Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

📣 Get a first look at the NVIDIA Photonics co-packaged optics switch with Lambda. At NVIDIA GB300 NVL72 scale, the network doesn't just move data between GPUs — it determines how fast your cluster thinks. Co-packaged optics cut switch power, reduce failure points, and deliver more tokens per watt....

NVIDIAAIInfra's profile picture

NVIDIA AI Infrastructure

68,868 subscribers

84,302 Aufrufe • vor 1 Monat •via X (Twitter)

BildungWissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

First look at $NVDA photonics co-packaged optics switch with Lambda. As GB300 NVL72 scales, networking becomes the layer that determines how fast the cluster thinks while co-packaged optics cut power and improve tokens per watt.
0:20

First look at $NVDA photonics co-packaged optics switch with Lambda. As GB300 NVL72 scales, networking becomes the layer that determines how fast the cluster thinks while co-packaged optics cut power and improve tokens per watt.

Shay Boloor

101,190 Aufrufe • vor 1 Monat

Nvidia CEO Jensen Huang talks about NVIDIA Co-Packaged Optics at GTC 2026
0:48

Nvidia CEO Jensen Huang talks about NVIDIA Co-Packaged Optics at GTC 2026

CNET

10,067,660 Aufrufe • vor 3 Monaten

The first-ever hydrogen-powered NVIDIA GB300 NVL72 systems are online. Training & inference at scale — with zero emissions, zero water. Lambda × Supermicro × ECL →
0:39

The first-ever hydrogen-powered NVIDIA GB300 NVL72 systems are online. Training & inference at scale — with zero emissions, zero water. Lambda × Supermicro × ECL →

Lambda

17,276 Aufrufe • vor 9 Monaten

Yesterday everyone sold optics because "CPO is delayed to 2027." Today I'm looking at this video of an $NVDA CPO switch running in production at Lambda, wired into GB300 clusters. Both headlines are true. CPO at the SWITCH is shipping now, 144 ports of 800G, liquid cooled, roughly 70% less power per port than pluggable optics. CPO at the GPU, the version that kills the module makers, that's the 2027-28 story, and it always was. Why does the switch get light first? Because the back end fabric is 86% of networking power in these clusters. At GB300 scale the network isn't plumbing anymore, it's part of the compute. Every watt you don't burn moving bits is a watt that generates tokens. Tokens per watt is the metric now. And the detail nobody will talk about again is even inside NVIDIA's most advanced CPO switch, the lasers sit OUTSIDE the package. 18 swappable lightsource modules. Field serviceable, because lasers fail. Pluggable, near package, copackaged, doesn't matter. Every road still runs through the laser makers…
0:20

Yesterday everyone sold optics because "CPO is delayed to 2027." Today I'm looking at this video of an $NVDA CPO switch running in production at Lambda, wired into GB300 clusters. Both headlines are true. CPO at the SWITCH is shipping now, 144 ports of 800G, liquid cooled, roughly 70% less power per port than pluggable optics. CPO at the GPU, the version that kills the module makers, that's the 2027-28 story, and it always was. Why does the switch get light first? Because the back end fabric is 86% of networking power in these clusters. At GB300 scale the network isn't plumbing anymore, it's part of the compute. Every watt you don't burn moving bits is a watt that generates tokens. Tokens per watt is the metric now. And the detail nobody will talk about again is even inside NVIDIA's most advanced CPO switch, the lasers sit OUTSIDE the package. 18 swappable lightsource modules. Field serviceable, because lasers fail. Pluggable, near package, copackaged, doesn't matter. Every road still runs through the laser makers…

Nicholas Mugalli

108,542 Aufrufe • vor 1 Monat

Nvidia just spent $4 billion on a technology 99% of people have never heard of. But in 3 years, every AI data center on Earth will need it. And Nvidia just LOCKED UP the supply. Here's what happened: Nvidia invested $2 billion in Coherent and $2 billion in Lumentum. You probably never heard of these companies. They make photonics technology. Systems that transmit data using LIGHT instead of electricity. Sounds like sci-fi. But this is the most important infrastructure bet in AI right now. Here's the problem Nvidia just solved for itself: AI data centers are hitting a wall that has nothing to do with chips, energy, or money... Copper wiring is dying. Every data center on Earth moves data between GPUs using copper cables. But at the speeds AI now demands, copper physically cannot keep up. Signal degrades. Heat explodes. Power consumption skyrockets. Right now, 30% of the electricity in an AI data center is wasted just MOVING data from point A to point B. An MIT researcher said: "Copper's not going to cut it. It gets too hot. Too much power consumption and loss." Jensen Huang admitted it himself too: "We use copper as far as we can, about a meter or two. But where data centers are the size of a stadium, we need something else." That something else is photonics. Replacing copper with laser-powered fiber optics built directly into the chip. The numbers are insane: - 3.5x more power efficient - 10x better network reliability - Data moving at 102 terabits per second Wells Fargo estimates the photonics market will hit $10-12 billion by 2030. And Nvidia just bought privileged access to the two companies that make the advanced lasers every single one of these systems will need. This is the Nvidia playbook on repeat. They did this with CoreWeave. Invested $2 billion, locked up GPU capacity, created a dependent customer. They did this with memory suppliers. Secured HBM allocations years in advance while competitors scrambled. Now they're doing it with photonics. Invest early. Lock up supply. Make the entire ecosystem dependent on companies that are dependent on Nvidia. By the time competitors realize photonics is the bottleneck, Nvidia already OWNS the supply chain. Every data center, AI factory, and GPU cluster will need this technology to function at scale. Nvidia will become even more important.
1:36

Nvidia just spent $4 billion on a technology 99% of people have never heard of. But in 3 years, every AI data center on Earth will need it. And Nvidia just LOCKED UP the supply. Here's what happened: Nvidia invested $2 billion in Coherent and $2 billion in Lumentum. You probably never heard of these companies. They make photonics technology. Systems that transmit data using LIGHT instead of electricity. Sounds like sci-fi. But this is the most important infrastructure bet in AI right now. Here's the problem Nvidia just solved for itself: AI data centers are hitting a wall that has nothing to do with chips, energy, or money... Copper wiring is dying. Every data center on Earth moves data between GPUs using copper cables. But at the speeds AI now demands, copper physically cannot keep up. Signal degrades. Heat explodes. Power consumption skyrockets. Right now, 30% of the electricity in an AI data center is wasted just MOVING data from point A to point B. An MIT researcher said: "Copper's not going to cut it. It gets too hot. Too much power consumption and loss." Jensen Huang admitted it himself too: "We use copper as far as we can, about a meter or two. But where data centers are the size of a stadium, we need something else." That something else is photonics. Replacing copper with laser-powered fiber optics built directly into the chip. The numbers are insane: - 3.5x more power efficient - 10x better network reliability - Data moving at 102 terabits per second Wells Fargo estimates the photonics market will hit $10-12 billion by 2030. And Nvidia just bought privileged access to the two companies that make the advanced lasers every single one of these systems will need. This is the Nvidia playbook on repeat. They did this with CoreWeave. Invested $2 billion, locked up GPU capacity, created a dependent customer. They did this with memory suppliers. Secured HBM allocations years in advance while competitors scrambled. Now they're doing it with photonics. Invest early. Lock up supply. Make the entire ecosystem dependent on companies that are dependent on Nvidia. By the time competitors realize photonics is the bottleneck, Nvidia already OWNS the supply chain. Every data center, AI factory, and GPU cluster will need this technology to function at scale. Nvidia will become even more important.

Ricardo

640,331 Aufrufe • vor 4 Monaten

Jensen Huang says Vera Rubin is now in full production and the supply chain is already twice the size of Grace Blackwell. $NVDA is officially ramping full AI factory systems with CPUs, GPUs and co-packaged optics built directly into the rack.
0:49

Jensen Huang says Vera Rubin is now in full production and the supply chain is already twice the size of Grace Blackwell. $NVDA is officially ramping full AI factory systems with CPUs, GPUs and co-packaged optics built directly into the rack.

Shay Boloor

70,283 Aufrufe • vor 1 Monat

🚨 NVIDIA just showed what the future of AI looks like at GTC Paris Here’s the top 7 wildest stuff they announced 👇 1/ NVIDIA’s new NVL72 system connects 72 Blackwell GPUs with 130 TB/s bandwidth
1:20

🚨 NVIDIA just showed what the future of AI looks like at GTC Paris Here’s the top 7 wildest stuff they announced 👇 1/ NVIDIA’s new NVL72 system connects 72 Blackwell GPUs with 130 TB/s bandwidth

Devin AI

873,095 Aufrufe • vor 1 Jahr

✨ The #NVIDIARubin platform is engineered to master multi-step problem-solving and massive long-context reasoning workflows at scale. Built using extreme co-design, Rubin delivers more AI tokens per watt and lower cost per token compared to the previous architecture. ✔️ 50 PFLOPS of NVFP4 inference ✔️ 22 TB/s of HBM4 bandwidth ✔️ 3.6 TB/s of NVIDIA NVLink bandwidth per GPU Learn more about NVIDIA Rubin ➡️
0:54

✨ The #NVIDIARubin platform is engineered to master multi-step problem-solving and massive long-context reasoning workflows at scale. Built using extreme co-design, Rubin delivers more AI tokens per watt and lower cost per token compared to the previous architecture. ✔️ 50 PFLOPS of NVFP4 inference ✔️ 22 TB/s of HBM4 bandwidth ✔️ 3.6 TB/s of NVIDIA NVLink bandwidth per GPU Learn more about NVIDIA Rubin ➡️

NVIDIA AI Infrastructure

10,475 Aufrufe • vor 5 Monaten

Jensen Huang shows off the NVIDIA GB200 NVL72: a data center superchip with 72 Blackwell GPUs, 1.4 exaFLOPS of compute and 130 trillion transistors
1:58

Jensen Huang shows off the NVIDIA GB200 NVL72: a data center superchip with 72 Blackwell GPUs, 1.4 exaFLOPS of compute and 130 trillion transistors

Tsarathustra

2,362,426 Aufrufe • vor 1 Jahr

The NVIDIA Vera Rubin platform is a multi-rack pod-scale system, built to process agentic AI and is now in full production. Built through extreme co-design, Vera Rubin unifies five connected rack-scale systems — the NVIDIA Vera Rubin NVL72, Vera CPU rack, Groq 3 LPX, Vera BlueField-4 STX, and Spectrum-6 SPX Ethernet — all purpose-built for global-scale agentic AI infrastructure. Vera Rubin delivers intelligence at scale, with the performance, efficiency and security needed to power the next industrial revolution. Congratulations to Microsoft, Dell Technologies, and CoreWeave for each standing up their Vera Rubin NVL72 engineering racks. Learn more:
4:14

The NVIDIA Vera Rubin platform is a multi-rack pod-scale system, built to process agentic AI and is now in full production. Built through extreme co-design, Vera Rubin unifies five connected rack-scale systems — the NVIDIA Vera Rubin NVL72, Vera CPU rack, Groq 3 LPX, Vera BlueField-4 STX, and Spectrum-6 SPX Ethernet — all purpose-built for global-scale agentic AI infrastructure. Vera Rubin delivers intelligence at scale, with the performance, efficiency and security needed to power the next industrial revolution. Congratulations to Microsoft, Dell Technologies, and CoreWeave for each standing up their Vera Rubin NVL72 engineering racks. Learn more:

NVIDIA

49,365 Aufrufe • vor 1 Monat

X-FLEXI announced a strategic partnership with NVIDIA to deploy the NVIDIA GB200 NVL72 cabinet.X-FLEXI announced a strategic partnership with NVIDIA to deploy the NVIDIA GB200 NVL72 cabinet
0:55

X-FLEXI announced a strategic partnership with NVIDIA to deploy the NVIDIA GB200 NVL72 cabinet.X-FLEXI announced a strategic partnership with NVIDIA to deploy the NVIDIA GB200 NVL72 cabinet

X-Flexi

131,205 Aufrufe • vor 8 Monaten

NVIDIA Vera Rubin is in full production and arrives just in time for the next frontier of AI. The #NVIDIARubin platform uses extreme co-design across six new chips to accelerate agentic AI, advanced reasoning and massive-scale MoE model inference. NVIDIA Vera Rubin NVL72: Six new chips, one AI supercomputer — see how we built it.
2:50

NVIDIA Vera Rubin is in full production and arrives just in time for the next frontier of AI. The #NVIDIARubin platform uses extreme co-design across six new chips to accelerate agentic AI, advanced reasoning and massive-scale MoE model inference. NVIDIA Vera Rubin NVL72: Six new chips, one AI supercomputer — see how we built it.

NVIDIA Data Center

27,643 Aufrufe • vor 6 Monaten

$NVDA networking is now running at a $60B+ annualized pace, growing ~200% YoY and nearly the size of $AVGO total 2025 revenue. As AI clusters scale, the bottleneck becomes how fast GPUs, CPUs, memory and storage move data together because even the best chips underperform if the network cannot keep up. This is why Nvidia is selling the full AI factory stack.
1:25

$NVDA networking is now running at a $60B+ annualized pace, growing ~200% YoY and nearly the size of $AVGO total 2025 revenue. As AI clusters scale, the bottleneck becomes how fast GPUs, CPUs, memory and storage move data together because even the best chips underperform if the network cannot keep up. This is why Nvidia is selling the full AI factory stack.

Shay Boloor

60,964 Aufrufe • vor 1 Monat

“NVIDIA’s cost per token is the lowest in the world.” — Jensen Huang, Founder & CEO of NVIDIA Token generation cost is a direct result of architecture excellence and extreme co-design, not just compute cost. Lowest cost per token and highest performance per watt are definitive measures of AI economics and the key to unlocking maximum profitability and AI revenue. ➡️
0:49

“NVIDIA’s cost per token is the lowest in the world.” — Jensen Huang, Founder & CEO of NVIDIA Token generation cost is a direct result of architecture excellence and extreme co-design, not just compute cost. Lowest cost per token and highest performance per watt are definitive measures of AI economics and the key to unlocking maximum profitability and AI revenue. ➡️

NVIDIA

88,885 Aufrufe • vor 3 Monaten

Groq's LPU is faster than GPUs, handling requests and responding more quickly. Groq's LPUs don't need speedy data delivery like Nvidia GPUs do because they don't have HBM in their system. They use SRAM, which is about 20 times faster than what GPUs use. Since inference runs use way less data than model training, Groq's LPU is more energy-efficient. It reads less from external memory and uses less power than a Nvidia GPU for inference tasks. The LPU works differently from GPUs. It uses a Temporal Instruction Set Computer architecture, so it doesn't have to reload data from memory as often as GPUs do with High Bandwidth Memory (HBM). This helps avoid issues with HBM shortages and keeps costs down. If Groq's LPU is used in places that do AI processing, you might not need special storage for Nvidia GPUs. The LPU doesn't demand super-fast storage like GPUs do. Groq is claiming that its technology could replace GPUs in AI tasks with its powerful chip and software.
7:57

Groq's LPU is faster than GPUs, handling requests and responding more quickly. Groq's LPUs don't need speedy data delivery like Nvidia GPUs do because they don't have HBM in their system. They use SRAM, which is about 20 times faster than what GPUs use. Since inference runs use way less data than model training, Groq's LPU is more energy-efficient. It reads less from external memory and uses less power than a Nvidia GPU for inference tasks. The LPU works differently from GPUs. It uses a Temporal Instruction Set Computer architecture, so it doesn't have to reload data from memory as often as GPUs do with High Bandwidth Memory (HBM). This helps avoid issues with HBM shortages and keeps costs down. If Groq's LPU is used in places that do AI processing, you might not need special storage for Nvidia GPUs. The LPU doesn't demand super-fast storage like GPUs do. Groq is claiming that its technology could replace GPUs in AI tasks with its powerful chip and software.

Brian Roemmele

87,128 Aufrufe • vor 2 Jahren

Nvidia just revealed Vera Rubin. Ships H2 2026. The numbers are wild: → 10x more performance per watt vs Blackwell → 10x cheaper inference token cost → 4x fewer GPUs to train the same MoE model Energy was the biggest bottleneck in AI. Nvidia just made it 10x cheaper.
0:27

Nvidia just revealed Vera Rubin. Ships H2 2026. The numbers are wild: → 10x more performance per watt vs Blackwell → 10x cheaper inference token cost → 4x fewer GPUs to train the same MoE model Energy was the biggest bottleneck in AI. Nvidia just made it 10x cheaper.

Min Choi

285,155 Aufrufe • vor 4 Monaten

Nvidia will now pay you to put a mini AI data center on your house It looks like a normal AC unit in the yard. But inside sits 16 Nvidia Blackwell GPUs and Dell servers. A startup called Span builds them, backed by Nvidia. They bolt onto your home and you get paid for the power and Wi-Fi. Some estimates put that around $1,000 a month in your pocket. That is rent money just for hosting a box outside. Span says it deploys way faster and cheaper than a real data center. The AI boom is literally moving into the suburbs. Save this, the grid is getting rebuilt in real time.
1:09

Nvidia will now pay you to put a mini AI data center on your house It looks like a normal AC unit in the yard. But inside sits 16 Nvidia Blackwell GPUs and Dell servers. A startup called Span builds them, backed by Nvidia. They bolt onto your home and you get paid for the power and Wi-Fi. Some estimates put that around $1,000 a month in your pocket. That is rent money just for hosting a box outside. Span says it deploys way faster and cheaper than a real data center. The AI boom is literally moving into the suburbs. Save this, the grid is getting rebuilt in real time.

winkle.

8,371,583 Aufrufe • vor 1 Monat

AI has turned data centers from 5–10 kW racks into power-hungry AI factories. At that scale, the question isn’t “who has the most FLOPs?” It’s: Who can deliver the most intelligence per watt and per dollar? Nvidia’s GPUs still dominate, but inside Google there’s a different answer: the Tensor Processing Unit (TPU):
1:09

AI has turned data centers from 5–10 kW racks into power-hungry AI factories. At that scale, the question isn’t “who has the most FLOPs?” It’s: Who can deliver the most intelligence per watt and per dollar? Nvidia’s GPUs still dominate, but inside Google there’s a different answer: the Tensor Processing Unit (TPU):

Computer

31,075 Aufrufe • vor 7 Monaten

NVIDIA $NVDA CEO, Jensen Huang just unveiled the N1X chip at Asia's largest tech expo Co-developed with MediaTek, this beast is built on the Blackwell architecture and runs the entire NVIDIA software stack flawlessly. "A chip that would take 33 years to build." Looks like future of AI hardware has officially arrived.
1:15

NVIDIA $NVDA CEO, Jensen Huang just unveiled the N1X chip at Asia's largest tech expo Co-developed with MediaTek, this beast is built on the Blackwell architecture and runs the entire NVIDIA software stack flawlessly. "A chip that would take 33 years to build." Looks like future of AI hardware has officially arrived.

The AI Colony R&D

22,916 Aufrufe • vor 1 Monat