Jensen Huang on Nvidia's Moat: Supply Chains, TPUs, CUDA, and the Case for Selling Chips to China
Dwarkesh Patel
In this long interview, Dwarkesh presses Nvidia CEO Jensen Huang on whether Nvidia's dominance can last. The questions cover whether Nvidia is really a software company that others manufacture for, whether its hold on scarce components is the true moat, whether Google's TPUs and other custom accelerators threaten it, why Nvidia doesn't become a cloud provider, and, at greatest length, whether the United States should allow Nvidia to sell AI chips to China. Huang's position throughout is that Nvidia's advantage is not any single factor. In his account it comes from a combination of programmability, ecosystem, install base, supply-chain reach, and an annual cadence of large performance gains. On China, the two sides of the conversation never fully converge.
Is Nvidia a Software Company That Can Be Commoditized?
Dwarkesh opens with a deliberately naive framing. Software company valuations have fallen because people expect AI to commoditize software. Nvidia sends a GDS2 design file to TSMC, which builds the logic dies and switches and packages them with HBM from SK Hynix, Micron, and Samsung. An ODM in Taiwan then assembles the racks. If Nvidia is fundamentally making software that other people manufacture, does it get commoditized too?
Huang says the framing matches his own mental model of the company: electrons go in, tokens come out, and Nvidia sits in the middle. He argues that turning electrons into tokens, and making tokens more valuable over time, is hard to commoditize. He compares it to making one molecule more valuable than another, and he says the artistry, engineering, and science involved are "far from deeply understood" and "the journey is far from over." The company's rule is to do "as much as necessary and as little as possible." Whatever Nvidia doesn't need to do, it hands to partners. He describes AI as a "five-layer cake" and says Nvidia has partner ecosystems across all five layers. The part Nvidia must do itself, he says, "is insanely hard," and he doesn't think it gets commoditized.
He also disputes the premise that software companies will be commoditized. He sees many of them as toolmakers, citing Excel, PowerPoint, Cadence, and Synopsys. His prediction runs the other way: the number of agents will grow exponentially, and so will the number of tool users. He expects the number of instances of tools like Synopsys Design Compiler, floor planners, layout tools, and design rule checkers to "skyrocket," because engineers today are limited by headcount and will soon be supported by many agents exploring the design space. In his view this hasn't happened yet only because agents aren't good enough at using tools. He expects it to come through some combination of software companies building their own agents and agents getting better.
Purchase Commitments and "Informing" the Supply Chain
Dwarkesh cites Nvidia's filings showing close to $100 billion in purchase commitments with foundries, memory makers, and packaging suppliers. He adds SemiAnalysis's report that the figure will reach $250 billion. Could the real moat be that Nvidia has locked up years of scarce components, so a rival with a good accelerator can't get the memory or logic to build it?
Huang accepts this as one thing Nvidia can do that others find hard, but he describes it as more than contracts. Some commitments are explicit. Others are implicit: suppliers make investments because he has personally explained to their CEOs how big the industry will be and why. They are willing to invest on Nvidia's behalf rather than someone else's, he says, because they know Nvidia has the downstream demand to buy their supply and sell it through. He compares this to cash flow. There is also "supply chain flow," and nobody builds a supply chain for an architecture whose business churns slowly.
He presents GTC, Nvidia's conference, as part of this effort. It puts upstream suppliers, downstream customers, and AI startups in one place so they can see each other and see firsthand what he has been telling them. He acknowledges that his keynotes can feel "a little torturous," closer to education than a string of announcements, and says that is deliberate. He wants the whole ecosystem to understand what is coming, when, and how big it will be, and to "reason about it systematically, just like I reason about it." If Nvidia's next several years reach a trillion dollars in scale, he says, the supply chain is there to support it.
Can the Upstream Keep Up With Doubling?
Dwarkesh pushes on the physical limits. Nvidia has roughly doubled revenue year over year and more than tripled the flops it provides annually. Citing SemiAnalysis, he notes that AI will account for about 60% of TSMC's N3 node this year and 86% next year. How do you double when you're already the majority, and how do you get twice as many EUV machines each year?
Huang argues that instantaneous demand exceeding supply is a healthy condition for an industry. When one component falls far behind, the industry "swarms it." His example is CoWoS packaging. People no longer talk about it much, he says, because for two years Nvidia "swarmed the living daylights out of it," doubling capacity several times over. TSMC now scales CoWoS and future packaging technologies in step with logic, and CoWoS and HBM, once specialty technologies, are now mainstream.
He says he was making today's predictions five years ago, and that some suppliers believed him early. He singles out Sanjay and the Micron team, recalling a meeting where he laid out what was going to happen. Micron doubled down across LPDDR and HBM, which he says has been "tremendous for the company." Others came later. Nvidia now tries to "prefetch" bottlenecks years in advance. Examples include investments with Lumentum, Coherent, and the silicon photonics ecosystem, which he says reshaped the supply chain around TSMC; partnering with TSMC on COUPE and licensing the resulting patents to keep the supply chain open; and developing new testing equipment such as double-sided probing.
Asked directly about EUV, Huang says none of this is impossible to scale quickly and all of it is "easy to do within two or three years" given a demand signal: "Once you can build one, you can build ten, and once you can build ten, you can build a million." He doesn't always go to ASML directly. "If I can convince TSMC, ASML will be convinced." His claim is that no bottleneck lasts longer than two or three years. Meanwhile Nvidia improves computing efficiency by 10x to 20x per generation, and in his account by 30x to 50x from Hopper to Blackwell.
What worries him is downstream: energy, and energy policy in particular. You can't reindustrialize the United States, bring back chip and packaging manufacturing, or build EVs, robots, and AI factories without energy, and those things take a long time. Asked which bottleneck is hardest, he answers "plumbers and electricians." This leads into a criticism of "doomers" who predict the end of work. If people are discouraged from becoming software engineers, he warns, there will be a shortage of software engineers. He notes that people were told a decade ago not to become radiologists, and "guess what we're short of? Radiologists." Dwarkesh remarks that other guests have told him the opposite about which bottlenecks are hard, and that he lacks the technical knowledge to adjudicate.
TPUs and the Case for General Programmability
Dwarkesh points out that arguably two of the top three models, Claude and Gemini, were trained on TPUs. Huang's first answer is that Nvidia builds something different: accelerated computing, not a tensor processing unit. Its GPUs are used for molecular dynamics, quantum chromodynamics, data processing, fluid dynamics, and particle physics as well as AI, so its market reach is "far greater than any TPU or ASIC can possibly have." Because Nvidia's systems are designed to be operated by others, they run in every cloud, including Google, Amazon, Azure, and OCI. They can also be run by a company for itself, as with Elon Musk's xAI, or used for a drug-discovery supercomputer at Lilly. By contrast, he says, most "home-built systems" require the builder to be the operator.
Dwarkesh presses the point. Nvidia isn't earning $60 billion a quarter from pharma and quantum. It's earning it from AI, and his AI researcher friends describe the TPU as a big systolic array ideal for matrix multiplies. GPUs spend die area on flexibility, such as warp schedulers and switches between threads and memory banks, that AI's predictable workloads may not need.
Huang replies that matrix multiplies are important but not the whole story. Inventing a new attention mechanism, disaggregating computation differently, building a hybrid SSM, or fusing diffusion with autoregressive techniques all call for a generally programmable architecture, and he argues that the ability to invent new algorithms is what makes AI advance so quickly. TPUs, like everything else, are subject to Moore's Law, which he puts at roughly 25% improvement per year. Getting 10x or 100x leaps requires changing the algorithm and how it is computed. He recounts announcing that Blackwell would be 35x more energy efficient than Hopper. Nobody believed it, and then Dylan (of SemiAnalysis) wrote that he had "sandbagged" and the real figure was 50x. That result, he says, came from models such as mixture-of-experts parallelized and distributed across the system, from new CUDA kernels, and from "extreme co-design," including offloading computation into the NVLink fabric or the Spectrum-X network. "Without CUDA to do that, I wouldn't even know where to start."
Does CUDA Matter to Customers Who Write Their Own Kernels?
Dwarkesh notes that about 60% of Nvidia's revenue comes from five hyperscalers. Unlike professors running experiments, these customers can write their own kernels. OpenAI has Triton and its own stack rather than relying on cuBLAS and NCCL, and that stack compiles to other accelerators. If the biggest customers can replace CUDA, how much does it matter?
Huang gives three reasons CUDA remains valuable. First, the ecosystem is rich and well tested. Nvidia supports every framework and contributes heavily to Triton, whose back end he says contains "huge amounts of Nvidia technology." He also lists vLLM, SGLang, and new reinforcement learning frameworks such as verl and NeMo RL. Because the foundation is so well "wrung out," a developer who hits a bug can reasonably assume it is in their own code rather than in the mountain of code underneath. Second, the install base: several hundred million GPUs across generations and form factors, including in robots, so software written for CUDA runs nearly everywhere. Third, Nvidia is in every cloud and on-prem, so an AI company that hasn't settled on a cloud partner can run anywhere.
Margins, Kernel Engineering, and Performance per TCO
Dwarkesh sharpens the question. If AI gets especially good at tasks with tight verification loops, and writing an efficient attention or MLP kernel is exactly that kind of task, hyperscalers could write their own kernels. Competition would then reduce to flops and memory bandwidth per dollar. Could Nvidia sustain margins above 70%?
Huang describes CPUs as a Cadillac, a comfortable cruiser anyone can drive, and Nvidia's accelerators as F1 cars that anyone can drive at 100 mph but that take real expertise to push to the limit. Nvidia assigns an "insane" number of engineers to AI labs to optimize their stacks, and it uses "a ton of AI" to create its own kernels. He says it is not unusual for this work to speed up a lab's model by 50%, 2x, or 3x. Across a lab's installed fleet of Hoppers and Blackwells, a 2x gain translates directly into doubled revenue.
He then makes his strongest competitive claim: Nvidia offers "the best performance per TCO in the world, bar none." He challenges competitors to prove otherwise on public benchmarks, naming Dylan's InferenceMAX and MLPerf. "TPU won't come, Trainium won't come," he says, and he invites Trainium to demonstrate the 40% advantage he says it "claim[s] all the time," and Google to demonstrate the claimed cost advantage of TPUs. "On first principles, it makes no sense."
He also disputes the idea that the hyperscaler concentration means a few internal buyers. Most Nvidia capacity at AWS serves external customers, he says, and at Azure and OCI all of it does. The clouds favor Nvidia because it brings them the most customers. He describes the flywheel this way. Among tens of thousands of AI startups, each would choose the most abundant architecture, the largest install base, and the richest ecosystem. Nvidia also offers the lowest-cost tokens through performance per dollar and the most tokens per watt, which matters because a one-gigawatt data center's revenue depends on how many tokens it produces. And for anyone renting out infrastructure, Nvidia has the most customers.
Anthropic, ASIC Margins, and a Missed Investment
Dwarkesh replies that the compute on these clouds is actually used by a few labs, such as Anthropic and OpenAI, that can make other accelerators work. Huang says the premise is wrong and asks to come back to it, because it is "too important to AI." Dwarkesh then asks about Anthropic's recently announced multi-gigawatt deal with Broadcom and Google for TPUs, which would make up the majority of its compute.
Huang calls Anthropic "a unique instance, not a trend." Without Anthropic, he asks, why would there be any TPU growth? "It's 100% Anthropic." He says the same about Trainium. When Dwarkesh raises OpenAI's deals with AMD and its own accelerator effort, Huang says OpenAI is still "vastly Nvidia." He says he isn't offended when customers try alternatives, since that is how they learn how good Nvidia is, and Nvidia must keep earning its position. He points to the number of canceled ASIC projects as evidence that building something better than Nvidia is hard.
Dwarkesh suggests the logic of an ASIC is that it only needs to be no more than 70% worse, given Nvidia's margins. Huang counters that ASIC vendors' margins are also high, around 65% by his estimate, so "what are you really saving?" He adds that ASIC makers are "quite proud of their incredible ASIC margins."
He then offers his own explanation for Anthropic's path, framed as a personal mistake. Early on he did not "deeply internalize" how hard it would be to build a foundation lab, or that labs like Anthropic needed huge investments from their suppliers. Google and AWS made large early investments, and Anthropic used their compute in return. Nvidia had never invested outside the company at that scale, and he assumed the labs could raise from VCs "like all companies do." He now recognizes that no VC would have put $5–10 billion into a lab on the hope it would become Anthropic. Even if he had understood this, he doesn't think Nvidia could have done it at the time. "I'm not going to make that same mistake again," he says, citing Nvidia's investment in OpenAI and later in Anthropic once it was in a position to invest. Though Nvidia's absence pushed Anthropic elsewhere, he says Anthropic's existence "is great for the world."
Why Nvidia Won't Become a Hyperscaler
Dwarkesh cites reports of Nvidia investing up to $30 billion in OpenAI and $10 billion in Anthropic, and backstopping CoreWeave for up to $6.3 billion on top of a $2 billion investment. With so much cash, why not become a cloud and rent compute directly?
Huang returns to "as much as needed, as little as possible." Nvidia should do the work that nobody else would do: building NVLink, building the whole stack, and "20 years of CUDA while losing money most of that time." He also points to the domain-specific CUDA-X libraries Nvidia began building about 15 years ago, for ray tracing, image generation, early AI, and structured and vector data processing, and to cuLitho for computational lithography. "If we didn't create it, nobody would have. I am completely certain of that." The world already has many clouds, however, and if Nvidia didn't build one, someone else would.
Dwarkesh points out an apparent contradiction: Huang says CoreWeave, Nscale, and Nebius wouldn't exist without Nvidia, which sounds like propping them up. Huang explains that these companies first had to want to exist and bring a business plan, expertise, and their own capabilities. If they needed investment to get off the ground, Nvidia would help, but it does not want to be in the financing business and prefers to work with financiers. Large investments like the one in OpenAI, which he describes as needed at "$30 billion scale" before its IPO, are made because the company needs them, not because Nvidia is trying to do more.
He adds that Nvidia deliberately avoids picking winners among model companies: "when I invest in one of them, I invest in all of them." His reason is humility drawn from history. When Nvidia started there were about 60 3D graphics companies, and Nvidia would have been at the top of the list of those not expected to survive. Its early graphics architecture was "precisely wrong," reasoned from good first principles but impossible for developers to support. Yet it was the only one that survived.
How Scarce GPUs Are Allocated
Dwarkesh describes Nvidia's reputation for splitting scarce supply among neoclouds like CoreWeave, Crusoe, and Lambda rather than selling to the highest bidder. Huang rejects the characterization. The first step is joint forecasting with customers, because chips and data centers take a long time to build. After that, "you still have to place an order," and allocation is essentially first in, first out. The main exception is that if a customer's data center or other components aren't ready, Nvidia may serve another customer first to maximize the throughput of its own factory.
He says the story that Larry Ellison and Elon Musk begged him for GPUs over dinner "never happened." They did have dinner, and it was a wonderful one, but "they just had to place an order."
Asked why Nvidia doesn't simply sell to the highest bidder, Huang calls it "a bad business practice." Nvidia sets a price and customers decide whether to buy, even if demand "goes through the roof." He acknowledges that others in the chip industry raise prices when demand spikes and says Nvidia never has. He ties this to dependability and compares it to Nvidia's nearly 30-year relationship with TSMC, which he says involves no legal contract: "sometimes I got a better deal, sometimes I got a worse deal," but he can completely trust and depend on them.
He extends the dependability claim to Nvidia's roadmap. Vera Rubin arrives this year, Vera Rubin Ultra next year, Feynman the year after, and then an as-yet-unnamed generation. He says customers can count on token cost decreasing "by an order of magnitude every single year," as reliably as a clock, and that no ASIC team can make that promise. Customers can order one graphics card or $100 billion of AI factory capacity, which he says only Nvidia (and, at the foundry level, TSMC) can offer. He says this position took "a couple of decades" of commitment to reach.
China, Part One: Mythos and the Cyber-Offense Argument
Dwarkesh says he is unsure whether selling chips to China is good and is playing devil's advocate, having argued the opposite side with Anthropic's Dario Amodei. He cites Anthropic's recently announced Mythos Preview. Anthropic is withholding it from public release because of its cyber-offensive capability, saying it found thousands of high-severity vulnerabilities across every major operating system and browser, including a 27-year-old bug in OpenBSD. If Chinese labs had the chips to train such a model and run millions of instances, would that threaten American security?
Huang says Mythos was trained on "fairly mundane capacity," in a fairly mundane amount, and that this kind of compute is "abundantly available in China." He asserts that China manufactures 60% or more of the world's mainstream chips, has abundant energy, and has about half of the world's AI researchers. He also says most researchers in American AI labs are Chinese. Given those assets, he argues that "victimizing them, turning them into an enemy" is probably not the safest approach. China is an adversary and the US should win, but he calls research dialogue "glaringly missing" and says American and Chinese researchers must talk and try to agree on what AI should not be used for.
On vulnerability discovery itself, he says finding bugs "is what AI is supposed to do." He highlights an ecosystem of AI security, privacy, and safety startups working toward a future in which one powerful agent is surrounded by thousands of agents keeping it safe. The idea of an agent "running around with nobody watching after it is kind of insane." That ecosystem, he argues, needs open source and open models, much of which comes from China, and the US "ought to not suffocate that." He considers it "extremely foolish" to end up with an open-source ecosystem running on a foreign tech stack and a closed one running on the American stack.
China, Part Two: Flops, 7nm, and Whether Energy Substitutes for Chips
Dwarkesh responds with estimates that, lacking EUV, China has about one-tenth the flops of the US. More compute lets American labs reach capabilities first. He notes that Anthropic is holding Mythos back while American companies patch vulnerabilities, and that inference compute determines whether an attacker runs a thousand hackers or a million. He adds that leaders at DeepSeek and Qwen have said they are bottlenecked on compute, so China's strong researchers make additional compute more dangerous, not less.
Huang agrees the US should always be first and have more, but says Dwarkesh's argument only works if China has no compute at all. China is the second-largest computing market in the world. AI is a parallel problem, and with abundant, essentially free energy, China can gang together 4x or 10x as many 7nm chips. He claims China has "ghost datacenters" sitting fully powered and empty, and overcapacity in mainstream chip manufacturing. The US, being energy-scarce, needs performance per watt, but "if your amount of watts is completely abundant… what do you care about performance per watt for?" He says 7nm chips are "essentially Hopper," and that today's models were largely trained on the Hopper generation. As evidence of volume, he cites Huawei's record year and shipments of "millions" of chips, "way more than Anthropic has."
Dwarkesh raises memory bandwidth: HBM2 versus the newest memory could mean close to an order-of-magnitude gap, and he believes advanced HBM requires EUV. Huang calls this "not at all true." He says Huawei, as a networking company, can gang chips together the way Nvidia does with NVL72, and has demonstrated silicon photonics connecting compute into a single supercomputer. He adds that compute-limited researchers invent smarter algorithms, and argues that most AI advances, including MoE and efficient attention mechanisms, came from algorithms rather than raw hardware. "DeepSeek is not an inconsequential advance."
He then states what he sees as the real danger: "The day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation." Dwarkesh asks why, since open models can run on any accelerator. Huang says models optimized for one architecture run worse on others: "I am the evidence." Nvidia's success shows that models created on its stack run best on its stack. If the world's AI models come out of the box running best on non-American hardware, particularly in the Global South and the Middle East, he calls that bad for the US. Dwarkesh presses that the hypothetical still has China finding American vulnerabilities first, merely on Nvidia hardware. Huang agrees that would not be good, "so let's not let it happen." He argues that if the next few years are critical, all the world's AI models should be built on the American tech stack during those years, while acknowledging there is "no guarantee either way."
China, Part Three: Analogies, Lock-In, and "Conceding the Market"
The exchange becomes more heated. Huang asks why one layer of the AI industry should lose an entire market to benefit another layer, and why Dwarkesh is "so fixated" on the model layer. He says the US has 100x more compute than anywhere else, that Nvidia gives US labs first access to new technology, and that Vera Rubin is for the United States. Nvidia is an American company, he argues, so why not write a more balanced regulation that lets it win around the world instead of "giving up the world"?
Dwarkesh brings up Dario Amodei's analogy comparing chip sales to China with Boeing boasting that it supplied the missile casings for North Korean nukes, and then compares compute to enriched uranium. Huang calls this "lunacy," a "lousy analogy," and "illogical." He says the answer to misuse is dialogue with researchers, with China, and with all countries, combined with ensuring the US has "mountains" of Vera Rubin and Blackwell. He says the same logic could be applied to microprocessors, DRAM, or electricity. Dwarkesh replies that the US does in fact export-control advanced DRAM-making technology and chipmaking equipment.
Dwarkesh then raises lock-in. Tesla and Apple sold excellent products in China without preventing Chinese EVs and smartphones from coming to dominate. Huang answers, "We're not a car." Switching car brands is easy, but computing ecosystems like x86 and ARM are sticky and costly to replace. He says 50% of AI developers are in China and the US should not give them up. When Dwarkesh argues that American labs already use multiple accelerators and Chinese labs could too, Huang says Nvidia's share is growing, not shrinking, and rejects the premise that it would lose China anyway: "You're not talking to somebody who woke up a loser."
Dwarkesh identifies what looks like a contradiction: Huang claims Nvidia would beat Huawei because its chips are better, and also that China would do the same things without Nvidia. Huang says both are true, because "in the absence of a better choice, you'll take the only choice you have." When Dwarkesh equates "better" with more compute, Huang says it is better because it is easier to program and has a better ecosystem. "And of course we're going to send them compute. So what?" He lists the benefits as American technology leadership, developers working on the American stack, and that stack being the best choice as models diffuse worldwide. He points to the American telecommunications industry, which he says was "policied out of basically the world" to the point that the US no longer controls its own telecommunications, and calls the current approach "a little narrow-minded." He says China accounts for about 40% of the world's technology industry and that Nvidia once had a large share of that market and no longer does.
Dwarkesh asks Huang to acknowledge a cost: compute feeds powerful models, and if China had reached Mythos-level capability first and deployed it widely, that would have been very bad. Huang responds with a cost of his own. Conceding the second-largest market lets China build scale and its own ecosystem, so future models are optimized for a different stack. Because Chinese models are open, their standards could become superior as AI spreads. Dwarkesh says he trusts Nvidia's kernel engineers and techniques like distillation to prevent long-term lock-in. Huang replies that AI is "more than kernel optimization," and lists what he calls facts: China is the largest contributor to open-source software and to open models, and today those are built on Nvidia's stack.
He broadens the argument into a critique of fear-driven framing. He says scaring the country into treating AI as a nuclear bomb, or scaring people away from software engineering or radiology, does the US a disservice. On radiology he distinguishes a job from a task: "The job of a radiologist is patient care. The task is to read a scan." He makes an explicit prediction. When the US wants to export its technology and standards to India, the Middle East, Africa, and Southeast Asia, he wants to revisit this conversation, and he expects to show how this policy "caused the United States to concede the second largest market in the world for no good reason at all." He says nobody is advocating shipping everything to China. The US should always have the best and most technology first while also competing worldwide, which he says "requires some amount of nuance, some amount of maturity instead of absolutes."
In a final round, Dwarkesh argues that for Chinese chips to set global standards, 7nm chips would have to beat Nvidia's future 1.6nm chips on exported models. Huang answers with his own numbers. The transistor improvement from Hopper to Blackwell, three years apart, was "call it 75%," while Blackwell delivers 50x Hopper. "Moore's Law is dead," and architecture, networking, energy, and computer science matter most, which he says is why Nvidia bought Mellanox. He says export restrictions have already backfired by accelerating China's chip industry and pushing its AI ecosystem toward domestic architectures. It is "not too late," but it has already happened. He adds that China will not stay stuck at 7nm, and that 7nm versus 5nm is not a 10x difference.
Would Nvidia Go Back to Older Process Nodes?
Returning to supply constraints, Dwarkesh asks whether Nvidia might, before 2030, use spare N7 capacity to build a Hopper- or Ampere-class chip updated with modern numerics. Huang says it is not necessary. Each generation involves so much packaging, stacking, numerics, and system architecture work beyond the transistors that porting back to an older node is "a level of R&D that no one could afford." Nvidia can afford to lean forward, not back. Only in a thought experiment where the world would never gain more capacity would he go back to 7nm, and then "in a heartbeat."
Why Not Pursue Several Radically Different Architectures?
Dwarkesh asks why Nvidia doesn't hedge by running parallel projects, such as Cerebras-style wafer-scale chips, Dojo-style large packages, or designs without CUDA. Huang says Nvidia could, but has no better idea. Nvidia simulates these alternatives and finds them "provably worse." It would add a different accelerator only if the workload changed dramatically, meaning the shape of the market rather than the algorithms.
He says that is what has now happened with Groq, which Nvidia recently added and will fold into the CUDA ecosystem. His reasoning is that tokens have become valuable enough to support different price tiers. A couple of years ago tokens were nearly free. Now some customers, such as highly paid software engineers, would pay for much more responsive tokens. Nvidia therefore wants to extend the Pareto frontier with a faster-response, lower-throughput inference segment, betting that high average selling prices per token can make up for lower factory throughput. Apart from that, he says, more money would go into Nvidia's existing architecture.
What Nvidia Would Be Without Deep Learning
In the closing question, Dwarkesh asks what Nvidia would be doing if the deep learning revolution hadn't happened. Huang answers: accelerated computing, as always. The company's premise is that general-purpose computing has largely run its course for many workloads, and that pairing a GPU running CUDA with a CPU to offload kernels can speed applications by 100x to 200x. Early domains included computer graphics, molecular dynamics, seismic processing for energy discovery, and image processing, and he lists particle physics, fluids, and structured data processing among others. Even without AI, he says, Nvidia "would be very, very large."
He adds that without AI he "would be very sad," but that Nvidia's computing advances are what democratized deep learning by letting any researcher or student do science on a PC or a GeForce card. He notes that the opening part of every GTC keynote, covering computational lithography, quantum chemistry, and data processing, has nothing to do with AI. He ends by saying that much important work is not AI-related and that "tensors are not the only way that you compute."
We've seen the valuations of a bunch of software companies crash because people are expecting AI to commoditize software. There's a potentially naive way of thinking about things, which is: look, Nvidia sends a GDSII file to TSMC. TSMC builds the logic dies, it builds the switches, then it packages them with the HBM that SK Hynix, Micron, and Samsung make. Then it sends it to an ODM in Taiwan where they assemble the racks. Nvidia is fundamentally making software that other people are manufacturing, and if software gets commoditized, does Nvidia get commoditized?
In the end, something has to transform electrons to tokens. The transformation of electrons to tokens and making those tokens more valuable over time is hard to completely commoditize.
The transformation from electrons to tokens is such an incredible journey. Making that token is like making one molecule more valuable than another molecule, making one token more valuable than another. The amount of artistry, engineering, science, and invention that goes into making that token valuable, obviously we're watching it happen in real time. The transformation, the manufacturing, all of the science that goes in there is far from deeply understood and the journey is far from over. I doubt that it will happen. We're going to make it more efficient, of course.
The way that you framed the question is my mental model of our company. The input is electrons, the output is tokens. In the middle is Nvidia. Our job is to do as much as necessary and as little as possible to enable that transformation to be done at incredible capabilities. What I mean by "as little as possible," whatever I don't need to do, I partner with somebody and make it part of my ecosystem.
If you look at Nvidia today, we probably have the largest ecosystem of partners, both in the supply chain upstream and downstream, all of the computer companies, application developers, and model makers. AI is a five-layer cake, if you will. We have ecosystems across the entire five layers. We try to do as little as possible, but the part that we have to do, as it turns out, is insanely hard. I don't think that gets commoditized.
In fact, I also don't think the enterprise software companies, the tools makers… Most software companies today are tool makers. Some of them are not. Some of them are workflow codification systems. But for a lot of companies, they're tool makers. For example, Excel is a tool, PowerPoint is a tool, Cadence makes tools, Synopsys makes tools. I actually see the opposite of what people see.
I think the number of agents is going to grow exponentially, and the number of tool users is going to grow exponentially. It's very likely that the number of instances of all these tools is going to skyrocket. It's very likely that the number of instances of Synopsys Design Compiler is going to skyrocket, along with the number of agents using the floor planners, our layout tools, and our design rule checkers.
Today we're limited by the number of engineers. Tomorrow, those engineers are going to be supported by a bunch of agents. We're going to be exploring the design space like you've never seen before, and we're going to use the tools that we use today. I think tool use is going to cause the software companies to skyrocket. The reason why it hasn't happened yet is because the agents aren't good enough at using their tools yet. Either these companies are going to build the agents themselves, or agents are going to get good enough to be able to use those tools. I think it's going to be a combination of both.
I think in your latest filings, you had almost $100 billion in purchase commitments with foundries, memory, and packaging. SemiAnalysis has reported that you will have $250 billion of these kinds of purchase commitments. One interpretation is that Nvidia's moat is really that you've locked up many years of these scarce components. Somebody else might have an accelerator, but can they actually get the memory to build it? Can they actually get the logic to build it? Is this really Nvidia's big moat for the next few years?
It's one of the things that we can do that is hard for someone else to do. We've made enormous commitments upstream. Some of it is explicit, these commitments that you mentioned. Some of it is implicit. For example, a lot of the investments that are upstream are made by our supply chain because I said to the CEOs, "Let me tell you how big this industry is going to be, let me explain to you why, let me reason through it with you, and let me show you what I see." As a result of that process of informing, inspiring, and aligning with CEOs of all different industries upstream, they're willing to make the investments.
Why are they willing to make the investments for me and not someone else? The reason for that is because they know that I have the capacity to buy their supply and sell it through my downstream. The fact is that Nvidia's downstream supply chain and our downstream demand is so large, they're willing to make the investment upstream.
If you look at GTC, people are marveled by the scale of it and the people that go. It's a full 360 degrees, the entire universe of AI all in one place. They're all in one place because they need to see each other. I bring them together so that the downstream can see the upstream, the upstream can see the downstream, and all of them can see the advances in AI. Very importantly, they can all meet the AI natives, all the AI startups being built, and all the amazing things happening so they can see firsthand all the things that I tell them.
I spend a lot of my time informing, directly or indirectly, our supply chain, partners, and ecosystem about the opportunity in front of us.
Some people always say, "Jensen, in most keynotes, it's one announcement after another." With our keynotes, there's always a part of it that's a little torturous in the sense that it almost comes across like education. In fact, that's exactly on my mind. I need to make sure the entire supply chain, upstream and downstream, the ecosystem, understands what is coming at us, why it's coming, when it's coming, how big it's going to be, and is able to reason about it systematically, just like I reason about it.
Regarding the moat as you describe it, we're able to build for a future. If our next several years are a trillion dollars in scale, we have the supply chain to do it. Without our reach, the velocity of our business… Just as there's cash flow, there's supply chain flow, there's churns. Nobody is going to build a supply chain for an architecture if the business churns are low. Our ability to sustain the scale is only because our downstream demand is so great. And they see it, they hear about it, they see it all coming. That allows us to do the things we're able to do at the scale we do them.
I do want to understand more concretely whether the upstream can keep up. For many years now, you guys have been 2x-ing revenue year over year. You've been more than tripling the amount of flops you're providing to the world year over year. And 2x-ing at this scale now is really incredible.
Exactly. But then you look at logic. You're the biggest customer on TSMC's N3 node, and you're one of the biggest on N2. AI as a whole this year is going to be 60% of N3. It's going to be 86% next year, according to SemiAnalysis. How do you double if you're the majority? And how do you do that year over year? Are we in a regime now where the growth rate in AI compute has to slow because of upstream? Do you see a way to get around this? How do we build 2x more fabs year over year, ultimately?
At some level, the instantaneous demand is greater than the supply upstream and downstream in the world. At any instant, we could be limited by the number of plumbers, which actually happens. The plumbers are invited to next year's GTC. By the way, great idea. But that's a good condition. You want an industry where the instantaneous demand is greater than the total supply of the industry. The opposite is obviously less good.
If we're too far apart, if one particular component is too far away, the industry swarms it. For example, notice people aren't talking very much about CoWoS anymore. The reason for that is because for two years we swarmed the living daylights out of it. We doubled, doubled, doubled on several doubles. Now I think we're in fairly good shape. TSMC now knows that CoWoS supply has to keep up with the rest of the logic demand and the memory demand. They're scaling CoWoS and future packaging technologies at the same level as they scale logic.
This is terrific, because for a long time, CoWoS and HBM memory were rather specialty. But they're not specialties anymore. People now realize they're mainstream computing technology.
Of course, we're now much more able to influence a larger scope of our supply chain. At the beginning of the AI revolution, all the things that I say now, I was saying five years ago. Some people believed in it and invested in it, for example, Sanjay and the Micron team. I still remember the meeting really well where I was clear about exactly what was going to happen, why it was going to happen, and the predictions of today. They really doubled down on it. We partnered with them across LPDDR and HBM memories, and they really invested in it. It obviously has been tremendous for the company. Some people came a little bit later, but now they're all here.
Each one of these bottlenecks gets a great deal of attention. Now we're prefetching the bottlenecks years in advance. For example, the investments that we've done with Lumentum, Coherent, and the silicon photonics ecosystem over the last several years really reshaped the supply chain. We built up an entire supply chain around TSMC. We partnered with them on COUPE, invented a whole bunch of technology, and licensed those patents to the supply chain to keep it nice and open. We're preparing the supply chain through the invention of new technologies, new workflows, new testing equipment like double-sided probing, investing in companies, and helping them scale up their capacity. You can see that we're trying to shape the ecosystem so that the supply chain is ready to support the scale.
It seems like some bottlenecks are easier than others. Scaling up CoWoS versus scaling up—
I went to the hardest one, by the way.
Which is?
Plumbers. Plumbers and electricians. This is one of the concerns that I have about the doomers describing the end of work and killing of jobs. If we discourage people from being software engineers, we're going to run out of software engineers. The same prediction happened ten years ago. Some of the doomers were telling people, "Whatever you do, don't be a radiologist." You might hear some of those videos still on the web saying radiology is going to be the first career to go and the world is not going to need any more radiologists. Guess what we're short of? Radiologists.
Going back to this point about how some things you can scale, and other things… How do you actually manufacture 2x the amount of logic a year? Ultimately, memory and logic are bottlenecked by EUV. How do you get to 2x as many EUV machines year over year?
None of that is impossible to scale quickly. All of that is easy to do within two or three years. You just need a demand signal. Once you can build one, you can build ten, and once you can build ten, you can build a million. These things are not hard to replicate.
How far down the supply chain do you go? Do you go to ASML and say, "Hey, if I look out three years from now, for Nvidia to be generating two trillion a year in revenue, we need way more EUV machines"?
Some of them I have to directly, some of them indirectly, and some of them… If I can convince TSMC, ASML will be convinced. We have to think about the critical pinch points. But if TSMC is convinced, you'll have plenty of EUV machines in a few years.
My point is that none of the bottlenecks last longer than a couple of years, two, three years, none of them. Meanwhile, we're improving computing efficiency by 10x, 20x, and in the case of Hopper to Blackwell, 30x to 50x. We're coming up with new algorithms because CUDA is so flexible. We're developing all kinds of new techniques so that we drive efficiency in addition to increasing capacity. None of those things worry me.
It's the stuff that's downstream from us. Energy policies that prevent energy from… You can't create an industry without energy. You can't create a whole new manufacturing industry without energy. We want to reindustrialize the United States. We want to bring back chip manufacturing, computer manufacturing, and packaging. We want to build new things like EVs and robots. We want to build AI factories. You can't build any of these things without energy, and those things take a long time. More chip capacity, that's a 2-3 year problem. More CoWoS capacity, 2-3 year problem.
Interesting. I feel like I have guests tell me the exact opposite thing sometimes. In this case, I just don't have the technical knowledge to adjudicate.
The beautiful thing is you're talking to the expert.
True. I want to ask about your competitors. If you look at the TPU, arguably two out of the top three models in the world, Claude and Gemini, were trained on TPU. What does that mean for Nvidia going forward?
We build a very different thing. What Nvidia built is accelerated computing, not a tensor processing unit. Accelerated computing is used for all kinds of things: molecular dynamics, quantum chromodynamics, data processing, data frames, structured data, and unstructured data. It's also used for fluid dynamics and particle physics. In addition, we use it for AI. Accelerated computing is much more diverse.
Although AI is the conversation today and is obviously very important and impactful, computing is much broader than that. Nvidia has reinvented the way computing is done, moving from general-purpose computing to accelerated computing. Our market reach is far greater than any TPU or ASIC can possibly have.
If you look at our position, we're the only company that accelerates applications of all kinds. We have a gigantic ecosystem. So all kinds of frameworks and algorithms run on Nvidia.
Because our computers are designed to be operated by other people, anyone who's an operator can buy our systems. With most of these home-built systems, you have to be your own operator because they were never designed to be flexible enough for others to operate. Because anybody can operate our systems, we're in every cloud, including Google, Amazon, Azure, and OCI. If you want to operate it to rent, you better have a large ecosystem of customers in many industries to be the offtakers. If you want to operate it for yourself, we obviously have the ability to help you operate it yourself, like we did for Elon with xAI.
And because we can enable operators in any company and any industry, you could use it to build a supercomputer for scientific research and drug discovery at Lilly. We can help them operate their own supercomputer and use it for the entire diversity of drug discovery and biological sciences that we accelerate.
There are just a whole bunch of applications that we can address that you can't do with TPUs. Nvidia built CUDA to be a fantastic tensor processing unit as well, but it also handles every life cycle of data processing, computing, AI, and so on. Our market opportunity is just a lot larger, and our reach is a lot greater. Because we support every application in the world now, you can build Nvidia systems anywhere and know that there will be customers for it. It's a very different thing.
This is going to be a long question. You have spectacular revenue, and you're not making $60 billion a quarter from pharma and quantum. You're making it because AI is an unprecedented technology that is growing unprecedentedly fast. The question then is what is best for AI specifically. I'm not in the details, but I talk to my AI researcher friends and they say, "Look,
when I use a TPU, it's this big systolic array that's perfect for doing matrix multiplies, whereas a GPU is very flexible. It's great when you have lots of branching or irregular memory access." But what is AI? It's just these very predictable matrix multiplies again and again and again. You don't have to give up any die area for warp schedulers or switches between threads and memory banks. And the TPU is really optimized for the bulk of this growth in revenue and use case for compute that is coming online right now. I wonder how you react to that.
Matrix multiplies are an important part of AI, but they're not the only part. If you want to come up with a new attention mechanism, disaggregate in a different way, or invent a whole new type of architecture altogether—like a hybrid SSM—you want an architecture that's generally programmable. If you want to create a model that fuses diffusion and autoregressive techniques, you want an architecture that's just generally programmable.
We run everything you can imagine. That's the advantage. It allows for the invention of new algorithms a lot more easily, because it's a programmable system. The ability to invent new algorithms is really what makes AI advance so quickly.
TPUs, like anything else, are impacted by Moore's Law, which we know is increasing by about 25% per year. The only way to really get 10x or 100x leaps is to fundamentally change the algorithm and how it's computed every single year. That's Nvidia's fundamental advantage. The only reason we were able to make Blackwell to Hopper 50x… When I first announced Blackwell was going to be 35x more energy efficient than Hopper, nobody believed it. Then Dylan wrote an article saying I sandbagged, and it's actually fifty times. You can't reasonably do that with just Moore's Law.
The way we solve that problem is with new models, like MoEs, that are parallelized, disaggregated, and distributed across a computing system. Without the ability to really get down and come up with new kernels with CUDA, it's really hard to do.
It's the combination of the programmability of our architecture and the fact that Nvidia is an extreme co-design company. We can even offload some of the computation into the fabric itself, like NVLink, or into the network with Spectrum-X. We could affect change across the processors, the system, the fabric, the libraries, and the algorithm simultaneously. Without CUDA to do that, I wouldn't even know where to start.
My sponsor Crusoe was among the first clouds to offer NVIDIA's Blackwell and Blackwell Ultra platforms. And they just announced their NVIDIA Vera Rubin deployment scheduled for later this year. But access to state-of-the-art hardware is only part of the story. For example, most inference engines already do KV caching for a single user's forward passes. But Crusoe does it across users and GPUs. So if a thousand agents are running on the same system prompt, Crusoe only has to compute the KV cache once for it to become available to every single GPU in the cluster. This is especially important as systems get more agentic and require much longer prefixes in order to use tools and access files. In a recent benchmark, Crusoe was able to deliver up to 10x faster time-to-first token and up to 5x better throughput than vLLM. This is just one among many reasons that you should run your inference workload with Crusoe. And if you need GPUs for training, you don't need to switch clouds. Crusoe's got you covered there too. Go to crusoe.ai/dwarkesh to learn more.
This gets at an interesting question about Nvidia's clientele. 60% of your revenue is coming from these big five hyperscalers. In a different era with different customers—let's say professors running experiments—they need CUDA. They can't use another accelerator. They just needed to run PyTorch with CUDA and have everything optimized. But these hyperscalers have the resources to write their own kernels. In fact, they have to in order to get that last 5% of performance they need for their specific architecture.
Anthropic and Google are mostly running their own accelerators or running TPUs and Trainium. But even OpenAI, using GPUs, has Triton because they need their own kernels. Down to CUDA C++, instead of using cuBLAS and NCCL, they've got their own stack which compiles to other accelerators as well. If most of your customers can and do make replacements for CUDA, to what extent is CUDA really the thing that is going to make frontier AI happen on Nvidia?
CUDA is a rich ecosystem. If you want to build on any computer first, building on CUDA first is incredibly smart. Because the ecosystem is so rich, we support every framework. If you want to create custom kernels… For example, we contribute enormously to Triton. So the back end of Triton has huge amounts of Nvidia technology. We're delighted to help every framework become as great as it can be.
There are lots and lots of frameworks. There's Triton, vLLM, SGLang, and more. Now there's a whole bunch of new reinforcement learning frameworks coming out, like verl and NeMo RL. With post-training and reinforcement learning, that entire area is just exploding. So if you want to build on an architecture, building on CUDA makes the most sense because you know the ecosystem is great.
You know that if something happens, it's more likely in your code and not in the mountain of code underneath. Don't forget the amount of code you're dealing with when building these systems. When something doesn't work, was it you or was it the computer? You would like it to always be you and to be able to trust the computer. Obviously, we still have lots of bugs ourselves, but our system is so well wrung out that you can at least build on top of the foundation. That's number one: the richness, programmability, and capability of the ecosystem.
The second thing is, if you're a developer building anything at all, the single most important thing you want is an install base. You want the software you write to run on a whole bunch of other computers. You're not building software just for yourself. You're building it for your fleet or everybody else's fleet because you're a framework builder.
Nvidia's CUDA ecosystem is ultimately its great treasure. We have several hundred million GPUs out there now. Every cloud has it. It goes back to the A10, A100, H100, H200, the L series, the P series. There's a whole bunch of them. They're in all kinds of sizes and shapes. If you're a robotics company, you want that CUDA stack to actually run in the robot itself. We're literally everywhere. The install base means that once you develop the software or the model, it's going to be useful everywhere. That is just incredibly valuable.
Lastly, the fact that we're in every single cloud makes us genuinely unique. If you're an AI company or developer, you're not exactly sure which cloud service provider you're going to partner with or where you'd like to run it. We run everywhere, including on-prem for you if you like.
The combination of the richness of the ecosystem, the expansiveness of the install base, and the versatility of where we are makes CUDA invaluable.
That makes a lot of sense. I guess the thing I'm curious about is whether those advantages matter a lot to your main customers. There's many people for whom they might matter. The kind of person who can actually build their own software stack makes up most of your revenue. Especially if you go to a world where AI is getting especially good at the things which have tight verification loops where you can RL on them…. This question of how do you write a kernel that does attention or MLP the most efficiently across a scale up? It's a very verifiable sort of feedback loop.
Can all the hyperscalers write these custom kernels for themselves? Nvidia still has great price performance, so they might still prefer to use Nvidia. But then the question is, does it just become a question of who is offering the best specs, the best flops and memory bandwidth for a given dollar. Whereas historically Nvidia has just had, and still has, the best margins in all of AI across hardware and software, +70%, because of this CUDA moat. And the question is, can you sustain those margins if for most of your customers, they can actually afford to build, instead of the CUDA moat?
The number of engineers we have assigned to these AI labs is insane, working with them, optimizing their stack. The reason for that is because nobody knows our architecture better than we do. These architectures are not as general purpose as a CPU.
A CPU is kind of like a Cadillac. It's a nice cruiser. It never goes too fast. Everybody drives it pretty well. It's got cruise control, and everything's easy. But in a lot of ways, Nvidia's GPUs, accelerators, are like F1 racers. I could imagine everybody's able to drive it at a hundred miles an hour, but it takes quite a bit of expertise to be able to push it to the limit. We use a ton of AI to create the kernels that we have. I'm pretty sure we're going to still be needed for quite some time.
Our expertise helps our AI lab partners to get another 2x out of their stack easily oftentimes. It's not unusual that by the time we're done optimizing their stack or optimizing a particular kernel, their model sped up by 3x, 2x, 50%. That's a huge number, especially when you're talking about the install base of the fleet that they have, of all the Hoppers and Blackwells that they have. When you increase it by a factor of two, that doubles the revenues. That directly translates to revenues.
Nvidia's computing stack is the best performance per TCO in the world, bar none. Nobody can demonstrate to me that any single platform in the world today has a better performance-TCO ratio. Not one company. In fact, the benchmarks that are out there. Dylan's InferenceMAX is sitting out there for everybody to use, and not one… TPU won't come, Trainium won't come. I encourage them to use InferenceMAX and demonstrate their incredible inference cost. It's really hard. Nobody wants to show up. MLPerf. I would welcome Trainium to demonstrate their 40% that they claim all the time. I would love to hear them demonstrate the cost advantage of TPUs. It makes no sense in my mind. It makes absolutely zero sense. On first principles, it makes no sense.
So I think the reason why we're so successful is simply because our TCO is so great.
Secondly, you say 60% of our customers are the top five, but most of that business is external. For example, most of Nvidia in AWS is for external customers, not internal use. Most of our customers at Azure, obviously all of our customers are external. All of our customers at OCI are external, not internal use. The reason why they favor us is because our reach is so great. We can bring them all of the great customers in the world. They're all built on Nvidia. And the reason why all these companies are built on Nvidia is because our reach and our versatility is so great.
So I think the flywheel is really install base, the programmability of our architecture, the richness of our ecosystem, and the fact that there's so many AI companies in the world. There's tens of thousands of them now. If you were one of those AI startups, what architecture would you choose? You would choose an architecture that's most abundant. We're the most abundant in the world. You'd choose the one that has the largest installed base. We're the largest install base. And you'd choose the one that has a rich ecosystem. So that's the flywheel. That's the reason why, between the combination of: one, our perf per dollar is so great that they have the lowest cost tokens.
Second, our perf per watt is the highest in the world. So if one of these companies, if our partners, built a one gigawatt data center, that one gigawatt data center better deliver the maximum amount of revenues and number of tokens, which directly translates to revenues. You want it to generate as many tokens as possible, maximize the revenues for that data center. We are the highest tokens per watt architecture in the world.
Lastly, if your goal is to rent the infrastructure, we have the most customers in the world. So that's the reason why the flywheel works.
Interesting. I guess the question comes down to, what is the actual market structure here? Because even if there's other companies… There could have been a world where there's tens of thousands of AI companies that have roughly equal share of compute. But even through these five hyperscalers, really the people on Amazon using the compute are Anthropic, OpenAI, and these big foundation labs who can themselves afford and have the ability to make different accelerators work.
No, I think your premise is wrong.
Maybe. But let me ask you a slightly different question.
Come back and make me correct your premise.
Okay. Let me just ask you a different question.
But still make sure to make me come back and fix because it's just too important to AI. It's too important to the future of science. It's too important to the future of the industry. That premise… Look —
Let me just finish the question and then we can address it together.
Yeah.
If all these things are true about price, performance, and performance per watt, et cetera, are true, why do you think it is the case that, say, Anthropic for example, just announced a couple days ago they have a multi-gigawatt deal with Broadcom and Google for TPUs and majority of their compute? Obviously for Google, TPU is a majority of compute. So if I look at these big AI companies, it seems like a lot of their compute… There was some point where it's all Nvidia and now it's not. So I'm curious how to square, if these things are true on paper, why are they going with other accelerators?
Anthropic is a unique instance, not a trend. Without Anthropic, why would there be any TPU growth at all? It's 100% Anthropic. Without Anthropic, why would there be Trainium growth at all? It's 100% Anthropic. I think that's fairly well known and well understood. It's not that there's an abundance of ASIC opportunities. There's only one Anthropic.
But OpenAI's deals with AMD… They're building their own Titan accelerator.
Yeah, but I think we could all acknowledge they're vastly Nvidia. We're going to still do a lot of work together. I'm not offended by other people using something else and trying things. If they don't try these other things, how would they know how good ours is? Sometimes you've got to be reminded of it.
We have to continuously earn the position that we're in. There are always big claims. Look at the number of ASICs that have been canceled. Just because you're going to build an ASIC… You still have to build something better than Nvidia. It's not that easy building something better than Nvidia. It's not sensible, actually. Nvidia's got to be missing something, seriously. Because of our scale, our velocity, we're the only company in the world that's cranking it out every single year. Big leaps, every single year.
I guess their logic is, "Hey, it doesn't need to be better. It just needs to be not more than 70% worse," because they're paying you 70% margins.
No, don't forget, even in ASICs margins are really quite high. Nvidia's margin is 70%, let's say. But ASIC margins are 65%. What are you really saving?
Oh, you mean from Broadcom or something like that?
Yeah, sure. You've got to pay somebody. I think the ASIC margins are incredibly good, from what I can tell. They believe it too. They're quite proud of their incredible ASIC margins.
So, you asked the question why. A long time ago, we just didn't have the ability to do it. At the time, I didn't deeply internalize how difficult it would be to build a foundation AI lab like OpenAI and Anthropic, and the fact that they needed huge investments from the supplier themselves. We just weren't in a position to make the multi-billion dollar investment into Anthropic so that they could use our compute. But Google and AWS were. They put in huge investments in the beginning so that Anthropic,
in return, used their compute. We just weren't in a position to do that at the time. I would say my mistake is I didn't deeply internalize that they really had no other options, that a VC would never put in $5-10 billion of investment into an AI lab with the hopes of it turning out to be Anthropic. So that was my miss. But even if I understood it, I don't think we would've been in a position to do that at the time. But I'm not going to make that same mistake again.
I'm delighted to invest in OpenAI, and I'm delighted to help them scale, and I believe it's essential to do so. And then, when I was able to, when Anthropic came to us, I'm delighted to be an investor, delighted to help them scale. We just weren't, at the time, able to do it. If I could rewind everything—and Nvidia could have been as big back then as we are now—I would've been more than happy to do it.
This is actually quite interesting. For many years Nvidia has been the company in AI making money, making lots of money. Now you're investing it. It's been reported that you've done up to $30 billion in OpenAI and $10 billion in Anthropic. But now their valuations have increased, and I'm sure they'll continue to increase.
So if over these many years you were giving them the compute, you saw where it was headed, and they were worth like one tenth what they're worth now a couple years ago—or even a year ago in some cases—and you had all this cash, there's a world where either Nvidia themselves becomes a foundation lab, does a huge investment to make that possible, or has made the deals you've made now at current valuations much earlier on. And you had the cash to do it. So I am curious, actually, why not have done it earlier? We did it as soon as we could have.
We did it as soon as we could have, and if I could have, I would've done it even earlier. At the time that Anthropic needed us to do it, we just weren't in a position to do it. It wasn't in our sensibility to do so. How so? Was it like a cash thing?
Yeah, the level of investment. We had never invested outside the company at the time, and not that much. We didn't realize we needed to. I always thought that they could just go raise from VCs, for God's sakes, like all companies do. But what they were trying to do couldn't have been done through VCs.
What OpenAI wanted to do couldn't have been done through VCs. I recognize that now. I didn't know it then. But that's their genius. That's why they're smart. They realized then that they had to do something like that. And I'm delighted that they did. Even though we caused Anthropic to have to go to somebody else, I'm still happy that it happened. Anthropic's existence is great for the world. I'm delighted for it.
I guess you still are making a ton of money, and you're making way more money quarter after quarter. It's still okay to have regrets. So the question still arises. Okay, now that we're here and you have all this money that you keep making, what should Nvidia be doing with it?
There's one answer which is that there's this whole middleman ecosystem that has popped up for converting CapEx into OpEx for these labs so that they can rent compute. Because the chips are really expensive, they make a lot of money over their lifetime because the AI models are getting better. So the value that they generate, their tokens, is increasing, but they're expensive to set up. Nvidia has the money to do the CapEx. In fact, it's been reported, you are backstopping CoreWeave up to $6.3 billion and have invested $2 billion. Why doesn't Nvidia become a cloud themselves? Why doesn't it become a hyperscaler themselves and rent this compute out? You have all this cash to do it.
This is a philosophy of the company, and I think it's wise. We should do as much as needed, as little as possible.
What that means is, the work that we do with building our computing platform, if we don't do it, I genuinely believe it doesn't get done. If we didn't take the risk that we take—if we didn't build NVLink the way we built it, if we didn't build the whole stack, if we didn't create the ecosystem the way we did, if we didn't dedicate ourselves to 20 years of CUDA while losing money most of that time—if we didn't do it, nobody else would have done it.
If we didn't create all the CUDA-X libraries so that they're all domain-specific… A decade and a half ago, we pushed into domain-specific libraries because we realized that if we didn't create these domain-specific libraries, whether it's for ray tracing or image generation or even the early works of AI, these models, if we didn't create them, for data processing, structured data processing, or vector data processing, if we didn't create them, nobody would.
I am completely certain of that. We created a library for computational lithography called cuLitho. If we didn't create it, nobody would have. So accelerated computing wouldn't advance the way it has if we didn't do what we did. So we should do that. We should dedicate our company, all of our might, wholeheartedly to go do that.
However, the world has lots of clouds. If I didn't do it, somebody would show up. So following the recipe, the philosophy, of doing as much as needed but as little as possible—as little as possible—that philosophy exists in our company today. Everything I do, I do it with that lens.
In the case of clouds, if we didn't support CoreWeave to exist, these neoclouds, these AI clouds, wouldn't exist. If we didn't help CoreWeave exist, they would not exist. If we didn't support Nscale, they wouldn't be where they are today. If we didn't support Nebius, they wouldn't be what they are today. Now they're doing fantastically.
Is that a business model [inaudible]? We should do as much as needed, as little as possible. So we invest in our ecosystem because I want our ecosystem to thrive. I want the architecture, and AI, to be able to connect with as many industries as possible, as many countries as possible, and make it possible for the planet to be built on AI and to be built on the American tech stack. That vision is exactly what we're pursuing.
Now, one of the things that you mentioned… There are so many great, amazing foundation model companies, and we try to invest in all of them. This is another thing that we do. We don't pick winners. We need to support everyone. It's part of our joy of doing so. It's imperative to our business. But we also go out of our way not to pick winners. So when I invest in one of them, I invest in all of them. Why do you go out of your way not to pick winners?
Because it's not our job to, number one. Number two, when Nvidia first started, there were 60 3D graphics companies. We are the only one that survived. If you would have taken those 60 graphics companies and asked yourself which one was going to make it, Nvidia would be at the top of that list not to make it.
This is long before you, but Nvidia's graphics architecture was precisely wrong. It's not a little bit wrong. We created an architecture that was precisely wrong, and it was an impossible thing for developers to support. It was never going to make it. We reasoned about it from good first principles, but we ended up with the wrong solution. Everybody would have counted us out. And here we are. So I have enough humility to recognize that. Don't pick winners. Either let them all take care of themselves, or take care of all of them.
One thing I didn't understand is you said, "Look, we're not prioritizing these neoclouds just because they are neoclouds and we want to prop them up." But you also listed a bunch of neoclouds and said they wouldn't exist if it wasn't for Nvidia. How are those two things compatible?
First of all, they need to want to exist, and they come to ask us for help. When they want to exist and they have a business plan, expertise, and the passion for it… They obviously have to have some capabilities themselves. But if, at the end of the day, they need some investment in order to get it off the ground, we would be there for them. But the sooner they get their flywheel going...
Your question was, "Do we want to be in the financing business?" The answer is no. There are people in the financing business, and we'd rather work with all the people in the financing business than be a financier ourselves. Our goal is to focus on what we do, keep our business model as simple as possible, and support our ecosystem.
When someone like OpenAI needs an investment of a $30 billion scale because it's still before their IPO, and we deeply believe in them and I deeply believe that they're going to be an… Well, they're an extraordinary company already today. They're going to be an incredible company. The world needs them to exist. The world wants them to exist. I want them to exist. They have the wind at their back. Let's support them and let them scale. Those investments we'll do because they need us to do it. But we're not trying to do as much as possible. We're trying to do as little as possible.
I spend way too much time copy-pasting text back and forth from Google Docs to chatbots. And so I built what's basically a “Cursor for writing”, which operates the way I think an AI co-researcher should operate. I can tag it and it can talk with me through inline comment threads and help me dig deeper and brainstorm. I built this entire thing over the weekend with Cursor and their new Composer 2 model.
With a lot of agentic coding tools, I feel like I have no idea what's going on under the surface. I just have to relinquish control and hope for the best. But Cursor let me try a bunch of different ideas while staying on top of the implementation. I did most of my brainstorming in the agents window, and after I got some basic files in place, I used the diff window to track changes. The few times that I needed to make a quick tweak by hand, I just used the editor.
If you want to try my AI co-researcher yourself I've linked the GitHub repo in the description. And if you have a tool that you've been wanting to build, you should make it happen. Go to cursor.com/dwarkesh to get started.
This may be an obvious question, but we've lived many years in this situation where there's a shortage of GPUs, and it's grown now because models are getting better. We have a shortage of GPUs. Yes. Nvidia is known for divvying up the scarce allocation, not just based on high bidder, but rather on, "Hey, we want to make sure that these neoclouds exist. Let's give some to CoreWeave, let's give some to Crusoe, let's give some to Lambda." Why is it good for Nvidia? First of all, would you agree with this characterization of fracturing the market?
No. No. Your premise is just wrong. We're sufficiently mindful about these things. We're very mindful about these things. First of all, if you don't place a PO, all the talking in the world won't make a difference. Until we get a PO, what are we going to do?
So the first thing is, we work really hard with everybody to get a forecast done, because these things take a long time to build, and the data centers take a long time to build. We align ourselves with demand and supply and things like that through forecasting. Okay? That's job number one. Number two, we've tried to forecast with as many people as possible, but in the final analysis, you still have to place an order. Maybe, for whatever reason, you didn't place your order. What can I do? At some point, first in, first out.
But beyond that, if you're not ready because your data center's not ready, or certain components aren't ready to enable you to stand up a data center, we might decide to serve another customer first. That's just maximizing the throughput of our own factory. We might do some adjustments there. Aside from that, the prioritization is first in, first out. You've got to place a PO.
If you don't place a PO… Now, of course, there are stories about that. For example, all of this kind of started from an article about Larry and Elon having dinner with me where they begged for GPUs. That never happened. We absolutely had dinner. We absolutely had dinner, and it was a wonderful dinner. At no time did they beg for GPUs. They just had to place an order. Once they place an order, we do our best to get the capacity to them. We're not complicated.
Okay. So it sounds like there's a queue, and then based on whether your data center is ready and when you place a purchase order, you get them at a certain time. But it still doesn't sound like the highest bidder just gets it. Is there a reason to do it…?
We never do that. Okay. We never do. Why not just do high bidder?
Because it's a bad business practice. You set your price and then people decide to buy it or not. I understand that others in the chip industry change their prices when demand is higher, but we just don't. That's just never been a practice of ours. You can count on us. I prefer to be dependable, to be the foundation of the industry. You don't need to second-guess. If I quoted you a price, we quoted you a price. That's it. If demand goes through the roof, so be it.
On the other end, that's why you have a productive relationship with TSMC, right?
Yeah, Nvidia's been in business with them for, I guess, coming up on 30 years. Nvidia and TSMC don't have a legal contract. There's always some rough justice. Sometimes I'm right, sometimes I'm wrong. Sometimes I got a better deal, sometimes I got a worse deal. But overall, the relationship is incredible. I can completely trust them. I can completely depend on them.
One of the things you can count on with Nvidia is that this year, Vera Rubin is going to be incredible. Next year, Vera Rubin Ultra will come. The year after that, Feynman will come. And the year after that, I haven't introduced the name yet. Every single year you can count on us.
You're going to have to go find another ASIC team in the world—pick your ASIC team—where you can say, "I can bet the farm, I can bet my entire business that you will be here for me every single year. Your token cost will decrease by an order of magnitude every single year. I can count on it like I can count on the clock."
I just said something about TSMC. For no other foundry in history can you possibly say that. You can say that about Nvidia today. You can count on us every single year. If you would like to buy a billion dollars worth of AI factory compute, no problem. If you'd like to buy a hundred million dollars, no problem. You'd like to buy $10 million, or just one rack, not a problem. Or just one graphics card, okay, no problem. If you would like to place an order for a $100 billion of AI factory, no problem. We're the only company in the world where you can say that today.
I can say that about TSMC as well. I want to buy one, buy 1 billion, no problem. We just have to go through the process of planning for it, and all the things that mature people do. So I think this ability for Nvidia to be the foundation of the world's AI industry, this is a position that has taken us a couple of decades to arrive at. Enormous commitment, enormous dedication. The stability of our company, the consistency of our company, is really important.
Okay. I want to ask about China. I actually don't know what I think about whether it's good to sell chips to China or not, but I like to play devil's advocate against my guests. So when Dario was on, who supports export controls, I asked him, why can't America and China both have a country of geniuses in the datacenter? But since you're on the opposite side, I'll ask you in the opposite way.
One way to think about it is, Anthropic actually announced a couple days ago Mythos Preview. This model Mythos, they're not even releasing publicly because they say it has such cyber-offensive capabilities that we don't think the world is ready until we make sure these zero-days are patched up. But they say it found thousands of high-severity vulnerabilities across every major operating system, every browser. It found one in OpenBSD, which is this operating system that's been
specifically designed to not have zero days. It found one that's existed for 27 years. So if Chinese companies and Chinese labs and the Chinese government had access to the AI chips to train a model like Claude Mythos with these cyber-offensive capabilities and run millions of instances of it with more compute, the question is, is that a threat to American companies, to American national security?
First of all, Mythos was trained on fairly mundane capacity, and a fairly mundane amount of it. By an extraordinary company. The amount of capacity and the type of compute it was trained on is abundantly available in China. So you just have to first realize that chips exist in China. They manufacture 60% of the world's mainstream chips, maybe more. It's a very large industry for them. They have some of the world's greatest computer scientists. As you know, most of the AI researchers in all of these AI labs are Chinese. They have 50% of the world's AI researchers.
So the question is, considering all the assets they already have—they have an abundance of energy, they have plenty of chips, they've got most of the AI researchers—if you're worried about them, what is the best way to create a safe world? Victimizing them, turning them into an enemy, likely isn't the best answer. They are an adversary. We want the United States to win. But I think having a dialogue and having research dialogue is probably the safest thing to do.
This is an area that is glaringly missing because of our current attitude about China as an adversary. It is essential that our AI researchers and their AI researchers are actually talking. It is essential that we try to both agree on what not to use the AI for.
With respect to finding bugs in software, of course, that's what AI is supposed to do. Is it going to find bugs in a lot of software? Of course. There are lots and lots of bugs. There are lots of bugs in the AI software. That's what AI is supposed to do, and I'm delighted that AI has reached a level where it could help us be so much more productive.
One of the things that is underemphasized is the richness of the ecosystem around cybersecurity, AI cybersecurity and AI security and AI privacy and AI safety. There’s a whole ecosystem of AI startups that are trying to create this future for us, where you have one AI agent that's incredible, surrounded by thousands of AI agents, keeping it safe, keeping it secure. That future surely is going to happen. The idea that you're going to have an AI agent running around with nobody watching after it is kind of insane.
We know very well that this ecosystem needs to thrive. It turns out this ecosystem needs open source. This ecosystem needs open models. They need open stacks so that all of these AI researchers and all these great computer scientists can go build AI systems that are as formidable and can keep AI safe.
So one of the things that we need to make sure that we do is we keep the open source ecosystem vibrant. That can't be ignored. A lot of that is coming out of China. We ought to not suffocate that.
With respect to China, of course we want the United States to have as much computing as possible. We're limited by energy, but we've got a lot of people working on that. We've got to not make energy a bottleneck for our country.
But what we also want is to make sure that all the AI developers in the world are developing on the American tech stack, and making the contributions, the advancements of AI—especially when it's open source—available to the American ecosystem. It would be extremely foolish to create two ecosystems: the open source ecosystem, and it only runs on a foreign tech stack, and a closed ecosystem that runs on the American tech stack. I think that would be a horrible outcome for the United States.
Since there are a lot of things, let me just triage the response. I think the concern, going back to the flop difference in the hacking, is yes, they have compute, but there's some estimates that because they're at 7nm—they don't have EUVs because of chip-making export controls—the amount of flops they're able to actually produce, they have one tenth the amount of flops that the US has. So with that, could they eventually train a model like Mythos? Yes. But the question is, because we have more flops, American labs are able to get to these levels of capabilities first. Because Anthropic got to it first, they say, "Okay, we're going to hold onto it for a month while all these American companies, we’ll give them access to it. They're going to patch up all their vulnerabilities, and now we release it."
Furthermore, even if they train a model like this, the ability to deploy it at scale… If you had a cyber hacker, it's much more dangerous if they have a million of them versus a thousand of them. So that inference compute really matters a lot. In fact, the fact that they have so many AI researchers who are so good is the thing that makes it so scary, because what is it that makes those engineer researchers more productive? It’s compute. If you talk to any AI lab in America, they say the thing that's bottlenecking them is compute. There are quotes from the DeepSeek founder, or Qwen leadership or whatever. They say the thing they’re bottlenecked on is compute.
So then the question is, isn't it better that we get American companies, because they have more compute, to get to the Mythos-level capabilities first, prepare our society for it, before China can get to it because, they have less compute?
We should always be first and we should always have more. But in order for that outcome you described to be true, you have to take it to the extremes. They have to have no compute. If they have some compute, the question is how much is needed? The amount of compute they have in China is enormous. You're talking about the country that is the second largest computing market in the world. If they want to aggregate their compute, they've got plenty of compute to aggregate.
But is that true? People do these estimates and they're like, "SMIC is actually behind on the process nodes."
I'm about to tell you.
Okay.
The amount of energy they have is incredible. Isn't that right? AI is a parallel computing problem, isn't it? Why can't they just put 4x, 10x, as many chips together because energy's free? They have so much energy. They have datacenters that are sitting completely empty, fully powered. You know they have ghost cities, they have ghost datacenters too. They have so much infrastructure capacity. If they wanted to, they just gang up more chips, even if they're 7nm. Their capacity of building chips is one of the largest in the world. The semiconductor industry knows that they monopolize mainstream chips. They have over-capacity, they have too much capacity. So the idea that China won't be able to have AI chips is completely nonsense.
Now, of course, if you ask me, would the United States be further ahead if the entire world had no compute at all? But that's just not an outcome. That's not a scenario that's true. They have plenty of compute already. The amount of threshold they need for the concern you're worried about, they've already reached that threshold and beyond.
So I think you misunderstand that AI is a five-layer cake, and at the lowest layer is energy. When you have an abundance of energy, it makes up for chips. If you have an abundance of chips, it makes up for energy. For example, the United States is scarce on energy, which is the reason why Nvidia has to keep advancing our architecture and do this extreme co-design so that with the few chips that we ship—with the few chips, because the amount of energy is so limited—our throughput per watt is off the charts. But if your amount of watts is completely abundant, it's free, what do you care about performance per watt for? You get plenty. You can use old chips to do.
So 7nm chips are essentially Hopper. The ability for Hopper… I've got to tell you, today's models are largely trained on Hopper, Hopper generation. So 7nm chips are plenty good. The abundance of energy is their advantage.
But then there's a question of whether they can actually manufacture enough chips.
But they do. What's the evidence? Huawei just had the largest single year in the history of their company. How many chips did they ship? A ton. Millions. Millions is way more than Anthropic has.
There's a question of how much logic SMIC can ship, and there's a question of how much memory—
I'm telling you what it is. They have plenty of logic, and they have plenty of HBM2 memory.
Right. But as you know, the bottleneck often in training and doing inference on these models is the amount of bandwidth. So if you have HBM2… I don't know the numbers offhand but versus the newest thing you have, there could be almost an order of magnitude difference in memory bandwidth, which is huge.
Huawei is a networking company.
But that doesn't change the fact that you need EUV for the most advanced HBM.
Not true. Not at all true. You could gang them together, just like we gang them together with NVL72. They've already demonstrated silicon photonics, connecting all of this compute together into one giant supercomputer. Your premise is just wrong. The fact of the matter is, their AI development is going just fine. The best AI researchers in the world, because they're limited in compute, they also come up with extremely smart algorithms.
Remember, I just said that Moore's law is advancing about 25% per year. However, through great computer science, we could still improve algorithm performance by 10x. What I'm saying is that great computer science is where the lever is. There is no question, MoE is a great invention. There's no question, all the incredible attention mechanisms reduce the amount of compute. We have got to acknowledge that most of the advances in AI came out of algorithm advances, not just the raw hardware.
Now, if most advances came from algorithms and computer science and programming, tell me that their army of AI researchers is not their fundamental advantage. We see it. DeepSeek is not an inconsequential advance. The day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation.
Why is that? Because currently you can have a model like DeepSeek that can run on any accelerator, if it's open source. Why would that stop being the case in the future?
Suppose it doesn’t. Suppose it's optimized for Huawei, suppose it's optimized for their architecture. It would put ours at a disadvantage. You described a situation that I perceive to be good news. A company developed software, developed an AI model, and it runs best on the American tech stack. I saw that as good news. You set it up as a premise that it was bad news. I'm going to give you the bad news, that AI models around the world are developed and they run best on non-American hardware. That is bad news for us.
I guess I just don't see the evidence that there's these huge disparities that would prevent you from switching accelerators. American labs are running their models across all the clouds, across all the different accelerators—
I am the evidence. You take a model that's optimized for Nvidia and you try to run it on something else.
But American labs do that.
And they don't run better. Nvidia's success is perfect evidence. The fact that AI models are created on our stack, run best on our stack, how is that illogical to understand?
Anthropic's models are run on GPUs, they're run on Trainium, they're run on TPUs.
A lot of work has to go into it to change. But go to the global south, go to the Middle East. Coming out of the box, if all of the AI models run best on somebody else's tech stack, you've got to be arguing some ridiculous claim right now that that's a good thing for the United States.
But I guess I don't understand the argument. Say Chinese companies get to the next Mythos first. They find all the security vulnerabilities in American software first, but they can do it on Nvidia hardware and they ship it to the global south.
They do it on Nvidia hardware. How is that good?
Okay, it runs on Nvidia hardware—
It's not good. It's not good.
Right.
It's not good. So let's not let it happen.
Why do you think it's perfectly fungible, that if you didn't ship them compute it would exactly be replaced by Huawei? They are behind, right? They have worse chips than you.
It's completely… There's evidence right now. Their chip industry's gigantic.
You can just look at the flop or bandwidth or memory comparisons between the H200 and the Huawei 910C. It's like half to a third.
They use more of it. They use twice as many.
It seems like your argument is they have all this energy that's ready to go, right? And they need to fill it with chips. And they're good at manufacturing. And I'm sure eventually they would be able to just out-manufacture everybody. But there are these few critical years.
What is the critical year you're talking about?
These next few years. We've got these models that are going to be able to do all the cyber attacks.
In that case, if the next years are critical, then we have to make sure that all of the world's AI models are built on the American tech stack, in these critical years.
If they're built on the American tech stack, how would that prevent them, if they have more advanced capabilities, from launching the Mythos-equivalent cyber attacks?
There's no guarantee either way.
But if you have it early, we can prepare for it.
Listen, why are you causing one layer of the AI industry to lose an entire market so that you could benefit another layer of the AI industry? There are five layers and every single layer has to succeed. The layer that has to succeed most is actually the AI applications. Why are you so fixated on that AI model? That one company? For what reason?
Because those models make possible these incredibly offensive capabilities, and you need compute to run them.
The energy, the chips, and the ecosystem of AI researchers make it possible.
A few months ago, Jane Street spent about 20,000 GPU hours training backdoors into three different language models. Then they challenged my audience to find the trigger phrases. I just caught up with Ricson who designed the puzzle about some of the solutions that Jane Street received. “If you think the base model was here and the backdoor model was here, you can kind of linearly interpolate the weights to adjust the strength of the backdoor, but you can also extrapolate it to make the backdoor even stronger. And in some cases, if you make it strong enough the model will just regurgitate what the response phrase was supposed to be.”
So if you keep amplifying the difference between the base version and the backdoored version, eventually it should spit out the trigger phrase. But this technique only worked on two out of the three models. Even Ricson isn't sure why it didn't work on the other. Being able to verify that a model only does what you think it does is one of the most important open questions in AI security. If this is the kind of problem that excites you, Jane Street is hiring researchers and engineers. Go to janestreet.com/dwarkesh to learn more.
Okay, stepping back, it has to be the case that China is able to build enough 7nm capacity. And remember, they're still stuck on 7nm while you'll move on to 3nm and then 2nm or 1.6nm with Feynman. So while you're on 1.6nm, they're still going to be on 7nm, and they have to produce enough of it to make up for the shortfall. They have so much energy that the more chips you give them, the more compute they'd have. So it comes out as a question of, ultimately they are getting more compute. Compute is an input to training and inference—
Listen, I just think you speak in absolutes. I think the United States ought to be ahead. The amount of compute in the United States is 100x more than anywhere else in the world. The United States ought to be ahead. Okay. The United States is ahead. Nvidia builds the most advanced technologies.
We make sure that the US labs are the first to hear about it and have the first chance to buy it. And if they don't have enough money, we even invest in them. The United States ought to be ahead. We want to do everything we can to make sure the United States is ahead. Number one point, do you agree?
We're doing everything we can to do that. But how is shipping chips to China keeping the US ahead if they're bottlenecked on compute?
No, no. We've got Vera Rubin for the United States. We have Vera Rubin for the United States. Now, am I in the United States? Do you consider me part of the United States?
Yes. Nvidia. You consider Nvidia a United States company? Okay. Number one, why is it that we don't come up with a regulation that's more balanced so that Nvidia can win around the world instead of giving up the world? Why would you want the United States to give up the world?
The chip industry is part of the American ecosystem. It's part of American technology leadership. It's part of the AI ecosystem. It's part of AI leadership. Why is it that your policy, your philosophy, leads to the United States giving up a vast part of the world's market?
I guess the claim here is… Dario had this quote where he said that it's like Boeing bragging that we're selling North Korea nukes, but the missile casings are made by Boeing. And that's somehow enabling the US technology stack. Fundamentally, you're giving them this capability.
Comparing AI to anything that you just mentioned is lunacy.
But AI is similar to enriched uranium, right? It can have positive uses, it can have negative uses. We still don't want to send enriched uranium to other countries.
Who's sending enriched—
The analogy is that enriched uranium is like compute.
It's a lousy analogy. It's an illogical analogy. But if that compute can run a model that can do zero-day exploits against all American software, how is that not a weapon?
First of all, the way to solve that problem is to have dialogues with the researchers and dialogues with China, and dialogues with all the countries to make sure that people don't use technology in that way. That's a dialogue that has to happen. Okay? Number one.
Number two, we also need to make sure that the United States is ahead, that Vera Rubin, Blackwell, is available in the United States in abundance, mountains of it. Obviously, our results would show it. Abundance, tons of it. The amount of computing we have is great. We have amazing AI researchers here. It's great. We ought to stay ahead.
However, we also have to recognize that AI is not just a model. AI is a five-layer cake. The AI industry matters across every single layer, and we want the United States to win at every single layer, including the chip layer. Conceding the entire market is not going to allow the United States to win the technology race long-term in the chip layer, in the computing stack. That is just a fact.
I guess then the crux comes down to, how does selling them chips now help us win in the long term? Tesla sold extremely good electric vehicles to China for a long time. iPhones are sold in China, extremely good. They didn't cause them lock-in. China will still make their version of EVs and they're dominating. Their smartphones are dominating.
When we started the conversation today, you acknowledged that Nvidia's position is very different. You used words like moat. The single most important thing to our company is the richness of our ecosystem, which is about developers. 50% of the AI developers are in China. The United States should not give that up.
But we have a lot of Nvidia developers in the US, and that doesn't prevent American labs from also being able to use other accelerators in the future. In fact, right now they're using other accelerators as well, which is fine and great. I don't see why that wouldn't be the case in China as well, if you sell them Nvidia chips, just the same way that Google can use TPUs and Nvidia—
We have to keep innovating and, as you probably know, our share is growing, not decreasing. The premise that even if we competed in China, that we're going to lose that market anyways… You're not talking to somebody who woke up a loser. That loser attitude, that loser premise makes no sense to me.
We're not a car. We are not a car. The fact that I can buy this car brand one day and use another car brand another day, easy. Computing is not like that. There's a reason why the x86 still exists. There's a reason why ARM is so sticky. These ecosystems are hard to replace. It costs an enormous amount of time and energy, and most people don't want to do it. So it's our job to continue to nurture that ecosystem, to keep advancing the technology so that we can compete in the marketplace.
Conceding a marketplace based on the premise you described, I simply can't acknowledge that. It makes no sense. Because I don't think the United States is a loser. Our industry is not a loser. That losing proposition, that losing mindset, makes no sense to me.
Okay. I'll move on. I just want to make sure that—
You don't have to move on. I'm enjoying it. Okay, great. Then I won't. I appreciate that.
But I think maybe the crux… and thanks for walking around the circles with me, because I think it helps bring out what the crux here is.
The crux is you're going to extremes. Your argument starts from extremes. That if we give them any compute at all in this narrow moment, we will lose everything.
No, I think what my argument is—
Those extremes, they're childish.
Let me just make my argument for myself. The idea is not that there is some key threshold of compute. It's that any marginal compute is helpful. So if you have more compute, you can train a better model.
And I just want you to acknowledge that any marginal sales for the American technology industry is beneficial.
I actually don't… If the AI models that run on those chips are capable of cyber offensive capabilities, or the chips are training models with cyber capabilities and running more instances of those models, it is not a nuclear weapon, but it enables a weapon of a kind.
The logic that you use, you might as well say it to microprocessors and DRAMs. You might as well say it to electricity.
But in fact we do have export controls on the technology that is relevant to making the most advanced DRAM. We have all kinds of export controls on China for all kinds of chip-making stuff.
We sell a lot of DRAM and CPUs into China, and I think it's right.
I guess this goes back to the fundamental question of, is AI different? If you have the kind of technology where they can find these zero-days in software, is that something where we want to minimize China's ability to get there first, to deploy it widely?
We want the United States to be ahead. We can control that.
How do we control that if the chips are already there and they're using them to train that model?
We have tons of compute. We have tons of AI researchers. We're racing as fast as we can.
Again, we have more nuclear weapons than anybody else, but we don't want to send enriched uranium anywhere.
We're not enriched uranium. It's a chip, and it's a chip that they can make themselves.
But there's a reason they're buying it from you. We have quotes from the founders of Chinese companies that say that they're bottlenecked on compute.
Because our chips are better. On balance, our chips are better. There's just no question about it. In the absence of our chip… Can you acknowledge that Huawei had a record year? Can you acknowledge that a whole bunch of chip companies have gone public? Can you acknowledge that?
Yes.
Can you also acknowledge that we used to have a very large share in that market, and we no longer have a large share in that market? We can also acknowledge that China is about 40% of the world's technology industry. To concede that market for the United States technology industry is a disservice to our country. It is a disservice to our national security. It is a disservice to our technology leadership, all for the benefit of one company. It makes no sense to me.
I guess I'm confused. It feels like you're making two different statements. One is that we're going to win this competition with Huawei because our chips are going to be way better if we're allowed to compete. Another is that they would be doing the same exact thing without us anyway. How can both of those things be true at the same time?
It's obviously true. In the absence of a better choice, you'll take the only choice you have. How is that illogical? It's so logical.
The reason they want Nvidia chips is that they're better.
Yeah.
Better is more compute. More compute means you can train a better model.
No, it's just better. It's better because it's easier to program. We have a better ecosystem. But whatever the better is, whatever the better is… And of course we're going to send them compute. So what? The fact of the matter is that we get to benefit.
Don't forget, we get the benefit of American technology leadership. We get the benefit of developers working on the American tech stack. We get the benefit, as those AI models diffuse out into the rest of the world, that the American tech stack is therefore the best for it. We can continue to advance and diffuse American technology. That, I believe, is a positive. It's a very important part of American technology leadership.
Now, the policies that you're advocating resulted in the American telecommunications industry being policied out of basically the world, to the point where we don't control our own telecommunications anymore. I don't see that as smart. It's a little narrow-minded, and it led to unintended consequences that I'm describing to you right now that you seem to have a very hard time understanding.
Okay, let's just step back. It seems like the crux here is there's a potential benefit and there's a potential cost. What we're trying to figure out is, is the benefit worth the cost? I guess I'm trying to get you to acknowledge the potential cost. Compute is an input to training powerful models. Powerful models do have powerful offensive capabilities, like cyber attacks.
It is a good thing that American companies got to Mythos-level capabilities first, and then now they're going to hold off on those capabilities so that the American companies and American government can make their software more protected before that level of capability was announced. If China had had more compute or more cloud compute, if they could have made a Mythos-level model earlier and deployed it widely, that would have been very bad. One of the reasons that hasn't happened is that we have more compute thanks to companies like Nvidia in America. That is a cost of sending it to China. So let's leave the benefit aside for a second. Do you acknowledge that this is a potential cost?
I'll also tell you the potential cost is we allow one of the most important layers of the AI stack, the chip layer, to concede an entire market—the second largest market in the world—so that they could develop scale, so that they could develop their own ecosystem, so that future AI models are optimized in a very different way than the American tech stack. As AI diffuses out into the rest of the world, their standards, their tech stack, will become superior to ours, because their models are open.
I guess I just believe enough in Nvidia's kernel engineers and CUDA engineers to think that they could optimize—
AI is more than kernel optimization, as you know.
Of course, but there are so many things you can do, from distilling to a model that's well-fit for your chips.
We're going to do our best.
You have all the software. It's just hard to imagine that there's a long-term lock-in to the Chinese ecosystem, even if they have a slightly better open source model for a while.
China is the largest contributor to open source software in the world. Fact. China's the largest contributor to open models in the world. Fact. Today it's built on the American tech stack, Nvidia's. Fact. All five layers of the tech stack for AI are important. The United States ought to go win all five of them. They're all important. The one that is the most important, of course, is the AI application layer. The layer that diffuses into society, the one that uses it most will benefit from this industrial revolution most. But my point is that every layer has to succeed.
If we scare this country into thinking that AI is somehow a nuclear bomb, so that everybody hates AI and everybody's afraid of AI, I don't know how you're helping the United States. You're doing it a disservice. If we scare everybody out of doing software engineering jobs because it's going to kill every software engineering job—and we don't have any software engineers as a result of that—we're doing a disservice to the United States.
If we scare everybody out of radiology so nobody wants to be a radiologist because computer vision is completely free and no AI is going to do a worse job than a radiologist, we misunderstand the difference between a job and a task. The job of a radiologist is patient care. The task is to read a scan. If we misunderstand that so profoundly and we scare everybody out of going to radiology school, we're not going to have enough radiologists and good enough healthcare.
So I'm making the case that when you make a premise that is so extreme, everything goes from zero or infinity, we end up scaring people in a way that's just not true. Life is not like that. Do we want the United States to be first? Of course we do. Do we need to be a leader in every layer of that stack? Of course we do. Of course we do.
Today you're talking about Mythos because Mythos is important. Sure. That's fantastic. But in a few years time, I'm making you the prediction that when we want the American tech stack, when we want American technology to be diffused around the world—out to India, out to the Middle East, out to Africa, out to Southeast Asia—when our country would like to export, because we would like to export our technology, we would like to export our standards, on that day, I want you and I to have that same conversation again.
I will tell you exactly about today's conversation, about how your policy and what you imagined literally caused the United States to concede the second largest market in the world for no good reason at all. We shouldn't concede it. If we lose it, we lose it. But why do we concede it?
Now nobody is advocating an all or nothing. Nobody's advocating all or nothing, meaning we ship everything to China at all times. Nobody's advocating that. We should always have the best technology here. We should always have the most technology here, and the first. But we should also try to compete and win around the world. Both of those things can simultaneously happen. It requires some amount of nuance, some amount of maturity instead of absolutes. The world is just not absolutes.
Okay. The argument hinges on this. They've built models that are specified for the best chips that they make in a few years. Those chips get exported around the world. That sets the standard. Because of EUV export controls, as we said, you're going to move on to 1.6nm. They're still going to be on 7nm, even after a few years from now. It may make sense that domestically they would prefer, "Hey, we've got so much energy, we can manufacture at scale. We'll still keep using 7nm." But on the exporting thing, their 7nm chips have to be competitive against your 1.6nm chips. Their models have to be so far optimized for the 7nm that it's better to run their models on 7nm than to run their models on your 1.6nm.
Can we just look at the facts then? Is Blackwell 50 times more advanced lithography than Hopper? Is it 50 times? Not even close. I just kept saying it over and over again. Moore's Law is dead. Between Hopper and Blackwell, from the transistors themselves, call it 75%. It was three years apart, 75%. Blackwell is 50 times Hopper. My point is, architecture matters.
Computer science matters. Semiconductor physics matters as well, but computer science matters. The impact of AI largely comes from the computing stack, which is the reason why CUDA is so effective, which is the reason why CUDA is so beloved. It's an ecosystem, a computing architecture that allows for so much flexibility that if you wanted to change an architecture completely—create something like MoE, create something like diffusion, create something that's disaggregated—you could do so. It's easy to do. So the fact of the matter is, AI is about the stack above as much as it is about the architecture below.
To the extent that we have architectures and software stacks that are optimized for our stack, for our ecosystem, it is obviously good, because we started the conversation today about how Nvidia's ecosystem is so rich. Why do people always love programming CUDA first? They do. They do. So do the researchers in China.
But if we are forced to leave China, if we're forced to leave China, first of all, it's a policy mistake. Obviously it has backlash. It has turned out badly for the United States. It enabled, it accelerated their chip industry. It forced all of their AI ecosystem to focus on their internal architectures. It's not too late, but nonetheless it has already happened.
You're going to see in the future, they're not stuck at 7nm, obviously. They're good at manufacturing. They will continue to advance from 7nm and beyond. Now, is there a 10x difference between 5nm and 7nm? The answer is no. Architecture matters. Networking matters. That's why Nvidia bought Mellanox. Networking matters. Energy matters. So all of that stuff matters. It's not simplistic, like the way you're trying to distill it.
We can move on from China, but that actually raises an interesting question. We were discussing earlier these bottlenecks at TSMC and memory and so forth. So if we're in this world where you're already the majority of N3—and at some point you'll be N2 and you'll be a majority of that—do you see that you could go back to N7, the spare capacity at an older process node, and say, "Hey, the demand for AI is so great and our capacity to expand the leading edge is not meeting it, so we're going to make a Hopper or Ampere, but with everything we know about numerics today and all the other improvements you described"? Do you see that world happening before 2030?
It's not necessary to. The reason for that is because with every generation, the architecture is more than just the transistor scale. You're doing so much engineering and packaging and stacking, and the numerics and the system architecture. When you run out of capacity, to easily go back to another node… That's a level of R&D that no one could afford. We could afford to lean forward. I don't think we could afford to go back.
Now, if the world simply says… If on that day, let's do the thought experiment, on that day we go, "Listen, we're just never going to have more capacity ever again." Would I go back and use 7nm? In a heartbeat, of course I would.
One question somebody I was talking to had is, why doesn't Nvidia run multiple different chip projects at the same time with totally different architecture? So you could do something like a Cerebras-style wafer scale. You could do a Dojo-style huge package. You could do one without CUDA. You have the resources and the engineering talent to do all of these in parallel. So why put all the eggs in one basket, given who knows where AI might go and architectures might go?
Oh, we could. It's just that we don't have a better idea. We could do all of those things. It's just not better. We simulate it all in our simulator, provably worse. So we wouldn't do it. We're working on exactly the projects that we want to work on.
If the workload were to change dramatically—and I don't mean the algorithms, I actually mean the workload, and that depends on the shape of the market—we may decide to add other accelerators. For example, recently we added Groq, and we're going to fold Groq into our CUDA ecosystem.
We're doing that now because the value of tokens has gone up so high that you could have different pricing of tokens. Back in the old days, just a couple years ago, tokens were either free or barely expensive. But now you can have different customers, and those customers want different answers. Because the customers make so much money—for example, our software engineers—if I can give them much more responsive tokens so that they're even more productive than they are today, I would pay for it.
But that market has only recently emerged. So I think we now have the ability to have the same model, based on the response time, have different segments. That's the reason why we decided to expand the Pareto frontier and create a segment of inference that is faster response time, even though it's lower throughput. Until now, higher throughput is always better. We think there could be a world where there could be very high ASP tokens, and even though the throughput is lower in the factory, the ASPs make up for it. That's the reason why we did it.
But otherwise, from an architecture perspective, if I had more money, I would put more behind Nvidia's architecture.
I think this idea of extremely premium tokens and just the disaggregation of the inference market is very interesting.
The segmentation of it.
Yeah. Alright, final question. Suppose the deep learning revolution didn't happen. What would Nvidia be doing? Obviously games, but given—
Accelerated computing, the same thing we've been doing all along.
The premise of our company is that Moore's law is going to… General purpose computing is good for a lot of things, but for a lot of computation it's not ideal. So we combined an architecture called a GPU, CUDA, to a CPU, so that we can accelerate the workload of the CPU. Different kernels of code or algorithms could be offloaded onto our GPU. As a result, you speed up an application by 100x, 200x. Where can you use that? Obviously engineering and science and physics, data processing, computer graphics, image generation, all kinds of things. Even if AI doesn't exist today, Nvidia would be very, very large.
The reason for that is fairly fundamental, which is that the ability for general purpose computing to continue to scale has largely run its course. And the only way… Not the only way, but the way to do that is through domain-specific acceleration. One of the domains that we started with was computer graphics, but there are many other domains. There's all kinds. Particle physics and fluids, structured data processing, all kinds of different types of algorithms that benefit from CUDA. Our mission was really to bring accelerated computing to the world and advance the type of applications that general purpose computing can't do, and scale to the level of capability that helps break through certain fields of science.
Some of the early applications were molecular dynamics, seismic processing for energy discovery, image processing of course, all of those kinds of fields where general purpose computing is just simply too inefficient to do so. If there were no AI, I would be very sad.
But because of the advances that we made in computing, we democratized deep learning. We made it possible for any researcher, any scientist, anywhere, any student, to be able to access a PC or a GeForce add-in card and do amazing science. That fundamental promise hasn't changed, not even a little bit.
If you watch GTC, there's the whole beginning part of it. None of it's AI. That whole part of it with computational lithography or our quantum chemistry work, data processing work, all of that stuff is unrelated to AI. And it's still very important. I know that AI is very interesting and quite exciting, but there's a lot of people doing a lot of very important work that's not AI related, and tensors are not the only way that you compute it. We want to help everybody.
Jensen, thank you so much. You're welcome. I enjoyed it.
Me too.
Article published