelerate adoption and innovation. Context: The Wall Street Journal yesterday came out with an article that OpenAI is condiering drasstically lowering the token prices to win more customers from Anthropic. The narrative "they" are trying to exacerbate the current AI selloff won't last long. This is a fundamental misunderstanding of what is going on, or what I already discussed for months and years. Followers and Subscribers already knew this for years, that this day would come, where token cost will bcome the central discussion among enterprises as there is no such thing as unlimited budget or Tokenmaxxing when they use $NVDA chips or In-house Hyperscalers chips. I will link various threads if you are interested in understanding the full picture from supply chain to recent TSMC Rapid 2nm expansion up to 12 Fabs total by 2027/2028. Hyperscalers and AI natives effectively have no choice but to buy more AMD system for Agentic AI as leadership in economical, power-aware, high-volume internal + agentic use. However, due to supply constraints where Supply is far behind Demand, this makes multi-vendor reality along with in-house chips drive faster industry progress, lower overall costs, and better sustainability. NVIDIA’s Vera Rubin cannot compete with a 2 years old EPYC Turin, but AMD under Dr. Lisa Su has engineered the lowest cost-per-million-tokens, highly competitive energy-efficient solutions, and superior CPU orchestration for agentic AI at scale with Helios. Dr. Su has championed this shift since at least 2023, foreseeing the rise of agentic workflows that demand far more orchestration, parallel agents, and balanced compute well before the industry fully embraced it. Her long-term vision of AI moving from simple prompts to always on, multi-agent systems has driven AMD’s investments in high-core EPYC CPUs and integrated rack-scale solutions, perfectly positioning the company for today’s realities. The OpenAI-AMD 1GW Helios deployment (starting H2 2026) represents a pivotal vertical integration move that directly supercharges the inference economics. This isn't incremental; it's a structural shift toward ownership of massive, optimized rack-scale capacity, enabling the lowest token costs and triggering the enterprise adoption flywheel. We need to be honest, $AMD is the only company that made a big bet on Inference since the day Chatgpt became sensational where $NVDA and others were betting big on Training. At the end of the day, Token bill from Anthropic has to obey economics. Meaning the bills rise, companies have to get more out of it to justify the cost. It cannot be an unlimited inference budget, and it has to show up on efficiency, profitability and operating leverage. 1. Tokenomics After you understand this, you will understand why Citi cited Anthropic is likely to sign a deal with $AMD along with Hyperscalers, AI Labs, Sovereign AI like Softbank 5GW in France and many other countries. However, OpenAI and $META are now wanting faster deployment, and they are AMD shareholders now, they have prioritized allocation. Anthropic and Hyperscalers just cannot compete when Helios Rack lower token cost to$0.0003–$0.0005 per million tokens at GW scale. Cost to build 1GW data center 1GW Helios Rack full build is estimated $30-$35B 1GW Rubin Rack full build is estimated $45-$55B Inference (Cost per Million Tokens) ~$NVDA B200 / HGX: ~$0.02–$0.08 on optimized workloads (FP4/MXFP4, speculative decoding). Significant improvement over Hopper but still premium-priced. GB200 NVL72 rack-scale: $0.05–$0.25+ ~$AMD Helios Racks: $0.0003-$0.0005 per M tokens, dramatically lower than NVIDIA equivalents in owned infra. MI355X node-level: Up to 40% more tokens per dollar vs. competing solutions ( B200), driven by higher memory capacity (up to 288GB+ HBM), strong bandwidth, and lower acquisition costs. Training ~$NVDA Rubin Rack is estimated $0.7-$1.2/M Tokens ~$AMD Helios Rack is estimated $0.65-$1.0/M Tokens Now, OpenAI, META and Hyperscalers can lower Inference cost even further with $AMD EPYC Venice "dense rack" or Agentic AI Rack. AMD published a detailed technical blog emphasizing that the future of agentic AI autonomous, multi-step AI systems requiring heavy orchestration, databases, caching, APIs, and control planes demands massive CPU-dense rack-scale infrastructure, not just GPUs. The catalyst prominently positions their upcoming 6th Gen EPYC "Venice" processors as the key enabler for next-generation dense racks, delivering leadership throughput under real-world power, cooling, and density constraints. ~EPYC Venice (Zen 6 architecture, up to 256 cores / 512 threads per socket) is projected to deliver exceptional rack-level performance. In AMD’s modeled 100 kW rack comparisons, Venice-powered systems are expected to achieve ~3.30x the throughput of NVIDIA’s Vera (88-core Olympus) baseline across a broad mix of agentic-supporting workloads. ~This builds on current-generation 5th Gen EPYC "Turin" (up to 192 cores), which already delivers ~2.37x rack throughput vs. Vera and ~1.6x vs. Intel’s Xeon 6980P (128 cores). ~ Liquid-cooled Turin deployments already support >27,000 CPU cores per rack today. Venice is architected to push this beyond 36,000 cores in the same rack class, dramatically increasing concurrent agent capacity and overall infrastructure efficiency. 2. Ownership vs renting compute from Hyperscalers matter to OpenAI and only owning $AMD chips can meaningfully lower token cost for enterprises. ~Eliminates cloud overhead: No provider margins, utilization buffers, or egress fees. Direct control over power contracts, cooling, scheduling, and orchestration at dedicated facilities. ~Helios optimizations at GW scale: Rack-level density (1.4+ exaFLOPS FP8 per rack), high HBM4 bandwidth, EPYC orchestration for agentic workloads, and superior TCO/TDP. AMD's long-standing focus on tokens per dollar/watt shines here 20-40%+ efficiency edges in inference-heavy scenarios. ~At 1GW+ optimized deployment, inference hits $0.0003–$0.0005 per million tokens (community/analyst models tied to Helios metrics). This is dramatically lower than typical rented/cloud equivalents, especially for high-volume output tokens in agentic flows. High token bills today, enterprises running heavy agentic/coding/analysis workloads can face $50-100M+/month at current API rates (flagship models $5-30+/M output, scaled to massive volumes). Post-Helios compression, same volume will drop to $10-15M/month (or better) via lower underlying costs passed through as pricing flexibility, volume tiers, caching, or batch discounts. ROI thresholds collapse. More companies greenlight pilots → production → massive scaling. Agentic AI (autonomous workflows) multiplies token demand exponentially, but affordability removes the friction. OpenAI gains flexibility, Unlike more cloud-dependent rivals (Anthropic), they can lower effective pricing, offer aggressive enterprise bundles, or absorb volume without margin destruction directly tackling "high token bill" complaints while maintaining profitability as usage explodes. 3. Agentic AI Models shifted CPU:GPU Ratio to 1:1 toward 3-5:1 with Explosively Token-Hungry Workloads Agentic AI (autonomous, multi-step agents with planning, tool use, iteration, and self-correction) is fundamentally more compute and token intensive than conversational or single-turn generative AI. Agentic AI. autonomous, multi-step workflows with orchestration, tool use, parallel agents, data movement, and enterprise integration has dramatically increased the importance of strong host CPUs alongside GPUs. This shifts the CPU-to-GPU ratio higher and makes balanced systems critical toward 1:1 to 5:1 as enterprises testing more than 5-10 agents. AMD EPYC Venice excels ~Leadership core density (up to 256 Zen 6 cores per socket) for running many agents in parallel, orchestration layers, and high-throughput control-plane tasks. ~Superior performance-per-core and power efficiency ( up to 2.1x higher perf/core and 2.26x better SPECpower vs. NVIDIA Grace in benchmarks). ~Tight integration in Helios: One Venice CPU + multiple MI450 GPUs per node, enabling efficient data feeding to GPUs ("zero-copy"), parallel execution, and full rack utilization for complex agentic loops. Hyperscalers (Meta, Microsoft, Amazon, Google, Softbank) and AI natives (OpenAI, Anthropic...) are adopting high-core EPYC at scale specifically for these agentic demands, as CPUs now handle a larger share of non-model work (orchestration, policy enforcement, tool calls). This complements AMD’s lower-cost GPUs for overall TCO wins. ~Agents often generate 10–100x+ more tokens per task due to iterative reasoning chains, multiple tool calls, verification loops, and long-context orchestration. ~Goldman Sachs forecasts token consumption multiplying 24x by 2030 (to 120 quadrillion tokens/month) largely driven by agentic adoption in consumer and enterprise. ~Enterprise data shows agent-pattern workloads growing at 680% annualized rates, projected to surpass conversational AI in token volume by Q3 2026. ~Daily enterprise agent token consumption is already in the billions, with complex workflows (coding, workflows, analysis) amplifying this dramatically. 4. Competitive Edge: Winning Customers from Anthropic Anthropic’s Claude models (especially Opus/Sonnet) excel in complex reasoning and agentic coding, commanding premium positioning. However, their higher underlying costs (heavier reliance on third-party cloud with margins) limit pricing flexibility compared to OpenAI’s owned Helios capacity. Anthropic is on track to generate $10.9 billion in Q2 revenue. The company expects to achieve its first-ever quarterly adjusted operating profit of $559 million. However, sustaining full-year profitability remains challenging due to immense computing and model training costs The truth is, Anthropic has no choice but to buy as much $AMD chips as possible if they want to compete with OpenAI or get investors attention. This 5% adjusted operating profit to revenue ratio is just pathetic. Current pricing dynamics (2026): OpenAI already undercuts on many tiers ( flagship output tokens significantly cheaper than equivalent Claude Opus). Nano/mini models offer 5–10x advantages for volume work. Anthropic holds edges in long-context flat pricing and certain reasoning quality. OpenAI after Helios Rack Ownership, At $0.0003–$0.0005/M effective costs, OpenAI gains massive headroom to: ~Aggressively discount high-volume agentic tiers or bundles. ~Offer “unlimited” enterprise plans or usage-based models that Anthropic struggles to match without margin erosion. ~Target cost-sensitive, high-throughput agent deployments (dev tools, automation platforms) where token bills explode. Enterprises facing $ millions in monthly agentic bills will migrate to the provider delivering better economics at scale. OpenAI’s combination of strong models (o-series reasoning) + lowest TCO positions it to erode Anthropic’s enterprise share, especially as agentic becomes the dominant token consumer. Cheaper tokens expand the total addressable market dramatically. This feeds the data/model improvement loop, justifying further capex. AMD benefits from proven scale pulling in more customers (Meta, Oracle, Microsfot, Amazon, Softbank, TensorWave, LumaAI ... already aligned on Helios). Conclusion: Dr. Lisa Su has been laser focused on inference economics since at least 2022–2023, repeatedly emphasizing that the real battleground for AI scalability would be TCO, power efficiency (TDP), and ultimately tokens per dollar and per watt not just raw training FLOPS. While many viewed inference as a secondary, commoditized workload, Dr. Su architected AMD’s roadmap around rack-scale systems optimized for high-volume, sustained inference that would dominate as models matured and usage exploded. Helios represents the culmination of that multi-year bet: a fully integrated, open platform designed precisely for the economics of massive token throughput. This deep, strategic partnership with OpenAI starting with the 1GW Helios deployment in H2 2026 and scaling to 6GW, is the embodiment of that shared vision. Both companies foresaw a future where agentic AI models evolve to become extraordinarily token-hungry: autonomous agents executing complex, iterative workflows with planning, tool use, verification loops, and long-context reasoning. These workloads can consume 100x+ more tokens per task than traditional chat or single-turn generation, driving exponential demand as capabilities improve and enterprises deploy them at scale. By owning and optimizing this massive Helios capacity at GW scale, OpenAI achieves inference costs as low as $0.0003–$0.0005 per million tokens. This structural cost advantage allows OpenAI to absorb the coming token explosion profitably, dramatically lower effective pricing for enterprises, and win high-volume agentic workloads from higher-cost competitors like Anthropic. What was once a prohibitive monthly token bill becomes an affordable accelerator for productivity and innovation. The OpenAI-AMD alliance validates Dr. Su’s prescient strategy and turns the Agentic flywheel into reality: Collapsing inference costs → explosive token consumption → richer data and better models → accelerate greater demand. This partnership doesn’t just address today’s economics, it positions both leaders at the center of the infrastructure buildout that will power AI’s next decade. By delivering the lowest inference economics at scale, OpenAI not only solves enterprise bill pain but gains a decisive weapon to win share from higher-cost rivals like Anthropic. And that is why OpenAI and $META will deploy EPYC Dense Rack Not Financial Advice! DYOR! Research Purpose Only!

Mike

84,951 просмотров • 1 месяц назад

$AMD $5 Trillion MC Is Inevitable Long Term👑 This thread will focus more on Inference! 2026 EPYC "Venice" $TSM 2nm to save Large GW Scale Inference by 40% more than Prior Turin gen. Context: EPYC Turin achieves ~$0.001 per million tokens for batch inference vs $0.02-$0.12/ million tokens as I wrote the thread below. Venice is going to lower cost down to $0.0005-$0.0006/Million Tokens. OpenAI spent roughly $20B on Inference and Training, where 80-90% of that was for Inference per Analysts. AKA Renting Compute is Expensive AF! In this thread, I want to focus on why most analysts and investors are underestimating the role EPYC "Venice" and future Gen on overall Data center revenue. And $TSM ramping up 2nm supply early is a confirmation that AMD will be a major buyer long term. I will also link the thread the Gap between AMD Analysts & Reality and 2nm Ramp Thread so you have more comprehensive view of what I'm writing here. Before I go into detail this is my 2026 Projection: AI GPUs: $35-$50B EPYC Data Center: $15B-$17B Client Segment: $12-$13B Gaming: $6B Embedded: $4B-$5B Total Revenue $70-$100B Non-GAAP net income $18B-$25B Non-GAAP EPS $10.97-$15.40 Foward P/E 55x-70x= $603-$1,078 AMD's Analysts are projecting $0 Revenue for MI450 and sluggish EPYC Growth. Meaning, all analysts are either full of 💩 or Sexist, you decide! Analysts are also projecting 0% growth on AMD "Secret Weapon" Chip as $MSFT said we are at significant Windows refresh and upgrade cycle. Do you think TSMC would allocate more 2nm supply to $AMD at $0 MI450 revenue and sluggish EPYC? 1. EPYC is going to be the leader in lowest Inference! Current Turin cost saving is 95% vs $NVDA or 98-99% on Inference cost when you factor in renting Inference compute from Amazon Web Services, Microsoft Azure, or $NVDA Neocloud pets. TSMC claimed: 10-15% higher performance at iso-power, 25-30% lower power at iso-speed, and ~15% higher transistor density compared to 3nm. This reduces operational expenses (energy, cooling) while increasing throughput per chip. EPYC Turin achieves ~$0.001 per million tokens for batch inference (via vLLM on models like Llama 3 70B), driven by high core counts and low hardware costs. EPYC Venice offers ~1.7x overall performance and up to 70% more compute capability per core, with up to 256 cores (512 threads). Enhanced vector/AI instructions and open-source firmware (openSIL) optimize for inference workloads. AMD Incorporates AI Engines (now part of AMD's XDNA) for on-chip acceleration, improving efficiency for low-latency and edge inference. This reduces reliance on discrete GPUs, lowering system complexity and TCO. Venice SKUs are projected at $3,000-$15,000 ($5,000 for 256-core flagship), far below NVIDIA Rubin ($50,000-$90,000) or AMD's own MI450 GPUs ($40,000-$50,000). High memory bandwidth (up to 1.6 TB/s) supports efficient batch inference. Venice is designed exactly for Large customers that want to lower Inference Cost and MI450 Helios is for Customers that want Training at lowest TCO, TDP as well as lower Upfront 1GW scale(Full build $35-$40B vs $NVDA $55B-$80B). 2. Real World Example: OpenAI's 2025 inference spend reached ~$20B, escalating to even higher total compute rental (mostly inference) amid token volume growth(from video generating). By 2026, with usage doubling (consistent with industry trends: token demand grows 2-5x YoY), assume OpenAI processes ~1,800 billion million-tokens annually $NVDA Blackwell at $0.02-$0.12 is $36B(most optimized) Rubin is projected to be at $0.01/million tokens or $18B annual Inference Cost vs $AMD Venice $0.0005/million tokens or $0.9B annual Inference Cost => Massive saving for OpenAI or anyone that are paying 80-90% Annual Bill for Inference compute. In short, it is unsustainable to pay this much rent vs owning for all current AI players for the medium to long term. Rubin excels in low-latency decode (if Groq integration from $20B deal in 2027-2028), but Venice dominates batch (80% of inference by 2030). Actual savings depend on deployment scale (OpenAI's 6GW AMD plans), electricity rates, and software maturity. If Rubin only hits $0.03, savings swell to $53.1B vs. $17.1B. 3. Will running Inference on Venice and future Gen slow down response generation in 2026 and beyond? Human perception of "fast enough" for chat, agents, search augmentation, summarization, coding assistance is roughly Meaning, EPYC may generate $100B a year on data center revenue, Hence $MSFT $AMZN $META $GOOGL OpenAI xAI and 42+ Countries are leaning AMD for Inference, because the cost saving is MASSIVE! 4. Regular users (you, me, people using ChatGPT, Claude, Gemini, Grok, Perplexity...) are extremely unlikely to notice any slowdown and in many cases might even experience slightly faster or more consistent response times if the industry heavily shifts toward AMD EPYC for inference. What actually happens when companies save massively on inference? When OpenAI , Anthropic , Gemini , Grok Meta .... save billions on the batch/enterprise/RAG layer using EPYC Venice, they typically do one or more of these things with the savings, none of which make your chat slower but enhancing their bottom line(Profit) ~Keep prices the same → make more profit ~Lower subscription prices / increase free tier limits ~Train bigger & better models more frequently ~Offer longer context windows ~Add more reasoning steps / tool calls / agents per query ~Improve multimodal capabilities ~Build more data centers / reduce throttling during peaks In practice the consumer experience usually gets better, not worse, when inference becomes dramatically cheaper. Prime example is $META leaning AMD heavily or currently AMD largest customer. or Grok 2 to Grok 3 heavily used AMD for Inference saving. And most Grok Users reported Groke responses snappier, not slower. 5. What does this mean for potential Revenue? Noted that TSMC is massively ramping 2nm supply for $AMD both MI450 and EPYC. EPYC Conservative projection: FY2025: $10.5B(best Est) FY2026: $16B FY2027: $29B FY2028: $49B FY2029: $75B FY2030: $100B Large customers: $META OpenAI $MSFT $AMZN $GOOGL xAI (Apple?) Smaller customer: $DELL $HPE $SMCI and 42+ other countries. The roadmap to $5 Trillion is very much inevitable as Inference Cost from Renting or owning $NVDA are too high, but $NVDA will still dominate Training market share, where MI families are likely to take 15-20% market share, but the TAM is also expanding Rapidly. Most Institutions are projecting $2-$3Trillion TAM by 2030. $NVDA said $4 Trillion. Dr. Lisa Su said $1 Trillion+ by 2030. So you decide on how much TAM. If you enjoy this kind of analysis, Slap the Like/Repost and Bookmark to please the X Algo as it is Free.99! If you want to support my work further, consider subscribe to see more in-depth analysis! Alright, that is it. Not Financial Advice!

$AMD $5 Trillion MC Is Inevitable Long Term👑 This thread will focus more on Inference! 2026 EPYC "Venice" $TSM 2nm to save Large GW Scale Inference by 40% more than Prior Turin gen. Context: EPYC Turin achieves ~$0.001 per million tokens for batch inference vs $0.02-$0.12/ million tokens as I wrote the thread below. Venice is going to lower cost down to $0.0005-$0.0006/Million Tokens. OpenAI spent roughly $20B on Inference and Training, where 80-90% of that was for Inference per Analysts. AKA Renting Compute is Expensive AF! In this thread, I want to focus on why most analysts and investors are underestimating the role EPYC "Venice" and future Gen on overall Data center revenue. And $TSM ramping up 2nm supply early is a confirmation that AMD will be a major buyer long term. I will also link the thread the Gap between AMD Analysts & Reality and 2nm Ramp Thread so you have more comprehensive view of what I'm writing here. Before I go into detail this is my 2026 Projection: AI GPUs: $35-$50B EPYC Data Center: $15B-$17B Client Segment: $12-$13B Gaming: $6B Embedded: $4B-$5B Total Revenue $70-$100B Non-GAAP net income $18B-$25B Non-GAAP EPS $10.97-$15.40 Foward P/E 55x-70x= $603-$1,078 AMD's Analysts are projecting $0 Revenue for MI450 and sluggish EPYC Growth. Meaning, all analysts are either full of 💩 or Sexist, you decide! Analysts are also projecting 0% growth on AMD "Secret Weapon" Chip as $MSFT said we are at significant Windows refresh and upgrade cycle. Do you think TSMC would allocate more 2nm supply to $AMD at $0 MI450 revenue and sluggish EPYC? 1. EPYC is going to be the leader in lowest Inference! Current Turin cost saving is 95% vs $NVDA or 98-99% on Inference cost when you factor in renting Inference compute from Amazon Web Services, Microsoft Azure, or $NVDA Neocloud pets. TSMC claimed: 10-15% higher performance at iso-power, 25-30% lower power at iso-speed, and ~15% higher transistor density compared to 3nm. This reduces operational expenses (energy, cooling) while increasing throughput per chip. EPYC Turin achieves ~$0.001 per million tokens for batch inference (via vLLM on models like Llama 3 70B), driven by high core counts and low hardware costs. EPYC Venice offers ~1.7x overall performance and up to 70% more compute capability per core, with up to 256 cores (512 threads). Enhanced vector/AI instructions and open-source firmware (openSIL) optimize for inference workloads. AMD Incorporates AI Engines (now part of AMD's XDNA) for on-chip acceleration, improving efficiency for low-latency and edge inference. This reduces reliance on discrete GPUs, lowering system complexity and TCO. Venice SKUs are projected at $3,000-$15,000 ($5,000 for 256-core flagship), far below NVIDIA Rubin ($50,000-$90,000) or AMD's own MI450 GPUs ($40,000-$50,000). High memory bandwidth (up to 1.6 TB/s) supports efficient batch inference. Venice is designed exactly for Large customers that want to lower Inference Cost and MI450 Helios is for Customers that want Training at lowest TCO, TDP as well as lower Upfront 1GW scale(Full build $35-$40B vs $NVDA $55B-$80B). 2. Real World Example: OpenAI's 2025 inference spend reached ~$20B, escalating to even higher total compute rental (mostly inference) amid token volume growth(from video generating). By 2026, with usage doubling (consistent with industry trends: token demand grows 2-5x YoY), assume OpenAI processes ~1,800 billion million-tokens annually $NVDA Blackwell at $0.02-$0.12 is $36B(most optimized) Rubin is projected to be at $0.01/million tokens or $18B annual Inference Cost vs $AMD Venice $0.0005/million tokens or $0.9B annual Inference Cost => Massive saving for OpenAI or anyone that are paying 80-90% Annual Bill for Inference compute. In short, it is unsustainable to pay this much rent vs owning for all current AI players for the medium to long term. Rubin excels in low-latency decode (if Groq integration from $20B deal in 2027-2028), but Venice dominates batch (80% of inference by 2030). Actual savings depend on deployment scale (OpenAI's 6GW AMD plans), electricity rates, and software maturity. If Rubin only hits $0.03, savings swell to $53.1B vs. $17.1B. 3. Will running Inference on Venice and future Gen slow down response generation in 2026 and beyond? Human perception of "fast enough" for chat, agents, search augmentation, summarization, coding assistance is roughly Meaning, EPYC may generate $100B a year on data center revenue, Hence $MSFT $AMZN $META $GOOGL OpenAI xAI and 42+ Countries are leaning AMD for Inference, because the cost saving is MASSIVE! 4. Regular users (you, me, people using ChatGPT, Claude, Gemini, Grok, Perplexity...) are extremely unlikely to notice any slowdown and in many cases might even experience slightly faster or more consistent response times if the industry heavily shifts toward AMD EPYC for inference. What actually happens when companies save massively on inference? When OpenAI , Anthropic , Gemini , Grok Meta .... save billions on the batch/enterprise/RAG layer using EPYC Venice, they typically do one or more of these things with the savings, none of which make your chat slower but enhancing their bottom line(Profit) ~Keep prices the same → make more profit ~Lower subscription prices / increase free tier limits ~Train bigger & better models more frequently ~Offer longer context windows ~Add more reasoning steps / tool calls / agents per query ~Improve multimodal capabilities ~Build more data centers / reduce throttling during peaks In practice the consumer experience usually gets better, not worse, when inference becomes dramatically cheaper. Prime example is $META leaning AMD heavily or currently AMD largest customer. or Grok 2 to Grok 3 heavily used AMD for Inference saving. And most Grok Users reported Groke responses snappier, not slower. 5. What does this mean for potential Revenue? Noted that TSMC is massively ramping 2nm supply for $AMD both MI450 and EPYC. EPYC Conservative projection: FY2025: $10.5B(best Est) FY2026: $16B FY2027: $29B FY2028: $49B FY2029: $75B FY2030: $100B Large customers: $META OpenAI $MSFT $AMZN $GOOGL xAI (Apple?) Smaller customer: $DELL $HPE $SMCI and 42+ other countries. The roadmap to $5 Trillion is very much inevitable as Inference Cost from Renting or owning $NVDA are too high, but $NVDA will still dominate Training market share, where MI families are likely to take 15-20% market share, but the TAM is also expanding Rapidly. Most Institutions are projecting $2-$3Trillion TAM by 2030. $NVDA said $4 Trillion. Dr. Lisa Su said $1 Trillion+ by 2030. So you decide on how much TAM. If you enjoy this kind of analysis, Slap the Like/Repost and Bookmark to please the X Algo as it is Free.99! If you want to support my work further, consider subscribe to see more in-depth analysis! Alright, that is it. Not Financial Advice!

Mike

102,223 просмотров • 6 месяцев назад

$NVDA $GFS NVIDIA’s reported agreement to acquire Groq for $20B in cash (per CNBC, amplified via Reuters and other wire coverage) represents a materially different strategic posture than NVIDIA’s prior M&A pattern, given both the headline size (largest reported NVIDIA acquisition to date) and the unusual carve-out that Groq’s early-stage cloud business would not be included. Public reporting indicates the information originated from Alex Davis, CEO of Disruptive (lead investor in Groq’s latest financing), and that neither NVIDIA nor Groq had issued an immediate confirmation at the time of publication. The same reporting frames the transaction as coming together quickly, only months after Groq raised $750M at a ~$6.9B valuation, and highlights Groq’s positioning as a high-performance inference chip vendor founded by ex-Google TPU engineers. Groq is best understood as a vertically integrated inference acceleration company whose core asset is an application-specific processor optimized for deterministic, low-latency execution of transformer-style workloads, paired with a compiler-led software stack and a distribution layer (GroqCloud) designed to reduce developer friction via OpenAI-compatible APIs and integrations. Groq brands its architecture as a Language Processing Unit (LPU) and consistently emphasizes that the design target is inference, not training. The company’s own architecture description centers on 1-core execution, large on-chip SRAM used as primary storage (explicitly not cache), a custom compiler that statically schedules compute and communication, and direct chip-to-chip connectivity intended to coordinate multi-chip execution without relying on conventional caching hierarchies or dynamic runtime scheduling. The technical premise is a deliberate inversion of the conventional GPU approach. GPUs deliver throughput via massively parallel, multi-core execution with dynamic scheduling, complex memory hierarchies, and heavy reliance on off-chip HBM bandwidth and sophisticated runtime/kernel optimization. Groq instead argues that inference bottlenecks are driven by latency variance (tail latency), synchronization overhead, and memory access unpredictability inherent in dynamically scheduled, cache-heavy architectures, particularly when workloads are latency sensitive and batch sizes cannot be inflated. Groq’s solution is to move “control” into the compiler: the full execution graph and inter-chip communication schedule are computed ahead of time down to clock-cycle granularity, with deterministic execution designed to reduce run-to-run variance. In Groq’s framing, the removal of caches, reorder buffers, speculative execution overhead, and other sources of contention enables predictable latency and high utilization without per-model kernel engineering typical of GPU tuning cycles. A critical nuance is that Groq’s determinism is not merely a software claim; it is tightly coupled to architectural constraints and system design choices that trade flexibility for predictability. Third-party technical commentary indicates Groq’s chip uses a fully deterministic VLIW-style approach with minimal buffering, no external memory, and heavy dependence on sharding models across many chips because on-chip SRAM capacity is limited. SemiAnalysis describes a ~725 mm^2 die on GlobalFoundries 14nm with ~230MB of SRAM and notes that “no useful models” fit on a single chip, forcing multi-chip partitioning for modern LLMs and driving a system-level design where networking and compilation are first-class scheduling problems rather than ancillary infrastructure. This is consistent with Groq’s own messaging that tensor parallelism across chips is a primary design goal, enabled by large on-chip SRAM and compile-time coordination of compute plus interconnect. The on-chip SRAM emphasis is central to Groq’s latency story and also its most constraining trade-off. Groq claims on-chip SRAM bandwidth “upwards of 80 TB/s” and contrasts that with off-chip HBM bandwidth “about 8 TB/s,” asserting a potential 10x advantage from bandwidth plus reduced trips across chip-to-memory boundaries. While these comparisons are marketing-oriented and depend on workload specifics, the architectural implication is clear: Groq prioritizes ultra-fast local weight/activation access and then scales capacity by adding chips, not by attaching large off-chip memory pools. This design can reduce latency for sequential inference layers and minimize unpredictable stalls, but it pushes complexity into partitioning strategy, interconnect topology, and compiler scheduling, and it increases the number of chips needed for very large parameter counts and large KV-cache footprints. Groq also highlights numeric formats and compiler-driven precision management as a performance lever. In its 2025 technical blog, Groq describes “TruePoint numerics,” including 100-bit intermediate accumulation and selective quantization choices (FP32 for attention-sensitive operations, block floating point for MoE weights, FP8 storage in error-tolerant layers), and claims 2-4x speedups versus BF16 without measurable accuracy degradation on benchmarks such as MMLU and HumanEval. Even if the absolute uplift is workload dependent, the strategic point is that Groq is pursuing performance via end-to-end co-design: precision policy is not just hardware capability (FP8/BF16) but compiler-enforced mapping of precision to error sensitivity, which can matter materially for inference cost-per-token if it reduces memory traffic and boosts throughput without forcing aggressive, accuracy-damaging quantization. Independent performance datapoints indicate Groq has been credible on latency-oriented inference speed, at least for certain regimes. EE Times reported in 2023 that Groq demonstrated Llama-2 70B inference at ~240 tokens/s per user on a cloud-based dev system described as 10 racks and 64 chips, using the company’s 1st-gen silicon introduced several years earlier. Separate Groq commentary around independent benchmarking cites results showing ~241 tokens/s throughput and ~0.8s time to receive 100 output tokens for a Llama-2 70B API configuration, positioning the platform as a step-change in “available speed” for certain interactive use cases. These figures do not settle total cost-of-ownership versus GPUs or hyperscaler ASICs, but they establish that Groq’s system-level architecture can deliver strong single-user throughput and latency on large models when properly partitioned and scheduled. GroqCloud is the commercial wrapper that packages this hardware/software stack as “tokens-as-a-service,” aiming to make Groq adoption feel like switching API endpoints rather than adopting new silicon. Groq’s documentation states its API is designed to be “mostly compatible” with OpenAI client libraries, and its pricing page provides model-specific token rates, published speeds (tokens/s), prompt caching discounts, and batch processing discounts. For example, pricing lists inputs as low as $0.05 per 1M tokens and outputs as low as $0.08 per 1M tokens for certain smaller LLM configurations, with higher prices for larger models and long-context or MoE variants; it also advertises prompt caching with a 50% discount on cached input tokens for certain models and a batch API offering 50% lower cost for asynchronous processing windows. These mechanics are economically important because they demonstrate Groq’s go-to-market is not simply “sell chips,” but “sell predictable unit economics per token,” with tooling (batch, caching) that directly targets inference cost drivers (reused prompts, throughput smoothing, and asynchronous workloads). The cloud footprint and distribution partnerships indicate Groq has been building an inference-native “edge within the cloud” strategy rather than competing head-on with hyperscalers on breadth of services. A 2025 Groq newsroom release describes a European deployment in Helsinki with Equinix, positioned as latency reduction and data governance for European customers, and explicitly references Equinix Fabric enabling private connectivity to GroqCloud over public, private, or sovereign infrastructure. The same release enumerates additional capacity in the U.S. (Equinix, DataBank), Canada (Bell Canada), and Saudi Arabia (HUMAIN), and states these sites collectively served more than 20M tokens/s across Groq’s global network at that time. That supply-side metric matters because it provides a directional sense that Groq is scaling capacity as a network, not merely as a chip vendor. Customer disclosure is inherently limited because Groq is private and many enterprise deployments are not public, but Groq’s marketing materials and partnerships provide signals about demand vectors. The company’s public website displays logos of large consumer and enterprise brands (e.g., Dropbox, Vercel, Chevron, Volkswagen, Canva, Robinhood, Riot Games, Workday, Ramp) and includes a published customer quote claiming a 7.41x chat speed increase and an 89% cost reduction after moving to GroqCloud, followed by a tripling of token consumption. While marketing claims should be treated as case-specific and not generalized, they indicate that Groq is targeting both AI-native developers (who measure success by latency and cost-per-token) and enterprise buyers (who care about predictable performance and governance). Supplier and dependency mapping for Groq spans 3 layers: silicon production, system integration, and cloud infrastructure. On silicon, third-party analysis indicates GlobalFoundries 14nm for the 1st-gen Groq chip, implying a supply chain less constrained by the most capacity-tight leading-edge nodes and advanced packaging bottlenecks that dominate high-end GPU supply (HBM stacks, CoWoS-type packaging constraints). If accurate, this is strategically meaningful because it suggests Groq capacity expansion could be gated more by conventional wafer supply, board assembly, and data center power than by the same HBM/advanced packaging scarcity that has constrained top-tier GPU ramp cycles. On systems and cloud, Groq’s own releases identify colocation and connectivity partners (Equinix, DataBank, Bell Canada) and a Middle East partner (HUMAIN), implying dependencies on data center real estate, power availability, and network connectivity, alongside procurement of standard server components, NICs/switching, racks, and cooling infrastructure. The Groq design narrative also emphasizes air cooling and reduced need for complex power/cooling infrastructure, which—if realized in deployments—can widen the set of feasible hosting locations and lower deployment friction relative to liquid-cooled, very high power density GPU racks. Against that backdrop, the strategic rationale for NVIDIA acquiring Groq can be framed as a set of overlapping objectives: inference silicon optionality, architectural hedging, competitive defense, and supply chain diversification, with the carve-out of GroqCloud signaling a preference to avoid direct cloud competition and to focus on IP and product portfolio control rather than operating a capital-intensive token-serving business. The deal, if confirmed, would occur at a valuation step-up of ~190% versus Groq’s reported ~$6.9B private valuation in the September $750M round, reinforcing that any acquisition logic would be predominantly strategic rather than a conventional financial multiple arbitrage. The most compelling strategic driver is inference. Training has historically been the center of gravity for cutting-edge GPU demand, but inference volume is structurally larger and more distributed as deployments scale, with economics dominated by cost-per-token, latency guarantees, and utilization under spiky demand. Inference workloads also create a strategic vulnerability for NVIDIA: hyperscalers and large platforms can justify bespoke ASICs (TPU, Trainium/Inferentia, Maia-class efforts) because inference is stable, repeatable, and can amortize software investment at massive scale. Groq’s core proposition—deterministic, compiler-scheduled inference with predictable latency—aligns directly with the segment where GPU generality is least valued and where “good enough” programmability plus superior unit economics can win share. Acquiring Groq would allow NVIDIA to own a credible inference-native architecture rather than relying solely on GPUs and software optimization to defend that segment. Competitive defense logic is also plausible. Groq occupies a specific competitive wedge: low-latency, high-throughput interactive inference, delivered via a simple API abstraction that reduces switching cost. That wedge directly pressures GPU inference margins in the long run because it makes inference price/performance comparisons more transparent at the token level, and it targets a developer persona that historically defaulted to CUDA-first ecosystems. Even if NVIDIA’s current-generation systems can achieve very high tokens/s per user with extensive optimization, the strategic risk is that competing architectures normalize the idea that inference is best served by special-purpose silicon with a simpler programming model, weakening CUDA lock-in at the application layer. NVIDIA has actively demonstrated that Blackwell-era systems can exceed 1,000 tokens/s per user in benchmarked configurations, but that performance leadership does not automatically translate to lowest cost-per-token across the full range of batch sizes, latency targets, and deployment environments. Groq’s existence as a credible alternative architecture forces NVIDIA to keep defending inference economics rather than only raw performance leadership. The “technology acquisition” rationale is unusually strong in this specific case because Groq’s differentiator is not a single block of silicon IP but an end-to-end methodology: compiler-led static scheduling, deterministic networking, and a system architecture designed around tensor-parallel inference rather than throughput-maximizing batch inference. NVIDIA’s stack is already compiler-heavy (TensorRT, Triton, CUDA graphs, kernel fusion, speculative decoding techniques), but GPUs remain dynamically scheduled devices with complex memory hierarchies and stochastic latency behaviors under contention. Groq’s approach provides an alternate design point: treating the entire inference execution (compute plus communication) as a statically schedulable program. In principle, that IP could be valuable even if Groq silicon itself is not adopted at massive scale, because it can inform how NVIDIA builds future inference-optimized products, compilers, and networking fabrics, especially as distributed inference with large models makes communication a first-order performance determinant. Supply chain diversification is a non-obvious but potentially important driver. If Groq’s mainstream product generation is truly based on a mature process node and avoids HBM, then the scaling constraints look different than those of state-of-the-art GPUs. NVIDIA’s ability to meet incremental demand has been tightly coupled to advanced packaging and HBM supply, and those constraints can remain binding even when wafer supply is available. An inference ASIC architecture that relies primarily on on-chip SRAM and scales by adding chips—while not costless—could reduce dependence on HBM availability and advanced packaging capacity, enabling NVIDIA to ship “inference capacity” in higher absolute volumes or into geographies and customer segments where the highest-end GPUs are economically or logistically difficult to deploy. This could be particularly relevant for latency-sensitive inference deployed in regional colocation footprints rather than centralized hyperscale campuses. The carve-out of GroqCloud, if accurate, is itself a strategic signal about NVIDIA’s priorities. Operating a token-serving cloud at scale is capital intensive, structurally lower margin than silicon IP rents, and creates channel conflict with hyperscalers and CSP partners who are core NVIDIA customers. NVIDIA has generally positioned its cloud offerings through partnerships rather than as a direct hyperscale competitor. Excluding GroqCloud would preserve neutrality with CSPs and avoid inheriting multi-region data residency obligations and partner contracts, while still allowing NVIDIA to acquire Groq’s silicon, compiler technology, and engineering talent. At the same time, excluding GroqCloud would also mean NVIDIA would not automatically acquire the commercial proof-point of Groq’s unit economics or the customer contracts that validate product-market fit at scale, increasing the importance of diligence on whether Groq’s cloud pricing is structurally profitable or partially subsidized by fundraising. There is also a “preemptive acquisition” angle. The reporting identifies recent investors in Groq’s latest round including large financial institutions and strategic/industry players. In that context, Groq represents an asset that could plausibly have been acquired by a competitor (AMD/Intel) or by a hyperscaler seeking to accelerate inference independence. NVIDIA acquiring Groq could be a defensive move to prevent a credible inference-native architecture from being weaponized by a rival with deep distribution. Even if GroqCloud is carved out, controlling the silicon roadmap and compiler IP would meaningfully constrain Groq’s ability to evolve into a standalone competitor, unless the carved-out entity retains long-term rights to the hardware and software stack. However, the strategic case is not one-sided; there are meaningful risks and potential contradictions that would need to be reconciled for the transaction to be value-accretive on a multi-year horizon. 1st, Groq’s architecture appears to rely on scaling out chip count to achieve capacity, which introduces system cost, networking complexity, and physical footprint considerations. The absence of external memory and limited on-chip SRAM implies very large models require substantial chip parallelism, and the economics then depend heavily on chip cost, yield, power efficiency, and interconnect overhead. SemiAnalysis explicitly frames Groq as trading space for time and raises questions about token economics and whether publicly advertised pricing reflects fully loaded costs or market share capture. 2nd, integration risk is non-trivial. Groq’s compiler-led deterministic model is philosophically and practically different from CUDA’s dominant programming and execution model. A poorly executed integration could create internal product confusion, dilute engineering focus, or alienate developers if the combined stack fragments. 3rd, there is cannibalization risk. If Groq-class inference silicon undercuts GPU inference economics, NVIDIA could face internal margin trade-offs, even if the goal is to defend share against hyperscaler ASICs. Cannibalization can still be rational if it prevents larger share loss, but it would require crisp portfolio segmentation and go-to-market discipline. The presence of NVIDIA’s own rapidly improving inference performance complicates the “need” for Groq but does not eliminate the “option value.” NVIDIA has demonstrated benchmark-leading tokens/s per user on Blackwell-based systems, suggesting that raw interactive throughput is not necessarily the limiting factor for NVIDIA’s product line. The more enduring strategic question is unit economics and architectural control: whether future inference demand is better monetized through general-purpose GPUs plus software optimization, or whether a bifurcated product portfolio (training GPUs plus inference-native ASICs) becomes necessary to defend total AI compute wallet share as hyperscaler ASIC penetration increases. Acquiring Groq could be a decisive move to ensure NVIDIA participates in both regimes rather than betting exclusively on GPUs to win inference forever. What is “special” about Groq’s technology relative to a typical accelerator roadmap is the tight coupling of determinism, compilation, and networking into a single scheduling problem. The LPU narrative emphasizes deterministic compute and networking, static scheduling, and direct chip-to-chip coordination that allows “hundreds” (more precisely, 100s) of chips to behave like a single scheduled resource. The architecture also explicitly targets tensor-parallel, latency-optimized distribution rather than pure data-parallel throughput scaling, which matters for real-time applications where a single response must arrive quickly rather than many requests being processed in bulk. The implication is that Groq is optimized for the time-to-first-token and steady token streaming behavior that defines user experience in interactive LLMs, and it attempts to achieve that without relying on large batch sizes that can degrade latency. From a portfolio manager’s perspective, the most important interpretation is that an NVIDIA-Groq combination would likely be less about “NVIDIA needs more inference speed” and more about controlling the architectural trajectory of inference acceleration and removing a fast-improving, developer-friendly competitor from the market. The carve-out of GroqCloud would reinforce that the transaction is aimed at IP, talent, and product optionality, not acquiring a cloud revenue stream. The valuation step-up implied by $20B versus $6.9B would therefore be justified only if the acquired assets materially reduce long-term competitive risk (hyperscaler ASIC displacement, inference margin compression) or enable new monetization vectors (inference ASIC product line, supply chain de-bottlenecking, improved software determinism) that would be difficult to achieve on a comparable timeline via internal R&D.

$NVDA $GFS NVIDIA’s reported agreement to acquire Groq for $20B in cash (per CNBC, amplified via Reuters and other wire coverage) represents a materially different strategic posture than NVIDIA’s prior M&A pattern, given both the headline size (largest reported NVIDIA acquisition to date) and the unusual carve-out that Groq’s early-stage cloud business would not be included. Public reporting indicates the information originated from Alex Davis, CEO of Disruptive (lead investor in Groq’s latest financing), and that neither NVIDIA nor Groq had issued an immediate confirmation at the time of publication. The same reporting frames the transaction as coming together quickly, only months after Groq raised $750M at a ~$6.9B valuation, and highlights Groq’s positioning as a high-performance inference chip vendor founded by ex-Google TPU engineers. Groq is best understood as a vertically integrated inference acceleration company whose core asset is an application-specific processor optimized for deterministic, low-latency execution of transformer-style workloads, paired with a compiler-led software stack and a distribution layer (GroqCloud) designed to reduce developer friction via OpenAI-compatible APIs and integrations. Groq brands its architecture as a Language Processing Unit (LPU) and consistently emphasizes that the design target is inference, not training. The company’s own architecture description centers on 1-core execution, large on-chip SRAM used as primary storage (explicitly not cache), a custom compiler that statically schedules compute and communication, and direct chip-to-chip connectivity intended to coordinate multi-chip execution without relying on conventional caching hierarchies or dynamic runtime scheduling. The technical premise is a deliberate inversion of the conventional GPU approach. GPUs deliver throughput via massively parallel, multi-core execution with dynamic scheduling, complex memory hierarchies, and heavy reliance on off-chip HBM bandwidth and sophisticated runtime/kernel optimization. Groq instead argues that inference bottlenecks are driven by latency variance (tail latency), synchronization overhead, and memory access unpredictability inherent in dynamically scheduled, cache-heavy architectures, particularly when workloads are latency sensitive and batch sizes cannot be inflated. Groq’s solution is to move “control” into the compiler: the full execution graph and inter-chip communication schedule are computed ahead of time down to clock-cycle granularity, with deterministic execution designed to reduce run-to-run variance. In Groq’s framing, the removal of caches, reorder buffers, speculative execution overhead, and other sources of contention enables predictable latency and high utilization without per-model kernel engineering typical of GPU tuning cycles. A critical nuance is that Groq’s determinism is not merely a software claim; it is tightly coupled to architectural constraints and system design choices that trade flexibility for predictability. Third-party technical commentary indicates Groq’s chip uses a fully deterministic VLIW-style approach with minimal buffering, no external memory, and heavy dependence on sharding models across many chips because on-chip SRAM capacity is limited. SemiAnalysis describes a ~725 mm^2 die on GlobalFoundries 14nm with ~230MB of SRAM and notes that “no useful models” fit on a single chip, forcing multi-chip partitioning for modern LLMs and driving a system-level design where networking and compilation are first-class scheduling problems rather than ancillary infrastructure. This is consistent with Groq’s own messaging that tensor parallelism across chips is a primary design goal, enabled by large on-chip SRAM and compile-time coordination of compute plus interconnect. The on-chip SRAM emphasis is central to Groq’s latency story and also its most constraining trade-off. Groq claims on-chip SRAM bandwidth “upwards of 80 TB/s” and contrasts that with off-chip HBM bandwidth “about 8 TB/s,” asserting a potential 10x advantage from bandwidth plus reduced trips across chip-to-memory boundaries. While these comparisons are marketing-oriented and depend on workload specifics, the architectural implication is clear: Groq prioritizes ultra-fast local weight/activation access and then scales capacity by adding chips, not by attaching large off-chip memory pools. This design can reduce latency for sequential inference layers and minimize unpredictable stalls, but it pushes complexity into partitioning strategy, interconnect topology, and compiler scheduling, and it increases the number of chips needed for very large parameter counts and large KV-cache footprints. Groq also highlights numeric formats and compiler-driven precision management as a performance lever. In its 2025 technical blog, Groq describes “TruePoint numerics,” including 100-bit intermediate accumulation and selective quantization choices (FP32 for attention-sensitive operations, block floating point for MoE weights, FP8 storage in error-tolerant layers), and claims 2-4x speedups versus BF16 without measurable accuracy degradation on benchmarks such as MMLU and HumanEval. Even if the absolute uplift is workload dependent, the strategic point is that Groq is pursuing performance via end-to-end co-design: precision policy is not just hardware capability (FP8/BF16) but compiler-enforced mapping of precision to error sensitivity, which can matter materially for inference cost-per-token if it reduces memory traffic and boosts throughput without forcing aggressive, accuracy-damaging quantization. Independent performance datapoints indicate Groq has been credible on latency-oriented inference speed, at least for certain regimes. EE Times reported in 2023 that Groq demonstrated Llama-2 70B inference at ~240 tokens/s per user on a cloud-based dev system described as 10 racks and 64 chips, using the company’s 1st-gen silicon introduced several years earlier. Separate Groq commentary around independent benchmarking cites results showing ~241 tokens/s throughput and ~0.8s time to receive 100 output tokens for a Llama-2 70B API configuration, positioning the platform as a step-change in “available speed” for certain interactive use cases. These figures do not settle total cost-of-ownership versus GPUs or hyperscaler ASICs, but they establish that Groq’s system-level architecture can deliver strong single-user throughput and latency on large models when properly partitioned and scheduled. GroqCloud is the commercial wrapper that packages this hardware/software stack as “tokens-as-a-service,” aiming to make Groq adoption feel like switching API endpoints rather than adopting new silicon. Groq’s documentation states its API is designed to be “mostly compatible” with OpenAI client libraries, and its pricing page provides model-specific token rates, published speeds (tokens/s), prompt caching discounts, and batch processing discounts. For example, pricing lists inputs as low as $0.05 per 1M tokens and outputs as low as $0.08 per 1M tokens for certain smaller LLM configurations, with higher prices for larger models and long-context or MoE variants; it also advertises prompt caching with a 50% discount on cached input tokens for certain models and a batch API offering 50% lower cost for asynchronous processing windows. These mechanics are economically important because they demonstrate Groq’s go-to-market is not simply “sell chips,” but “sell predictable unit economics per token,” with tooling (batch, caching) that directly targets inference cost drivers (reused prompts, throughput smoothing, and asynchronous workloads). The cloud footprint and distribution partnerships indicate Groq has been building an inference-native “edge within the cloud” strategy rather than competing head-on with hyperscalers on breadth of services. A 2025 Groq newsroom release describes a European deployment in Helsinki with Equinix, positioned as latency reduction and data governance for European customers, and explicitly references Equinix Fabric enabling private connectivity to GroqCloud over public, private, or sovereign infrastructure. The same release enumerates additional capacity in the U.S. (Equinix, DataBank), Canada (Bell Canada), and Saudi Arabia (HUMAIN), and states these sites collectively served more than 20M tokens/s across Groq’s global network at that time. That supply-side metric matters because it provides a directional sense that Groq is scaling capacity as a network, not merely as a chip vendor. Customer disclosure is inherently limited because Groq is private and many enterprise deployments are not public, but Groq’s marketing materials and partnerships provide signals about demand vectors. The company’s public website displays logos of large consumer and enterprise brands (e.g., Dropbox, Vercel, Chevron, Volkswagen, Canva, Robinhood, Riot Games, Workday, Ramp) and includes a published customer quote claiming a 7.41x chat speed increase and an 89% cost reduction after moving to GroqCloud, followed by a tripling of token consumption. While marketing claims should be treated as case-specific and not generalized, they indicate that Groq is targeting both AI-native developers (who measure success by latency and cost-per-token) and enterprise buyers (who care about predictable performance and governance). Supplier and dependency mapping for Groq spans 3 layers: silicon production, system integration, and cloud infrastructure. On silicon, third-party analysis indicates GlobalFoundries 14nm for the 1st-gen Groq chip, implying a supply chain less constrained by the most capacity-tight leading-edge nodes and advanced packaging bottlenecks that dominate high-end GPU supply (HBM stacks, CoWoS-type packaging constraints). If accurate, this is strategically meaningful because it suggests Groq capacity expansion could be gated more by conventional wafer supply, board assembly, and data center power than by the same HBM/advanced packaging scarcity that has constrained top-tier GPU ramp cycles. On systems and cloud, Groq’s own releases identify colocation and connectivity partners (Equinix, DataBank, Bell Canada) and a Middle East partner (HUMAIN), implying dependencies on data center real estate, power availability, and network connectivity, alongside procurement of standard server components, NICs/switching, racks, and cooling infrastructure. The Groq design narrative also emphasizes air cooling and reduced need for complex power/cooling infrastructure, which—if realized in deployments—can widen the set of feasible hosting locations and lower deployment friction relative to liquid-cooled, very high power density GPU racks. Against that backdrop, the strategic rationale for NVIDIA acquiring Groq can be framed as a set of overlapping objectives: inference silicon optionality, architectural hedging, competitive defense, and supply chain diversification, with the carve-out of GroqCloud signaling a preference to avoid direct cloud competition and to focus on IP and product portfolio control rather than operating a capital-intensive token-serving business. The deal, if confirmed, would occur at a valuation step-up of ~190% versus Groq’s reported ~$6.9B private valuation in the September $750M round, reinforcing that any acquisition logic would be predominantly strategic rather than a conventional financial multiple arbitrage. The most compelling strategic driver is inference. Training has historically been the center of gravity for cutting-edge GPU demand, but inference volume is structurally larger and more distributed as deployments scale, with economics dominated by cost-per-token, latency guarantees, and utilization under spiky demand. Inference workloads also create a strategic vulnerability for NVIDIA: hyperscalers and large platforms can justify bespoke ASICs (TPU, Trainium/Inferentia, Maia-class efforts) because inference is stable, repeatable, and can amortize software investment at massive scale. Groq’s core proposition—deterministic, compiler-scheduled inference with predictable latency—aligns directly with the segment where GPU generality is least valued and where “good enough” programmability plus superior unit economics can win share. Acquiring Groq would allow NVIDIA to own a credible inference-native architecture rather than relying solely on GPUs and software optimization to defend that segment. Competitive defense logic is also plausible. Groq occupies a specific competitive wedge: low-latency, high-throughput interactive inference, delivered via a simple API abstraction that reduces switching cost. That wedge directly pressures GPU inference margins in the long run because it makes inference price/performance comparisons more transparent at the token level, and it targets a developer persona that historically defaulted to CUDA-first ecosystems. Even if NVIDIA’s current-generation systems can achieve very high tokens/s per user with extensive optimization, the strategic risk is that competing architectures normalize the idea that inference is best served by special-purpose silicon with a simpler programming model, weakening CUDA lock-in at the application layer. NVIDIA has actively demonstrated that Blackwell-era systems can exceed 1,000 tokens/s per user in benchmarked configurations, but that performance leadership does not automatically translate to lowest cost-per-token across the full range of batch sizes, latency targets, and deployment environments. Groq’s existence as a credible alternative architecture forces NVIDIA to keep defending inference economics rather than only raw performance leadership. The “technology acquisition” rationale is unusually strong in this specific case because Groq’s differentiator is not a single block of silicon IP but an end-to-end methodology: compiler-led static scheduling, deterministic networking, and a system architecture designed around tensor-parallel inference rather than throughput-maximizing batch inference. NVIDIA’s stack is already compiler-heavy (TensorRT, Triton, CUDA graphs, kernel fusion, speculative decoding techniques), but GPUs remain dynamically scheduled devices with complex memory hierarchies and stochastic latency behaviors under contention. Groq’s approach provides an alternate design point: treating the entire inference execution (compute plus communication) as a statically schedulable program. In principle, that IP could be valuable even if Groq silicon itself is not adopted at massive scale, because it can inform how NVIDIA builds future inference-optimized products, compilers, and networking fabrics, especially as distributed inference with large models makes communication a first-order performance determinant. Supply chain diversification is a non-obvious but potentially important driver. If Groq’s mainstream product generation is truly based on a mature process node and avoids HBM, then the scaling constraints look different than those of state-of-the-art GPUs. NVIDIA’s ability to meet incremental demand has been tightly coupled to advanced packaging and HBM supply, and those constraints can remain binding even when wafer supply is available. An inference ASIC architecture that relies primarily on on-chip SRAM and scales by adding chips—while not costless—could reduce dependence on HBM availability and advanced packaging capacity, enabling NVIDIA to ship “inference capacity” in higher absolute volumes or into geographies and customer segments where the highest-end GPUs are economically or logistically difficult to deploy. This could be particularly relevant for latency-sensitive inference deployed in regional colocation footprints rather than centralized hyperscale campuses. The carve-out of GroqCloud, if accurate, is itself a strategic signal about NVIDIA’s priorities. Operating a token-serving cloud at scale is capital intensive, structurally lower margin than silicon IP rents, and creates channel conflict with hyperscalers and CSP partners who are core NVIDIA customers. NVIDIA has generally positioned its cloud offerings through partnerships rather than as a direct hyperscale competitor. Excluding GroqCloud would preserve neutrality with CSPs and avoid inheriting multi-region data residency obligations and partner contracts, while still allowing NVIDIA to acquire Groq’s silicon, compiler technology, and engineering talent. At the same time, excluding GroqCloud would also mean NVIDIA would not automatically acquire the commercial proof-point of Groq’s unit economics or the customer contracts that validate product-market fit at scale, increasing the importance of diligence on whether Groq’s cloud pricing is structurally profitable or partially subsidized by fundraising. There is also a “preemptive acquisition” angle. The reporting identifies recent investors in Groq’s latest round including large financial institutions and strategic/industry players. In that context, Groq represents an asset that could plausibly have been acquired by a competitor (AMD/Intel) or by a hyperscaler seeking to accelerate inference independence. NVIDIA acquiring Groq could be a defensive move to prevent a credible inference-native architecture from being weaponized by a rival with deep distribution. Even if GroqCloud is carved out, controlling the silicon roadmap and compiler IP would meaningfully constrain Groq’s ability to evolve into a standalone competitor, unless the carved-out entity retains long-term rights to the hardware and software stack. However, the strategic case is not one-sided; there are meaningful risks and potential contradictions that would need to be reconciled for the transaction to be value-accretive on a multi-year horizon. 1st, Groq’s architecture appears to rely on scaling out chip count to achieve capacity, which introduces system cost, networking complexity, and physical footprint considerations. The absence of external memory and limited on-chip SRAM implies very large models require substantial chip parallelism, and the economics then depend heavily on chip cost, yield, power efficiency, and interconnect overhead. SemiAnalysis explicitly frames Groq as trading space for time and raises questions about token economics and whether publicly advertised pricing reflects fully loaded costs or market share capture. 2nd, integration risk is non-trivial. Groq’s compiler-led deterministic model is philosophically and practically different from CUDA’s dominant programming and execution model. A poorly executed integration could create internal product confusion, dilute engineering focus, or alienate developers if the combined stack fragments. 3rd, there is cannibalization risk. If Groq-class inference silicon undercuts GPU inference economics, NVIDIA could face internal margin trade-offs, even if the goal is to defend share against hyperscaler ASICs. Cannibalization can still be rational if it prevents larger share loss, but it would require crisp portfolio segmentation and go-to-market discipline. The presence of NVIDIA’s own rapidly improving inference performance complicates the “need” for Groq but does not eliminate the “option value.” NVIDIA has demonstrated benchmark-leading tokens/s per user on Blackwell-based systems, suggesting that raw interactive throughput is not necessarily the limiting factor for NVIDIA’s product line. The more enduring strategic question is unit economics and architectural control: whether future inference demand is better monetized through general-purpose GPUs plus software optimization, or whether a bifurcated product portfolio (training GPUs plus inference-native ASICs) becomes necessary to defend total AI compute wallet share as hyperscaler ASIC penetration increases. Acquiring Groq could be a decisive move to ensure NVIDIA participates in both regimes rather than betting exclusively on GPUs to win inference forever. What is “special” about Groq’s technology relative to a typical accelerator roadmap is the tight coupling of determinism, compilation, and networking into a single scheduling problem. The LPU narrative emphasizes deterministic compute and networking, static scheduling, and direct chip-to-chip coordination that allows “hundreds” (more precisely, 100s) of chips to behave like a single scheduled resource. The architecture also explicitly targets tensor-parallel, latency-optimized distribution rather than pure data-parallel throughput scaling, which matters for real-time applications where a single response must arrive quickly rather than many requests being processed in bulk. The implication is that Groq is optimized for the time-to-first-token and steady token streaming behavior that defines user experience in interactive LLMs, and it attempts to achieve that without relying on large batch sizes that can degrade latency. From a portfolio manager’s perspective, the most important interpretation is that an NVIDIA-Groq combination would likely be less about “NVIDIA needs more inference speed” and more about controlling the architectural trajectory of inference acceleration and removing a fast-improving, developer-friendly competitor from the market. The carve-out of GroqCloud would reinforce that the transaction is aimed at IP, talent, and product optionality, not acquiring a cloud revenue stream. The valuation step-up implied by $20B versus $6.9B would therefore be justified only if the acquired assets materially reduce long-term competitive risk (hyperscaler ASIC displacement, inference margin compression) or enable new monetization vectors (inference ASIC product line, supply chain de-bottlenecking, improved software determinism) that would be difficult to achieve on a comparable timeline via internal R&D.

TheValueist

101,296 просмотров • 6 месяцев назад

Been discussion on Blessing Adeoye Jr. of Kinda Funny taking issue with #Xbox going to Jez Corden of Windows Central, for new CEO Asha Sharma's interview. Don't think it's "jealousy" contrary to Xbox fandom musings. He listed numerous far more reputable journalists/outlets for the task, with no mention of Kinda Funny. Instead, think it's a matter of credibility and authenticity. Jez's often referred to as an Xbox and Microsoft "shill" with good reason (reasons later). Even by important figures such as Obsidian's co-founder. Xbox going to Jez signals they wanted controlled PR and favoured narratives, not opportunities for hard hitting insight or truth. Inadvertently very revealing. Xbox's moves this last week let us know how Xbox is still operating. +First Asha immediately begins aggressive toxic parasocial antics and "I'm one of you!" PR on social media that Xbox execs are famed for, to evangelise the fandom and have them run positive PR. +She creates a new Gamertag just a month ago with very irregular play, later admits it was a shared account played by others. +Her X responses seem robotic and inorganic. +Xbox have Tom Warren (another Xbox leaning journalist) of The Verge run a hit piece attempting to scapegoat Sarah Bond, which doesn't stand up to scrutiny. +In Satya Nadella's very long statement on the shake ups, Sarah Bond isn't mentioned once. +Asha's first interview is with Jez. +Messaging is a whole load of nothing. Teases things fans want to hear without actually committing to changes. What this all indirectly tells us is it's the same slippery and not so authentic Xbox, and we shouldn't really put much stock into anything they or their execs are currently saying. Only folk who are going to be won over by this controlled PR are the same naive echo chamber Xbox hopefuls who've fallen for or parrot every Xbox narrative/PR, despite countless flip flops, pivots and failings. So why's Jez Corden discredited or seen as an Xbox "shill" by so many? Myriad reasons. This is the same guy who said: +Development on Everwild was going well, gameplay loop was nailed and we'd see it soon, before it was cancelled shortly after. +He wasn't worried about The Initiative or Perfect Dark based on research he'd done, despite layoffs/Crystal Dynamics stuff. Was canned and studio shut shortly after. +Xbox wasn't porting games to PS5, rumour was fanboy hopium, before Xbox announced PS5 ports a few months later. +Implied all Laura Fryer worked on was Xbox One after her critical vid on Xbox, when she'd worked on OG Xbox and 360. +Physical games retail was dead in the UK after unfavourable Xbox sales, when data showed ~40% of sales of the Top 20 games were physical. +Said WuKong never mentioned Series S as a reason for Xbox version delay, implied it was a deal, despite WuKong director pinning blame on Series S. +Xbox might drop prices on games, only to suggest we knew price increases were coming when Xbox increased prices on GP. +No aspect of Halo rumours were true, when Halo moved to UE5 per rumours. +Rumours of Phil's retirement were made up, just months before he retired. +He'd never covered Avowed when called out, when he'd done an article hyping it. +Mischaracterised Tim Stuart's quote on bringing games to more platforms with clarification question to Phil on Game Pass instead of "first party experiences". +Sony's shown "no interest in mobile" when they acquired multiple mobile game studios, been open on mobile expansion and bringing more IP to mobile, set up a mobile gaming division in 2022 and have one of the worlds highest earning mobile games in Fate/Grand Order. +Promoted ABK acquisition and how good it'd be for devs/gamers, when Xbox has suffered 4k+ layoffs since (FAR higher than industry avg in % of total workforce), most in ABK, and GP has seen a 100% price increase in last 2+ years alone. I could go on and on. Point is, so much of the time Jez's running PR, defence and favoured narratives for Xbox, often with misinformation

Been discussion on Blessing Adeoye Jr. of Kinda Funny taking issue with #Xbox going to Jez Corden of Windows Central, for new CEO Asha Sharma's interview. Don't think it's "jealousy" contrary to Xbox fandom musings. He listed numerous far more reputable journalists/outlets for the task, with no mention of Kinda Funny. Instead, think it's a matter of credibility and authenticity. Jez's often referred to as an Xbox and Microsoft "shill" with good reason (reasons later). Even by important figures such as Obsidian's co-founder. Xbox going to Jez signals they wanted controlled PR and favoured narratives, not opportunities for hard hitting insight or truth. Inadvertently very revealing. Xbox's moves this last week let us know how Xbox is still operating. +First Asha immediately begins aggressive toxic parasocial antics and "I'm one of you!" PR on social media that Xbox execs are famed for, to evangelise the fandom and have them run positive PR. +She creates a new Gamertag just a month ago with very irregular play, later admits it was a shared account played by others. +Her X responses seem robotic and inorganic. +Xbox have Tom Warren (another Xbox leaning journalist) of The Verge run a hit piece attempting to scapegoat Sarah Bond, which doesn't stand up to scrutiny. +In Satya Nadella's very long statement on the shake ups, Sarah Bond isn't mentioned once. +Asha's first interview is with Jez. +Messaging is a whole load of nothing. Teases things fans want to hear without actually committing to changes. What this all indirectly tells us is it's the same slippery and not so authentic Xbox, and we shouldn't really put much stock into anything they or their execs are currently saying. Only folk who are going to be won over by this controlled PR are the same naive echo chamber Xbox hopefuls who've fallen for or parrot every Xbox narrative/PR, despite countless flip flops, pivots and failings. So why's Jez Corden discredited or seen as an Xbox "shill" by so many? Myriad reasons. This is the same guy who said: +Development on Everwild was going well, gameplay loop was nailed and we'd see it soon, before it was cancelled shortly after. +He wasn't worried about The Initiative or Perfect Dark based on research he'd done, despite layoffs/Crystal Dynamics stuff. Was canned and studio shut shortly after. +Xbox wasn't porting games to PS5, rumour was fanboy hopium, before Xbox announced PS5 ports a few months later. +Implied all Laura Fryer worked on was Xbox One after her critical vid on Xbox, when she'd worked on OG Xbox and 360. +Physical games retail was dead in the UK after unfavourable Xbox sales, when data showed ~40% of sales of the Top 20 games were physical. +Said WuKong never mentioned Series S as a reason for Xbox version delay, implied it was a deal, despite WuKong director pinning blame on Series S. +Xbox might drop prices on games, only to suggest we knew price increases were coming when Xbox increased prices on GP. +No aspect of Halo rumours were true, when Halo moved to UE5 per rumours. +Rumours of Phil's retirement were made up, just months before he retired. +He'd never covered Avowed when called out, when he'd done an article hyping it. +Mischaracterised Tim Stuart's quote on bringing games to more platforms with clarification question to Phil on Game Pass instead of "first party experiences". +Sony's shown "no interest in mobile" when they acquired multiple mobile game studios, been open on mobile expansion and bringing more IP to mobile, set up a mobile gaming division in 2022 and have one of the worlds highest earning mobile games in Fate/Grand Order. +Promoted ABK acquisition and how good it'd be for devs/gamers, when Xbox has suffered 4k+ layoffs since (FAR higher than industry avg in % of total workforce), most in ABK, and GP has seen a 100% price increase in last 2+ years alone. I could go on and on. Point is, so much of the time Jez's running PR, defence and favoured narratives for Xbox, often with misinformation

NIB

19,584 просмотров • 4 месяцев назад