Patel says GPT-4 used 24,000 GPUs, GPT-5 uses 100,000, and Microsoft is building toward a million GPUs.
“GPT four was trained with 24,000 GPUs roughly, and GPT five is on the order of a 100,000. And then they're trying to build this data center over the next few years. That's a million.”
Patel reveals OpenAI signed a deal with Oracle for 200 megawatts because Microsoft cannot provide enough compute capacity.
“Again, a gigawatt is like 500,000 plus GPUs. That's billions, tens of billions of dollars. So OpenAI signed a deal for just 200 megawatts.”
Patel says CoreWeave can build $10 billion clusters from start to completion much faster than competitors.
“it lets you get to these, you know, call it $10,000,000,000 clusters, right, in in time scales, from project start to completion that are much much shorter than anyone else.”
Patel notes CoreWeave got H100 services online six months before every hyperscaler in many cases.
“Get services online with h one hundreds before every hyperscaler by a factor of, like, as much as, like, six months in many cases.”
Patel argues competing with NVIDIA requires leaping forward beyond supply chain advantages across networking, memory, manufacturing and components.
“They're gonna have better negotiations with whether it's TSMC or SK Hynix and the memory and silicon side or all the rack people or, like, copper cables, everything, they're gonna have better cost efficiency.”
Patel says GPU hardware and networking represent 80% of data center costs, with power and cooling only 20%.
“It's the GPU purchases. It's the networking. It's the it's the physical data center conversion power conversion equipment. All of this stuff is, like, 80% of the cost.”
Patel lists NVIDIA's structural advantages across networking, HBM, process nodes, speed to market, and supplier negotiations.
“NVIDIA's gonna have better networking than you. They're gonna have better, HBM. They're gonna have better process node. They're gonna come to market faster.”
Patel says OpenAI and Anthropic are receiving 30% of all GPU chips being produced this year.
“30% of the chips are going to them, just those two companies. But that's actually like, okay, well, 70% of the stuff, who's making off well, one third of it is ads, whether it be ByteDance or Meta or many of the other people who are doing ads.”
Patel explains competitors need 5x hardware advantage over NVIDIA but risk failure if AI workloads shift before shipping.
“So now you need to do something, you know, that will give you five x advantage, right, in hardware efficiency for a certain type of workload, and then pray the workload doesn't shift.”
Patel shows B200 has 15x raw performance advantage over H100 but only 10x performance per TCO.
“If we don't divide by TCO, then it looks like the performance of B200 is actually 15x that of H100, versus the performance TCO is only 10x,”
Patel reports GB200 is 10x more power efficient than H200 at certain interactivity rates.
“But it turns out at certain interactivity rates, I. E. Tokens per second per user, it's 10x more efficient per watt, right, compared to H200.”
Patel notes NVIDIA Blackwell performance improved dramatically from launch to present, as did AMD hardware.
“If you tried to use NVIDIA's Blackwell six months ago, the numbers were not amazing, right? But now they're actually amazing. So how did that progress over time? Same with AMD, right?”
Patel finds AMD MI355 beats NVIDIA B200 on performance TCO in certain publicly usable configurations.
“So we do different scenarios. We do document processing, which is 8,000 context in, 1,000 out. We do chat, which is 1,000 in, 1,000 out.”
Patel identifies GB200's power efficiency advantage but notes deployment challenges with backplane and liquid cooling.
“GB200 has a huge power efficiency advantage, right? Everyone here understands the challenges of running and deploying GB200. There's a lot of challenges with the backplane.”
Patel says the standard inference deployment unit has shifted from single nodes to hundreds of GPUs.
“the standard unit for an inference deployment being hundreds of GPUs instead of a single node. And then there's all these different things about traffic.”
Patel states OpenAI has over 500,000 GPUs for R&D but GPT-5 pre-training uses under 100,000 GPUs.
“OpenAI has over 500,000 GPUs working on r and d. Right? Let's call it r and d.”
Patel predicts Google will have zero cash flow profit in 2027, spending all revenue on AI infrastructure.
“There's no reason why Google will have any profit in '27 at all, right, in terms of cash flow. They will just spend every dollar they make on on AI infrastructure.”
Patel calculates Anthropic added $1.5 billion in compute infrastructure in one month to serve new revenue growth.
“If Anthropic added 2,500,000,000 of revenue, and their gross margin is 40%, they added like $1,500,000,000 of compute in one month. Mhmm. Right?”
Patel details that producing one gigawatt of Rubin chips requires specific wafer volumes across multiple manufacturing nodes.
“a gigawatt of, you know, NVIDIA's Rubin chips. Right? So Rubin is announced at GTC, I believe, the week this podcast goes live.”
Patel argues NVIDIA is setting up supply chain to manufacture tens of millions of AI chips.
“I can't buy millions, tens of millions, which is what Jensen's, setting his supply chain up for.”
Patel says NVIDIA has locked up over 60% of supply chain capacity this year in long-term contracts.
“The NVIDIA has locked up over 60% of the capacity this year in long term contracts alone, and they're buying more on top of that.”
Patel reports NVIDIA is negotiating over $250 billion in supply contracts across components this year.
“And as you look at what they're negotiating in the market today, there's over 250,000,000,000 of wafers, of memory, of substrates, of PCBs, of networking equipment that they're going to sign this year,”
Patel calls Samsung's capacity planning organizational structure the most retarded ever
“Learned about how Samsung does capacity planning. Most retarded organizational structure ever”
Patel says no ASML lithography tool can be made or operated without US supply chain
“@voidtarget @lithos_graphein Machine = tool And no, no ASML litho tool can be made and operated without US supply chain.”