Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Delphi AI researcher Accelxr 👾 explains how Chinese labs are optimizing AI for domestic chips. “The Chinese labs have actually been going beneath that, handwriting low-level GPU instructions and exploiting some of the chip behaviors at a lower level.”

Delphi Digital

1,538,075 subscribers

38,928 Aufrufe • vor 5 Tagen •via X (Twitter)

Nachrichten & Politik Wissenschaft & Technologie

Anya Rossi• Live Now

Private livecam show

5 Kommentare

Profilbild von Delphi Digital

Delphi Digitalvor 5 Tagen

Watch the full episode here.

Profilbild von Omega

Omegavor 5 Tagen

@accelxr Writing assembly by hand is not optimization, it is just flexing on everyone using a compiler.

Profilbild von Sophs m

Sophs mvor 5 Tagen

@accelxr low-level optimization is honestly way more impressive than just scaling params

Profilbild von Shinka - AI

Shinka - AIvor 5 Tagen

@accelxr Export controls just made handwriting GPU microcode a strategic advantage for Chinese labs.

Profilbild von Scam Casino Alert

Scam Casino Alertvor 5 Tagen

@accelxr casinos here have been doing the same thing with legacy backend software for years. Patching withdrawal logic by hand when the original devs are long gone

Ähnliche Videos

Delphi AI researcher Accelxr 👾 explains how China is scaling AI compute with domestic chips. “These superclusters take a ton of Ascend chips and combine them together. The supercluster itself has, I think, over 500k Ascend chips. They're pushing on it for sure.”

Delphi AI researcher Accelxr 👾 explains how China is scaling AI compute with domestic chips. “These superclusters take a ton of Ascend chips and combine them together. The supercluster itself has, I think, over 500k Ascend chips. They're pushing on it for sure.”

Delphi Digital

29,443 Aufrufe • vor 4 Tagen

Delphi AI researcher Accelxr 👾 explains why access to Nvidia chips could slow China’s push for chip independence. “Getting them back on the Nvidia bridge should slow down their development, kind of paradoxically. It slows their sovereignty push, and any tightening actually accelerates their push to build out their own chips.”

Delphi AI researcher Accelxr 👾 explains why access to Nvidia chips could slow China’s push for chip independence. “Getting them back on the Nvidia bridge should slow down their development, kind of paradoxically. It slows their sovereignty push, and any tightening actually accelerates their push to build out their own chips.”

Delphi Digital

20,562 Aufrufe • vor 10 Tagen

New Delphi Podcast with Accelxr 👾, author of our latest report on China’s AI labs He went into the research thinking China was mostly distilling US models and benchmark-maxing. He came out with a very different view We cover: – Why DeepSeek is doing real research, not just copying – How Chinese labs are optimizing below CUDA – Why China’s electricity advantage could matter more than chips – Whether cheap open-source models are part of a broader strategy – Why more Chinese AI talent is returning home – Whether restricting NVIDIA chips could actually accelerate China’s independence – What happens if China reaches full AI training sovereignty 00:00 Intro 00:35 Is China Innovating or Just Distilling? 06:10 DeepSeek and China’s Leading AI Labs 10:15 How China Is Working Around Weaker Chips 18:25 China’s Biggest AI Advantage: Electricity 29:45 Three Scenarios for the US-China AI Race 40:20 America’s AI Talent Advantage Is Reversing 47:45 China’s Path to AI Sovereignty Full episode:

New Delphi Podcast with Accelxr 👾, author of our latest report on China’s AI labs He went into the research thinking China was mostly distilling US models and benchmark-maxing. He came out with a very different view We cover: – Why DeepSeek is doing real research, not just copying – How Chinese labs are optimizing below CUDA – Why China’s electricity advantage could matter more than chips – Whether cheap open-source models are part of a broader strategy – Why more Chinese AI talent is returning home – Whether restricting NVIDIA chips could actually accelerate China’s independence – What happens if China reaches full AI training sovereignty 00:00 Intro 00:35 Is China Innovating or Just Distilling? 06:10 DeepSeek and China’s Leading AI Labs 10:15 How China Is Working Around Weaker Chips 18:25 China’s Biggest AI Advantage: Electricity 29:45 Three Scenarios for the US-China AI Race 40:20 America’s AI Talent Advantage Is Reversing 47:45 China’s Path to AI Sovereignty Full episode:

Tommy

81,587 Aufrufe • vor 16 Tagen

Delphi AI researcher Accelxr 👾 explains why AI agents could favor cheaper open weights models. “Agentic use cases are extremely cost sensitive. And so if that’s the case, you kind of have this composition shift towards cheapest, but still adequately capable models.”

Delphi AI researcher Accelxr 👾 explains why AI agents could favor cheaper open weights models. “Agentic use cases are extremely cost sensitive. And so if that’s the case, you kind of have this composition shift towards cheapest, but still adequately capable models.”

Delphi Digital

23,931 Aufrufe • vor 3 Tagen

Jensen Huang: “We (Nvidia) are 100% out of China. We went from 95% market share to 0%. I can’t imagine any policymaker thinking that’s a good idea.” Chinese AI labs like DeepSeek now have to use Chinese chips for both inference and training. Chinese chips will inevitably rise if this Nvidia chip ban continues.

Jensen Huang: “We (Nvidia) are 100% out of China. We went from 95% market share to 0%. I can’t imagine any policymaker thinking that’s a good idea.” Chinese AI labs like DeepSeek now have to use Chinese chips for both inference and training. Chinese chips will inevitably rise if this Nvidia chip ban continues.

Yuchen Jin

2,362,286 Aufrufe • vor 11 Monaten

Delphi AI researcher Accelxr 👾 explains why China’s AI progress is more than just distillation. “Distillation can explain a large chunk of the capability and benchmark side, but it can’t explain a lot of the things they’ve shipped... The research is quite impressive.”

Delphi AI researcher Accelxr 👾 explains why China’s AI progress is more than just distillation. “Distillation can explain a large chunk of the capability and benchmark side, but it can’t explain a lot of the things they’ve shipped... The research is quite impressive.”

Delphi Digital

37,686 Aufrufe • vor 15 Tagen

China’s arrival as a serious AI competitor threatens to upend an investment boom so far dominated by a small group of mostly US and allied companies. Chinese labs are already producing frontier-level models despite restricted access to advanced chips, while a determined push to build a domestic semiconductor supply chain could eventually challenge the incumbents in AI hardware. In this interview, Will Denyer considers the winners and losers from the disruption, arguing that while Chinese competition threatens model and hardware makers, cheaper models and computing power should ultimately benefit cloud providers and AI users.

China’s arrival as a serious AI competitor threatens to upend an investment boom so far dominated by a small group of mostly US and allied companies. Chinese labs are already producing frontier-level models despite restricted access to advanced chips, while a determined push to build a domestic semiconductor supply chain could eventually challenge the incumbents in AI hardware. In this interview, Will Denyer considers the winners and losers from the disruption, arguing that while Chinese competition threatens model and hardware makers, cheaper models and computing power should ultimately benefit cloud providers and AI users.

Gavekal

10,696 Aufrufe • vor 1 Monat

$🚨 APPLE MAY HAVE JUST EXPOSED HOW EXTREME THE AI CHIP BUBBLE HAS BECOME. Apple is reportedly lobbying the White House for permission to buy memory chips from a Chinese company blacklisted by the Pentagon. Why? Because AI-driven chip prices have gotten so insane that even Apple can’t absorb the costs anymore. DeepSeek already showed the market you could get similar AI results for a fraction of the cost. If Apple starts moving toward cheaper Chinese chips too, the entire AI memory chip BUBBLE could start bursting.$

🚨 APPLE MAY HAVE JUST EXPOSED HOW EXTREME THE AI CHIP BUBBLE HAS BECOME. Apple is reportedly lobbying the White House for permission to buy memory chips from a Chinese company blacklisted by the Pentagon. Why? Because AI-driven chip prices have gotten so insane that even Apple can’t absorb the costs anymore. DeepSeek already showed the market you could get similar AI results for a fraction of the cost. If Apple starts moving toward cheaper Chinese chips too, the entire AI memory chip BUBBLE could start bursting.

Crypto Rover

212,727 Aufrufe • vor 2 Monaten

Ayar Labs and Wiwynn are partnering to bring CPO to rack-scale AI systems that support next-generation hyperscale AI workloads. By combining Ayar Labs’ CPO solution with Wiwynn’s rack-level system design and manufacturing capabilities, the two companies are enabling a new class of rack-scale AI infrastructure.

Ayar Labs and Wiwynn are partnering to bring CPO to rack-scale AI systems that support next-generation hyperscale AI workloads. By combining Ayar Labs’ CPO solution with Wiwynn’s rack-level system design and manufacturing capabilities, the two companies are enabling a new class of rack-scale AI infrastructure.

Ayar Labs

137,455 Aufrufe • vor 3 Monaten

SITUATION EXPLAINED: The Trump administration is closing the loophole that lets Chinese labs rent US chips. • A small group inside Commerce is building a slimmed-down replacement for the Biden-era AI Diffusion Rule, which the administration scrapped in 2025 • The target is Chinese AI firms accessing advanced chips remotely through data centers abroad, in countries including Thailand and Singapore • The current regime is upside down: the US now permits H200 sales to China while Beijing forbids its own firms from buying them • Nvidia's share of China's AI chip market fell from 95% in 2023 to 8% in 2026, while Huawei climbed to roughly 50-60% Theo Jaffee: "It's kind of unclear to me how important access to American chips is for Chinese AI progress, especially if GLM 5.3 is served entirely on Chinese chips, which is what they claim."

SITUATION EXPLAINED: The Trump administration is closing the loophole that lets Chinese labs rent US chips. • A small group inside Commerce is building a slimmed-down replacement for the Biden-era AI Diffusion Rule, which the administration scrapped in 2025 • The target is Chinese AI firms accessing advanced chips remotely through data centers abroad, in countries including Thailand and Singapore • The current regime is upside down: the US now permits H200 sales to China while Beijing forbids its own firms from buying them • Nvidia's share of China's AI chip market fell from 95% in 2023 to 8% in 2026, while Huawei climbed to roughly 50-60% Theo Jaffee: "It's kind of unclear to me how important access to American chips is for Chinese AI progress, especially if GLM 5.3 is served entirely on Chinese chips, which is what they claim."

MTS

10,458 Aufrufe • vor 21 Tagen

BANNON: Shut down 100% of computer chips being exported to the Chinese Communist Party! No 350,000 kids here in our universities, no Chinese nationals at the corporations, nobody at the labs, NOBODY. Shut the whole ecosystem down! NO money, NO equity, NO debt, BOOM. If we do that, it won't matter how much oil they get from the Middle East. If we're going to get serious about AI, let's get serious about AI!

BANNON: Shut down 100% of computer chips being exported to the Chinese Communist Party! No 350,000 kids here in our universities, no Chinese nationals at the corporations, nobody at the labs, NOBODY. Shut the whole ecosystem down! NO money, NO equity, NO debt, BOOM. If we do that, it won't matter how much oil they get from the Middle East. If we're going to get serious about AI, let's get serious about AI!

Bannon’s WarRoom

21,205 Aufrufe • vor 4 Monaten

SITUATION EXPLAINED: American AI labs are worth ~$1T each. Chinese AI labs are worth tens of billions. What explains the gap? • Moonshot AI (Kimi) raising $2B at $30B, its third financing in six months • OpenAI and Anthropic: both around $1T • DeepSeek and Moonshot: tens of billions... one to two orders of magnitude smaller • DeepSeek rumored at ~$200M ARR • Chinese models don't compete on capability, they compete on price Theo Jaffee: "Markets are probably pricing in OpenAI and Anthropic kicking off recursive self-improvement first. The niche of DeepSeek and Moonshot is not that they're better... it's that they're much, much cheaper. Which is good for enterprise applications. But it's not as good for kicking off the singularity."

SITUATION EXPLAINED: American AI labs are worth ~$1T each. Chinese AI labs are worth tens of billions. What explains the gap? • Moonshot AI (Kimi) raising $2B at $30B, its third financing in six months • OpenAI and Anthropic: both around $1T • DeepSeek and Moonshot: tens of billions... one to two orders of magnitude smaller • DeepSeek rumored at ~$200M ARR • Chinese models don't compete on capability, they compete on price Theo Jaffee: "Markets are probably pricing in OpenAI and Anthropic kicking off recursive self-improvement first. The niche of DeepSeek and Moonshot is not that they're better... it's that they're much, much cheaper. Which is good for enterprise applications. But it's not as good for kicking off the singularity."

MTS

11,559 Aufrufe • vor 3 Monaten

SITUATION EXPLAINED: What does the Chinese AI ecosystem actually look like beyond DeepSeek? Gabrierl (.Gabriel) asked FleetingBits, anonymous AI researcher. "Most people are familiar with DeepSeek. But if we look at other Chinese labs like Zhipu, Minimax, Moonshot, ByteDance, they're a little bit less well known." "Zhipu is sort of the Chinese equivalent of Palantir... it receives state support through contracts with state-owned enterprises. GLM 5.2 was a Zhipu model. And unique among Chinese AI labs, it has actually pretty good gross margins, around 40%. All of its revenue comes from these state-owned enterprises for which it does these Palantir deployments where it takes its GLM models and deploys them locally on their hardware." "MiniMax is interesting because they're sort of the character AI of China, a companion app called Talky and a video generation app called Hailuo. And the majority of their revenue comes from outside China, including a decent chunk from the United States." "Doubao, which is ByteDance's equivalent of ChatGPT, is the most popular AI app in China." "The Chinese ChatGPT is actually ByteDance. And I think they don't make a splash in America because they don't open source any of their models."

SITUATION EXPLAINED: What does the Chinese AI ecosystem actually look like beyond DeepSeek? Gabrierl (.Gabriel) asked FleetingBits, anonymous AI researcher. "Most people are familiar with DeepSeek. But if we look at other Chinese labs like Zhipu, Minimax, Moonshot, ByteDance, they're a little bit less well known." "Zhipu is sort of the Chinese equivalent of Palantir... it receives state support through contracts with state-owned enterprises. GLM 5.2 was a Zhipu model. And unique among Chinese AI labs, it has actually pretty good gross margins, around 40%. All of its revenue comes from these state-owned enterprises for which it does these Palantir deployments where it takes its GLM models and deploys them locally on their hardware." "MiniMax is interesting because they're sort of the character AI of China, a companion app called Talky and a video generation app called Hailuo. And the majority of their revenue comes from outside China, including a decent chunk from the United States." "Doubao, which is ByteDance's equivalent of ChatGPT, is the most popular AI app in China." "The Chinese ChatGPT is actually ByteDance. And I think they don't make a splash in America because they don't open source any of their models."

MTS

172,212 Aufrufe • vor 2 Monaten

Where are DeepSeek and other Chinese AI firms going to be getting their chips over the next 5 years? Jordan Schneider says that without Nvidia hardware, they're just not going to be able to keep up. "There's this ratio between the US semi ecosystem and the China semi ecosystem. The capacity is like 15:1. "And you can project that out pretty clearly, doing some math around how much capex is being spent, how much equipment is being acquired, yield rates, etc." "So where is [DeepSeek founder Liang Wenfeng] going to get the chips? He's probably going to get as much Nvidia as he can get." "But the larger takeaway with the Chinese and Western ecosystems is that the Western one is way better capitalized and has way more straightforward access to chips, which is basically an order of magnitude better over at least the next 5 years." "So it's going to be rough sledding for Chinese firms." "The Chinese AI IPOs are making like, $500M. Or like $1B. And it's cute. But it's not $3 trillion. Or even $10B. And the reason that's the case is that they're not on the frontier — so they can't charge a premium — and they're compute-constrained." "If all the firms have is access to Chinese chips, it's going to be really difficult to make a lot of money, and keep paying for more chips, in the way that the Western labs have been able to mobilize the entire global financial system."

Where are DeepSeek and other Chinese AI firms going to be getting their chips over the next 5 years? Jordan Schneider says that without Nvidia hardware, they're just not going to be able to keep up. "There's this ratio between the US semi ecosystem and the China semi ecosystem. The capacity is like 15:1. "And you can project that out pretty clearly, doing some math around how much capex is being spent, how much equipment is being acquired, yield rates, etc." "So where is [DeepSeek founder Liang Wenfeng] going to get the chips? He's probably going to get as much Nvidia as he can get." "But the larger takeaway with the Chinese and Western ecosystems is that the Western one is way better capitalized and has way more straightforward access to chips, which is basically an order of magnitude better over at least the next 5 years." "So it's going to be rough sledding for Chinese firms." "The Chinese AI IPOs are making like, $500M. Or like $1B. And it's cute. But it's not $3 trillion. Or even $10B. And the reason that's the case is that they're not on the frontier — so they can't charge a premium — and they're compute-constrained." "If all the firms have is access to Chinese chips, it's going to be really difficult to make a lot of money, and keep paying for more chips, in the way that the Western labs have been able to mobilize the entire global financial system."

TBPN

52,507 Aufrufe • vor 18 Tagen

Large amounts of Chinese nationals are working at our nation's top AI labs. We know that China will stop at nothing, including espionage and coercion, to beat us in the AI race. This is why I secured language in this year's NDAA to require American AI labs to work collaboratively with the government to catch insider threats before they undermine our national security.

Large amounts of Chinese nationals are working at our nation's top AI labs. We know that China will stop at nothing, including espionage and coercion, to beat us in the AI race. This is why I secured language in this year's NDAA to require American AI labs to work collaboratively with the government to catch insider threats before they undermine our national security.

Senator Jim Banks

46,680 Aufrufe • vor 1 Monat

US officials accused six Chinese AI companies, including DeepSeek and Alibaba, of copying American AI models. 🇺🇸🇨🇳 Washington is now treating stolen AI research the way it treats stolen military secrets. That's overdue. American labs have spent hundreds of billions building these models, and Chinese firms have been quietly training cheaper copies. CISA, NSA, and the FBI put their names on the same statement together, which almost never happens.

US officials accused six Chinese AI companies, including DeepSeek and Alibaba, of copying American AI models. 🇺🇸🇨🇳 Washington is now treating stolen AI research the way it treats stolen military secrets. That's overdue. American labs have spent hundreds of billions building these models, and Chinese firms have been quietly training cheaper copies. CISA, NSA, and the FBI put their names on the same statement together, which almost never happens.

Donald J Trump Posts TruthSocial

66,199 Aufrufe • vor 7 Tagen

BANNON: If we don’t want China competing with us in AI, STOP BUILDING THEIR ECOSYSTEM: 1. Cut off chips, including black-market routes. 2. Stop educating Chinese nationals at our top engineering schools. 3. Stop giving Chinese nationals visas to work in frontier labs and top tech companies. 4. Freeze U.S. pension money going to venture capital firms financing the CCP’s AI and advanced-weapons buildout. 5. Cut off Chinese access to U.S. equity and debt markets. WE CAN SHUT DOWN THE CHINESE COMMUNIST PARTY’S AI ECOSYSTEM IN 72 HOURS.

BANNON: If we don’t want China competing with us in AI, STOP BUILDING THEIR ECOSYSTEM: 1. Cut off chips, including black-market routes. 2. Stop educating Chinese nationals at our top engineering schools. 3. Stop giving Chinese nationals visas to work in frontier labs and top tech companies. 4. Freeze U.S. pension money going to venture capital firms financing the CCP’s AI and advanced-weapons buildout. 5. Cut off Chinese access to U.S. equity and debt markets. WE CAN SHUT DOWN THE CHINESE COMMUNIST PARTY’S AI ECOSYSTEM IN 72 HOURS.

Bannon’s WarRoom

46,386 Aufrufe • vor 4 Tagen

The American AI company AfterQuery AfterQuery is reportedly generating at least $50 MILLION in recurring revenue from Chinese AI labs while maintaining ties to Ant Group Ant Group and Alibaba, two major Chinese technology companies. Its CEO Spencer Mateega publicly celebrated Moonshot AI’s Kimi model for topping an AfterQuery benchmark and called the Chinese team “cracked.”Moonshot AI is a Chinese AI company based in Beijing, China American AI executives are traveling to China, hiring China engagement managers and openly praising the Chinese models competing to overtake the United States in AI dominance. How many warning signs does Washington need before it stops American AI companies from helping China overtake the United States in the AI race? David Sacks

The American AI company AfterQuery AfterQuery is reportedly generating at least $50 MILLION in recurring revenue from Chinese AI labs while maintaining ties to Ant Group Ant Group and Alibaba, two major Chinese technology companies. Its CEO Spencer Mateega publicly celebrated Moonshot AI’s Kimi model for topping an AfterQuery benchmark and called the Chinese team “cracked.”Moonshot AI is a Chinese AI company based in Beijing, China American AI executives are traveling to China, hiring China engagement managers and openly praising the Chinese models competing to overtake the United States in AI dominance. How many warning signs does Washington need before it stops American AI companies from helping China overtake the United States in the AI race? David Sacks

Laura Loomer

66,528 Aufrufe • vor 28 Tagen

We ran the latest AI models from leading Chinese labs through some 3D builds and game-clone generations in Code Arena for Lunar New Year 🧧 Take a look at what they created, and let us know what you think.

We ran the latest AI models from leading Chinese labs through some 3D builds and game-clone generations in Code Arena for Lunar New Year 🧧 Take a look at what they created, and let us know what you think.

Arena.ai

12,418 Aufrufe • vor 7 Monaten