Jordan Schneider, Dylan Patel, and Jordan Nanos of SemiAnalysis record an emergency pod after Jensen Huang’s resignation as Nvidia CEO. We discuss:
Why Jensen’s exit could become cover for “really stupid laws” — and why Dylan backs the MATCH Act but thinks RASA is a mistake.
The game theory of remote access: renting compute to Chinese firms as margin capture and a kill switch, versus closing the Malaysia loophole outright.
“Flip the table”: why equipment controls beat chip and model controls, and how close Huawei’s silicon now is.
Why no American lab has out-built Kimi or DeepSeek in open models.
Nvidia without Jensen: talent retention, Rubin and Feynman execution risk, and who should run the company.
Colette Kress’s exposure, Supermicro and Megaspeed, and a fight over whether this amounts to the “nationalization” of Nvidia.
Listen now on your favorite podcast app.
Jensen Resigns
Jordan Nanos: Hey, Dylan. Hey, Jordan. Did you guys hear the CEO of the world’s largest and most important company just resigned?
Dylan Patel: No.
Jordan Nanos: Are you serious? What could possibly have led to that? Well, there are allegations of some pretty serious conduct violations. Apparently you can’t sleep with your subordinates and keep your job.
Jordan Schneider: Wait, Jordan — we’ve got Elon, we’ve got Sergey, everyone in Silicon Valley. Once you get to a trillion dollars, I think you’re legislated to start sleeping with your subordinates. There must be something else.
Jordan Nanos: The trillion-dollar class might be fine, but once you get to $5 trillion, you have problems with this sort of stuff.
Jordan Schneider: Then woke 2.0 really starts coming after you.
Jordan Nanos: We’re rotating back.
Jordan Schneider: This is truly a momentous day for the future of America, China, and artificial intelligence — and Dylan somehow looks bored, or just drained. How are you processing it?
Dylan Patel: Dude, I don’t know. You look up to this guy, and then he does these things. It’s not fun.
There’s a lot going on and a lot to contemplate. We shouldn’t even talk about the future of Nvidia — we should talk about the future of open source, the future of AI, and US-China relations. I’ve obviously been a hawk — “or a fucking bald eagle,” as we said on a prior podcast — for quite some time. But I’m worried this is going to be used as cover to pass really stupid laws.
Nvidia without Jensen is a much weaker company. Whether Nvidia’s interests are America’s national interest — most people would say yes, but some people in this audience would probably say no. It’s pretty clear Nvidia’s chips were ending up in China, knowingly or not. It’s probably just knowingly.
Jensen and Nvidia were the biggest opponents of cutting off chip sales to China, and now they’re out of play. So what regulations pass? I hope this means the MATCH Act passes. That is far more important for the US national interest than the Remote Access Security Act (RASA). RASA is a pretty dumb bill.
[Context: The Remote Access Security Act (RASA) would amend export control law to clarify that remote access to U.S.-controlled chips counts as an export. This makes cloud rental subject to the same rules as shipping the chip itself. RASA passed the House 269-22 in January of this year, but no Senate vote has been held or scheduled. A parallel Senate version is stuck in the banking committee, and no version of the current NDAA text includes RASA. The Senate version of RASA directly targets “foreign persons of concern” (generally interpreted as China, Russia, Iran, North Korea, companies whose ultimate parent is headquartered in those countries, and Entity List firms). As designed, RASA would cover only single-customer clusters that are big enough to provide frontier training. This threshold is Commerce’s to set following consultation with chipmakers, AI labs, and cloud providers. RASA would not cover inference rentals, smaller allocations, API access, smuggled chips, or chips already in China. Perhaps most importantly, RASA does not ban anything itself. The text in the Senate says “may” and the House text acts only if the Secretary “determines” a serious risk. There is no deadline nor minimum in place here. Likewise, Commerce can choose whether to wind down existing leases or run them to term. These flexibilities should be viewed as a feature not a bug. This bill settles the legal authority, adds material penalties, and sets up the system to survive a change of administration. RASA, then, gives Commerce the tools to titrate access to rented chips as it deems appropriate.]
Dylan Patel: Biden-style diffusion controls should totally be done. But if the US government passes RASA, that’s going to be really bad for American interests — short, medium, and long term.
The other undertone is that you can’t just ban everything. China can and will exploit the fact that its components are in every single Nvidia server, every Google server, every Amazon server. You cannot build these things without them. So there is a level of tit for tat here. That’s my main take for this audience.
MATCH vs. RASA: How Much Compute Should China Rent?
Jordan Schneider: Let’s start with MATCH and RASA. For those who haven’t been following my MATCH Act shilling over the past few months, the idea is that advanced DUV lithography — alongside EUV, which already can’t get into China — shouldn’t be able to go into China. RASA is slightly different. It aims to close the cloud-access loophole through which Alibaba, Tencent, and ByteDance can currently, legally, access GPUs that are banned from import into China, by setting up cutouts in Malaysia — which you guys have done such a great job untangling.
Dylan, why don’t you expand on why you think RASA is such a bad thing?
Dylan Patel: What is the US national interest? It’s (a) to maintain as large a lead in AI as possible, for as long as possible, and (b) to control China’s access to AI as much as possible. If that’s your goal, you do not prevent China from having any compute, because that actually means the US will have no compute either. If you fully ban everything, I believe China will start waging economic war too.
In 2022, I stayed up all night reading the October 7th chip restrictions. I wrote that China and the US were officially in a new Cold War — an economic war. I found about 20 loopholes and posted about them that very night, and I’m just one dude, not an army of lawyers. I was hoping they’d get patched. Many still aren’t.
One key premise of the original restriction, which I think is mostly true, is civil-military fusion: the Chinese military and state can and will appropriate any resource for military and intelligence purposes, even if it’s completely commercial. We’ve seen it with Douyin and TikTok, which is about as consumer as you can get.
At the same time, China will respond. We’ve seen it with the trade wars and the auto tariffs — almost shutting down American auto plants, threatening rare earths. China has threatened a lot of actions, carried out some, then pulled back to show its fangs. If you ban chips from China completely, including rental and deployment, there is no reason for China not to blow the whole thing up. You have to have some tit for tat.
The optimal American policy is the minimum amount of compute China can have, with the maximum American oversight, such that China won’t blow up the supply chain — because we very clearly do not have the stomach to disentangle the supply chain. We just don’t, over and over again.
Given that China can’t have zero chips, the way to minimize its compute and maximize control is Biden-style diffusion rules. Yes, it’s basically a bill written for hyperscalers, but hyperscalers can rent to Chinese companies, and America has the power to say, “Turn it off at any moment.” You want all of China’s compute to look like that.
If Alibaba, ByteDance, Moonshot, whoever, rent compute from American companies outside China, and that’s their main source of compute, this is supremely awesome. First, China won’t want to send its data abroad in many cases — and if it does, it’s probably stuff the US intelligence apparatus can handle. Second, “if shit goes tits up, you can turn it all off.”
If you cut off remote access, there’s even more incentive to (a) smuggle and (b) ramp the domestic industry — which they’ll do anyway, by the way, which is why you need the MATCH Act. And (c), you remove any reason for them to continue the status quo, in which a lot of Chinese industry is dedicated to making electronics for American companies. The largest PCB supplier to Nvidia is a Chinese company, for example — and it’s not just Nvidia.
You have to come at this from a game-theory perspective. The best thing for America to do is allow remote access to GPUs, because the alternatives are much worse.
Jordan Schneider: A few pushbacks. There’s a piece of the Chinese ecosystem — Xi Jinping likely included — that is deeply embarrassed that these companies so badly want to import GPUs. Part of what we’ve learned during the Trump administration is that even when Trump offers them, the Chinese government says, “I’m not sure we even want these companies to buy them.”
So yes, we have an equilibrium right now, but I can definitely see chunks of the Chinese government looking at something like RASA and saying, “Great. These companies thought they could get cute and not invest in the domestic ecosystem. Now they’ll really have to play ball.” There are also other things going on with Xi that would make him less likely to declare economic warfare on the US by doing something as dramatic as shutting off global chip manufacturing —
Dylan Patel: But we already shut them off. Why wouldn’t they retaliate?
Jordan Schneider: Because then you get into a tit for tat. And by the way, their economy isn’t growing and unemployment’s at 20%.
Dylan Patel: Every time we’ve gotten into that tit for tat, it’s led to a compromise. Tariffs got reduced. Some restrictions went up, but mostly China didn’t just take it. They retaliated, and the US economy and administration weren’t willing to stomach the pain.
Jordan Schneider: They didn’t retaliate in 2022.
Dylan Patel: Well, the 2022 rules had so many loopholes.
Jordan Schneider: Okay, they didn’t in 2023. They didn’t in 2024. And look, Trump kept doing the war in Iran even though it cost him 10 Senate seats. There’s a tolerance for pain, especially in a lame-duck second term when the whole cabinet is leaking against Lutnick because they’re pissed. There was an Axios report saying China’s already breaking the trade deal and already waging economic warfare against the US. I don’t think escalation dominance should be assumed on the part of the Chinese government.
Dylan Patel: The other argument: under remote access, if China has compute outside China, it’s paying humongous margin dollars to the US ecosystem. Take a megawatt of compute outside China — there are multiple layers of margin stacking. Nvidia makes about $7 million of profit per megawatt deployed. The cloud provider makes another $7–10 million at today’s pricing.
That’s a tremendous amount of profit for us. Meanwhile, American companies like Google, Amazon, and Meta either own the cloud or deploy their own chips, so their cost structure is completely different. Why does China win? Economies of scale and a better cost structure. Why does American AI win? Our cost structure is better and our economies of scale are larger. So let them ship all this money out. If ByteDance wants to compete with Meta globally on video recommendation, and you don’t let ByteDance buy chips directly but do let it rent them, Meta potentially has a quarter of the compute cost.
Jordan Schneider: But here’s the thing. Zuck was totally fine letting TikTok spend billions of dollars on Instagram ads. If you just don’t let them play the game, you win by default.
Also, Dylan, even in the shaky AI times of the past week, isn’t there enormous demand? Wouldn’t the US be better off taking however many gigawatts are currently flowing through Malaysia into the Chinese ecosystem and using them to help OpenAI, Anthropic, and DeepMind make better models — and more money?
Dylan Patel: There’s a reflexive nature to this. I’m going to devil’s-advocate this for a while — I think we’re generally pretty close, but it’s fun so people hear the argument.
This was, to some extent, Jensen’s argument: AI diffusion across the US is largely aided by Chinese models. If you add up all the application-layer companies running on open models, it’s well north of $10 billion in ARR. That’s not some tiny amount of open-model revenue going to American companies. DeepSeek’s revenue from DeepSeek V4 Flash is less than “fucking Baseten and Fireworks” and all the other guys serving the Chinese model. America is making so much money off Chinese models, and it lets us accrue and invest more and more.
Yes, ByteDance bought a bunch of ads on Instagram and Zuck didn’t see the threat immediately. Of course — everyone’s short-sighted. Everyone makes mistakes. “Jensen got a little horny, right? It’s his short-sighted mistake at the end of the day.”
Jordan Nanos: Come on.
Jordan Schneider: He did more than that.
Dylan Patel: Yes — it’s clear what he did over a long period of time.
Jordan Schneider: Again, it’s not just the women. It’s doing more than turning a blind eye while his suppliers not only did the Malaysia thing — which is kind of legal, even today — but brought the chips in.
Dylan Patel: The Malaysia thing is totally legal. Not “kind of” legal.
Jordan Schneider: Sure, but it’s jumping through unethical hoops.
Jordan Nanos: It’s about the letter of the law versus the spirit of the law. Jensen is in closed-door meetings with a lot of US leaders who clearly want to slow China’s advance, and he has decisions he can make —
Jordan Schneider: I don’t think that’s why he’s losing his job. I think he lost his job because there’s evidence they knew the chips were going into China.
Dylan Patel: 1,000% — that’s completely illegal.
Jordan Nanos: Look at the circumstantial evidence around Jensen. His Taiwan sales manager was indicted for smuggling chips out of Taiwan. His closest OEM partner, Supermicro, had co-founder Wally Liaw arrested. One of Supermicro’s sales managers is currently a fugitive in Taiwan. So it’s not out of the blue that they’re cracking down, and that people around him are being caught with clear evidence. Take the picture of the Supermicro Taiwan guy with a hair dryer in his hand, moving stickers between real and fake servers to pass an audit in September of last year. That’s pretty damning evidence that they were aware.
Dylan Patel: And no honeypot was required to get Jensen to do that with Supermicro.
Jordan Nanos: Well, the evidence that has to pass the test would be Jensen actually knowing about this, being complicit, and doing something. Either way, I hope they can establish what he knew.
Dylan Patel: If I had to guess, the board has evidence on this.
Jordan Schneider: How else do you explain Nvidia no longer reporting Singapore revenue?
Jordan Nanos: The circumstantial evidence is there. Even with the affairs, people would say, “Well, it’s not technically illegal” — except the spirit of his actions runs against people who hold a different view of what the US-China relationship should be and of the technology involved. And at a convenient time — right around the midterms — he gets whacked.
Dylan Patel: The cover is gone. The cabinet basically took the blue wave as, “Okay, it’s fine, we can blow the cover now,” and it’s over for Lutnick and Jensen. Trump and Lutnick always wanted to make deals, and they used export restrictions as a deal-making tool. But in another sense, “it’s the fucking law, right? You can’t negotiate with the law.” You make a law, and you enforce it. So there are clearly issues there.
Jordan Nanos: Enforcement has been a big question so far. The Biden-era diffusion framework was passed and then just not enforced. It’s still on the books — Trump has just directed people not to enforce it.
Dylan Patel: Really? They didn’t remove the diffusion rules?
Jordan Schneider: It was super weird. There was a comment and a footnote, but they never repealed it in 2025.
Dylan Patel: So these are laws in effect? I thought we’d have to repass them.
Jordan Nanos: It’s not really a law in the traditional sense.
Dylan Patel: An executive action?
Jordan Nanos: No — it’s basically guidance to BIS.
Jordan Schneider: And Kessler never published real guidance saying, “We are no longer doing this.” I don’t know what the fuck Kessler has been doing. Apparently he shows up at the office. He just put a footnote somewhere saying, “By the way, guys, don’t worry about this, we’ll bring you a new thing soon.” That was in May 2025. The indecision, incompetence, and cognitive dissonance in this policy process are really breathtaking.
Jordan, what’s your RASA take? Where do you set the dial on how many chips we let China access remotely?
Jordan Nanos: What Dylan said matters: the US being able to monitor what China does with the chips, and having the option to cut off access and use the chips for its own purposes if the intelligence wars escalate. That’s important optionality.
The other thing is how clearly American companies benefit from a mutually beneficial relationship with China — with Chinese companies, Chinese researchers, Chinese software, and Chinese people. Lots of people at the frontier labs today were born in mainland China. We sometimes work with China-based engineers at Nvidia and AMD who have access to early products, bring up new GPUs, and work on software and low-level hardware. As tensions escalate, all of these relationships break down and everything gets harder for everybody. So insofar as anybody is pro-AI, this looks like a bad day for AI progress. That’s my high-level take.
On RASA specifically: as you said, China is trying to get off Nvidia GPUs. In our tracking, GLM was released in February with support for seven domestic accelerators — not just Huawei, but Moore Threads, Hygon, Cambricon, Kunlunxin, MetaX, and Enflame. DeepSeek published a megakernel for Huawei Ascend with its V4 model. Meituan claims it trained LongCat entirely on Ascend. In the US, people don’t even train on AMD — it’s only Nvidia or TPUs. So that shift is happening with or without remote-access restrictions; RASA just affects motivation.
It’s also about inference. Are you going to let ByteDance compete globally using inference capacity on Nvidia GPUs from these providers? Or will you keep it out of the US and Southeast Asia and make it serve from within China? That’s an open question.
Finally, I and many people I know at American labs learn from people in China. If you cut off those direct working relationships, it’s a one-way street. It’s really hard to see how you roll that back.
Dylan, I’d love your take on this: I think you can get all the meaningful restrictions by focusing on ASML and the MATCH Act. Expand controls beyond EUV to DUV tools. China needs a lot of them to produce the chips that go into these data centers. Make them build lithography from scratch. That’s the easiest way to stop their progress.
Flip the Table: Equipment, Scale, and Open Models
Dylan Patel: What people don’t understand is that if you remove one item at the top of the supply chain, they’ll just engineer around it. I’ve used this analogy 100 times: it’s a jigsaw puzzle. “You remove one piece, they’ll fucking engineer the one piece. If you blow up the board, it’s harder.”
This is an economic war. If you think you can solve it without huge ramifications across supply chains, you won’t. You have to flip the table. The best way to do that is to block them from buying the relatively low-value goods and force them to buy the expensive stuff. Don’t let them buy ASML equipment and make their own chips and models. Don’t let them buy chips — make them rent. Don’t let them rent chips — make them buy the AI. Push them up the value chain, where the value gets larger and larger. Do that, and China is at a massive economic disadvantage. This is a scale game, and therefore the West will win.
They’re going to be pissed about whatever you ban. So ban the thing at the bottom of the totem pole and chop the legs out from underneath them. Don’t just attack the tip of the spear — the AI, the chips. Focus on how you make the chips: the equipment. That’s what matters more.
Jordan Nanos: It’s also a timing thing. We’re watching China rise, and in response to a lot of these actions, they don’t want to escalate, because the longer they wait in these intelligence wars, the more time they have to catch up. There’s a lot of discourse about whether the gap is widening or shrinking. I’d say the gap in model capabilities is wider at this point, but the gap in chip capabilities is as close as it’s ever been.
Jordan Schneider: But the question isn’t just capability — it’s scale. That’s where it comes back to equipment. It’s not ideal that China broke through the 14-nanometer line we tried to draw in the sand in 2022. But it’s a scale game. Making 1,000 chips isn’t enough; you need 100,000. You need megawatts upon megawatts of advanced hardware to train and deliver AI that makes a difference to your economy. Get the equipment side right, and it’s so much harder for the Chinese ecosystem to compete if it can’t rely on the Japanese and the Dutch. China already can’t buy from the Americans, so this isn’t screwing over US companies on the SME side. It’s making sure the Japanese, Dutch, and Koreans can’t backfill what we won’t sell.
I hear you, Jordan, on the interpersonal stuff. There’s an aspect of this that is deeply tragic. AI has the potential to deliver incredible things for humanity, and there’s a side of this that’s a beautiful, humanity-scale experiment. The problem is that the US and China have a lot of issues with each other, and China happens to be run by an authoritarian leader presiding over a Leninist system.
That’s not Liang Wenfeng’s fault — it absolutely isn’t. There are old blog posts where he talks about how he’s not into it, and you can read in his writing a vision of the world that’s much more unified and holistic than what you see in Xi Jinping’s speeches, where the frame is a struggle for the commanding heights. But Liang Wenfeng doesn’t run the country. Xi does — he’s been doing it for well over a decade, and he’ll be around for a long time to come. That’s the reality people need to operate under, and Jensen clearly didn’t. He didn’t buy into this framework. He didn’t think China advancing at the same clip as the US was a problem. And there’s enough consensus among America’s elected officials that he was wrong that we’ve ended up where we are.
I want your take on how bad it is for American innovation that Jensen’s now off the table. And we should talk about open models — he was the big champion. Dylan, as you said earlier, I still don’t understand why someone didn’t spend a few billion dollars to build something better than Kimi or DeepSeek and open-source it.
Jordan Nanos: It’s not easy.
Dylan Patel: It’s not money. It’s talent, man. The most talented people primarily work at Anthropic, OpenAI, DeepMind, and Meta’s TBD Lab. Those who don’t are split across 100 different startups. Yes, Thinking Machines has a crack team, but is Thinking Machines as good as Moonshot? No. Skill-wise, Moonshot is probably just as good as the US labs — they just have a big deficit in compute and deployment.
Without a scale advantage, you don’t win. These smaller labs have spent a lot of money on models, but DeepMind, Grok, and Meta have regularly fallen behind the Chinese labs’ frontier while spending more. So it’s not just “spend a few billion dollars and you get a better model.” The innovation and the quality of the engineers and researchers at Chinese labs is “just very fucking good.”
Jordan Nanos: By the way, Nvidia did spend that money, and still didn’t produce anything with Nemotron better than the Chinese labs’ open-source models. Jensen’s a big supporter of open source, but pouring in money and time doesn’t guarantee results. It’s definitely hard. You need compute, people, and data.
Jordan Schneider: So we just have to get Alexandr Wang on the case. That’s the only successful zero-to-one we’ve had in the past three years.
Jordan Nanos: I mean, they hired a ton of great researchers, have a ton of compute, and spent a bunch of money on high-quality data. That’s part of it.
Nvidia After Jensen
Jordan Nanos: I see the discourse shifting toward two things that I’m not completely against but want to add nuance to. First, that ramping up export controls via RASA, MATCH, and the diffusion framework is obviously good and should happen immediately. I’m not sure people are thinking through all the implications, which is what we tried to discuss here.
Second, people immediately say, “Jensen leaves Nvidia, so the company’s going to zero. It could never run without him; it has such a unique work structure.” That’s all true, but the more nuanced take is that in the short term, Nvidia is fine. They need to deliver on customer commitments and build a bunch of GPUs with the supply chain they’ve got. Nobody could set them up for success as well as Jensen, but now they just need to execute on what he set up — Rubin and Rubin Ultra. A lot of people know their marching orders, their responsibilities, the mission.
The really big test is talent. Jensen has been incredibly good at retaining talent, especially on his executive team. There have been very few high-profile departures. A lot of people retire, and basically only Rene Haas has left to become CEO somewhere else, at Arm. EVPs, SVPs, and VPs rarely leave Nvidia. Now Jensen and his mission aren’t at the helm, and somebody needs to keep giving these people compensation and direction to retain them. This is when the sharks are in the water: VCs will offer people billions of dollars to start their own chip companies and neolabs.
We’ll see in the medium to long term whether Nvidia can operate without Jensen. If a year from now we’ve seen a bunch of departures, I’ll be really concerned about its future. If a year from now everybody’s still rowing in the same direction — Colette is driving things, or a technical leader like Ian Buck has emerged, or they move Michael Kagan up from CTO, whatever that looks like — I’d be much more confident.
But dude, people are underestimating Nvidia’s dominance: the supply chain, sales and support, the relationships with the neoclouds, the investments in labs and startups, the balance sheet it can use to backstop everybody and make them partners. It can invest in everybody and acquire companies as soon as they’re successful. It’s still Nvidia’s to lose, and there’s a very high chance it’s doing just fine two, three, four years from now.
Jordan Schneider: Another open question: say some people get poached, raise big rounds, and start new things. Is that really the worst thing for the Western ecosystem — different people trying new bets? The other question is that we’re going from a founder to a non-founder. Beyond executing on the roadmap, there must be some big, wacky, out-of-consensus bets Jensen would have made over the next five or 10 years that might not have paid off for another five or 10. Do those still happen, but half as much? Or do they just happen at some EVP’s new startup instead of inside Nvidia? That’s my question as we go from the Jobs era to the Tim Cook era of Nvidia.
Jordan Nanos: I’m certainly not saying this is a good day for Nvidia.
Dylan Patel: I don’t understand why Nvidia’s not down a lot more.
Jordan Nanos: It’s really bad, for sure. But people tell me Nvidia’s a zero, and I say, take the Apple and Steve Jobs analogy. If you bought Apple stock when Jobs went on leave and then passed away, you’d have made a ton of money since.
Dylan Patel: But didn’t he come back and —
Jordan Schneider: No, no — gone 2.0. This is Jobs leaving in 2011, not in 1985.
Dylan Patel: Okay, got it. I thought you meant the original exit.
Jordan Nanos: Apple in 2011, looking back —
Dylan Patel: The thing is, Steve Jobs was a maniacal guy who cared about every aspect of his company, but I don’t think he exerted as much control over Apple as Jensen does over Nvidia, or as Elon does over his companies.
Jordan Schneider: Really?
Dylan Patel: I don’t think so. I don’t know Apple as well — I’ve only read some books. He was tremendously important, but it’s not the same level of work. Steve Jobs was a pure visionary. Jensen is negotiating every supply-chain contract. Steve Jobs wasn’t doing that.
Jordan Schneider: So, Dylan, if they start faltering, on what dimensions?
Dylan Patel: Rubin. If there are any manufacturing issues with Rubin and Jensen’s not there, things get solved slower. They get solved by committee. They don’t get escalated. I don’t know how many times at SemiAnalysis we’ve fucking emailed Jensen to get something solved. Why are we emailing Jensen to fix something in some open-source software? Yet that’s the only way it gets solved — and we’re irrelevant compared to actual companies in the space. From what I’ve heard around the ecosystem, Jensen goes out and solves problems a lot. How many times does Elon tell Jensen something, Jensen tells his team, and the team fixes it — because working team-to-team hadn’t worked?
Jordan Nanos: Right. But all of these reasons Nvidia could have problems going forward were already present with Jensen at the helm. They have a big problem with all the custom chips coming up —
Dylan Patel: Exactly. But I’m saying Rubin could slip. The yield curve could look worse. Feynman could slip, or be less aggressive. One of my favorite stories: Volta didn’t have Tensor Cores until three months before tape-out. Jensen decided that. There were bickering little people, and Jensen said, “No, AI is the way. Let’s do fucking Tensor Cores. I don’t care that we’re three months from tape-out and you say it’ll take six months to incorporate the change. You’re going to do it, and you’re going to do it now.”
Stack up 100 of those decisions. Without them, Feynman will be much worse and come later. Rubin’s yield ramp could be slower. Rubin Ultra could be slower, or might not get done at all, so Nvidia ships lower volumes. There are a bajillion software decisions too. DeepSeek is open-sourcing all this software, like DeepEP, and Moonshot is making KDA work. That stuff will take much longer to run at high efficiency on Nvidia than before. Maybe it’s only a week each time, but add up all these decisions, his leadership, and the supply chain giving Nvidia 70% of its capacity because it’s Jensen asking. All of it compounds, and it could be that in 2029 the best chips aren’t Nvidia’s and Nvidia doesn’t have the majority of the supply chain.
Jordan Schneider: That’s relevant to the stock. But this is 2026, not 2022. There are lots of other competent teams going after this, where before there was really only one.
Dylan Patel: Nvidia is still the high-water mark. It’s still the leader, still driving the industry. If Nvidia slows down, it’s not that someone else accelerated — Nvidia slowed down, so someone else is the best.
Jordan Nanos: It’s a question mark. Maybe it depends on your inference situation.
Dylan Patel: No matter what, the industry progresses slower. Hardware progresses slower without Jensen at the helm of Nvidia, period. Yes, there are competitors, but without Nvidia in the lead, people run less fast — or even if they run as fast as before, the leader is no longer as relevant, so progress has slowed. This is a real slowdown.
Who’s next — and is this nationalization?
Jordan Schneider: Man, this is pretty morbid for all of us. It’s sad. Very Shakespearean.
Dylan Patel: Dude, I’m so sad. I’ve been a Jensen fanboy since I was a teenager. I don’t think people understand how much I fucking look up to the man and love him.
Jordan Nanos: It’s a sad circumstance. A lot of us were always impressed by Nvidia, and watching people try to excuse conduct this bad — it’s just awful to read about. I see all the memes —
Dylan Patel: Dude, the memes are fire. The past few days are up there with the 2016 Republican primaries. The memes are that fucking good.
Jordan Nanos: I laugh and then I’m sad. That’s what the memes are doing to me right now.
Jordan Schneider: When I was a little high schooler into politics, Eliot Spitzer was my hero. He was crusading against corruption, then became governor of New York. Then it came out that he’d signed an anti-prostitution law while paying for $2,000-an-hour escorts, and he had to resign. That was my wake-up moment: men with egos aren’t necessarily all they’re cracked up to be. So it’s not surprising, but it is depressing. Humans are frail. Don’t meet your heroes. But we can still look up to Lisa Su, right? She’s still there for us.
Dylan, who’s your dream new Nvidia CEO? Who are we putting in if we can snap our fingers?
Dylan Patel: If I could snap my fingers, Lisa Su, but I don’t think she’d do it. Charlie Kawwas, Johny Srouji — Charlie Kawwas would be the best.
Jordan Schneider: How about you, Jordan?
Jordan Nanos: If Charlie’s on the table, I’ll just say Hock Tan, man. Bring the guy in, tear it to pieces.
Dylan Patel: Hock is too much of a businessman. Charlie is a bit more technical and less of a businessman than Hock, even though he’s a really good businessman too. For this dynamic, you actually want Charlie.
Jordan Nanos: My personal pick is a lady named Colette Kress. I really look forward to working with her.
Dylan Patel: We do love Colette.
Jordan Nanos: She’s great.
Jordan Schneider: I know you guys are buds, but there is no way the CFO of Nvidia isn’t implicated in whatever has been happening with these chips. It’s not possible. I can’t fathom a world —
Dylan Patel: We don’t know how deep this web goes. But I don’t see how Jensen and Colette weren’t both — I don’t know. It’s hard. Colette’s not the one in Asia making the agreements. Maybe she could say, “I didn’t know. I signed a lot of pieces of paper.”
Jordan Schneider: But she’s going to have to testify in front of Congress about just how much she didn’t know. There will be congressional investigations. It may be true, but that’s why you want a technical CEO, or someone from outside. I don’t understand how someone deeply involved in operations and interacting with clients —
Dylan Patel: Right. And a new CEO won’t be called to testify for days and days. They get to focus on the company.
Jordan Nanos: Look, the fact that this originates in the intelligence community makes it unlikely that whoever is immediately installed as CEO has deep knowledge of it. The board has clearly been briefed on something Rubio knew back in May, before the China visit. Anything else would be really strange.
Jordan Schneider: No, this isn’t about a Chinese spy. Look at all the facts with Supermicro. It’s not the affair anymore. It’s Supermicro. It’s Megaspeed. It’s chip smuggling into China and the willful covering of your eyes —
Jordan Nanos: Willful ignorance.
Jordan Schneider: Anyone complicit in that is going to have a really hard time staying out of the line of fire.
Jordan Nanos: I’m watching the Supermicro indictment closely to see how the prosecution plays out. They flipped the hair-dryer guy who was moving the stickers — the one who deleted his encrypted messaging app the day he got arrested. Apparently he’s working with prosecutors now. In the connected world we live in, there aren’t many ways to hide this stuff and stay willfully ignorant when you’re running things in Taiwan from the US. But again, I’d find it really strange if they immediately installed somebody complicit in the same conduct that’s motivating the change.
Jordan Schneider: I hope not. I hope she does a great job, brings all our new chips online, and brings me AGI. We’ll see.
Jordan Nanos: My closing thought: there’s potential for Nvidia to go into civil war here, because she’s interim, not permanent. Every company has a constant clash between finance, technology, and sales. Sales wants to do deals, technology wants to make big bets, and finance has to figure out how to pay for it. Finance is in charge now. So we know what happens to the business: it’s probably very profitable for a while, and then in the medium to long term it probably looks a lot different.
On the political side, I really hope they pass the fucking MATCH Act. Think this stuff through, guys.
Dylan Patel: You can be a hawk and still debate RASA, but it’s hard to debate MATCH.
Jordan Nanos: It seems obvious. Anyway, we’ll be watching closely.
Jordan Schneider: Tune in for more. I did my emergency pod on Friday and thought, “How many days before I have to do another one?” Six days. Tight turnaround.
Jordan Nanos: Daily emergency pod. Internally, we’re saying every one of our pods has to be titled “emergency” right now. Dylan, what happens next week? How do we ramp up from here? We thought Anthropic was the biggest story in the capital markets, and now the CEO of the biggest, most valuable, most important company in the world is gone.
Dylan Patel: AI 2027 said China wakes up to AI in November. So we’ve got a big bear awakening.
Jordan Nanos: Next November, or —
Dylan Patel: No, no — we’ve got to nationalize it.
Jordan Schneider: We’ve got to nationalize Anthropic and SpaceX, and then Dario gets thrown in jail. No, no — I’m just saying, for season two, we’ve got to up the stakes for our next emergency pod.
Jordan Nanos: Just to be clear, how would nationalizing a company that size functionally play out?
Jordan Schneider: I have no idea. It’s not actually going to happen. Sorry — we should have ended this podcast three minutes ago.
Jordan Nanos: We lost the thread. We can cut all this —
Dylan Patel: But I know Jordan is a socialist and loves and admires China so much.
Jordan Schneider: So Canadian.
Dylan Patel: No, no — Schneider, Schneider.
Jordan Nanos: Oh, Schneider.
Jordan Schneider: You’d need to pass some war-powers law. There would need to be legislation around AGI. I don’t think the executive branch can just nationalize it.
Jordan Nanos: Cut this part out. But what I’m getting at is that we are watching the nationalization of Nvidia play out right now. This is the equivalent of Jack Ma disappearing.
Jordan Schneider: I don’t think so at all.
Dylan Patel: I disagree completely.
Jordan Schneider: All he had to do was not sell chips to people smuggling them into China, and he would have been fine. That’s all he had to do.
Jordan Nanos: So he does something against the national interest, and the intelligence community comes in, takes him out, and installs somebody it has more influence over.
Jordan Schneider: When you break the law, there are consequences.
Dylan Patel: Just don’t break the fucking law.
Jordan Schneider: Just don’t break the law.
Jordan Nanos: Okay, sorry, guys. Sorry for the conspiracy theory.
Jordan Schneider: He can advocate all he wants. He can scream it from the rafters.
Dylan Patel: He could have pushed H20s all he wanted — that wasn’t breaking the law. H100s were.
Jordan Nanos: Wait, all you’re saying is that he deserved to be nationalized — which, fair enough.
Dylan Patel: No, there’s a huge difference between “don’t break the law” and nationalization. “This is a fucking massive chasm.”
Jordan Nanos: Are you saying this is not —
Dylan Patel: They are not nationalizing Nvidia right now. That’s crazy.
Jordan Schneider: No chance. Absolutely no chance.
Jordan Nanos: What would they have to do for it to look like nationalization to you?
Jordan Schneider: What does nationalization even mean to you?
Dylan Patel: Does it mean a CEO who doesn’t break the law? Any of the people we’ve suggested for CEO wouldn’t be nationalizing the company. They’re just logical CEOs.
Jordan Nanos: Right. So let’s see who it is. Jared Kushner for CEO.
Jordan Schneider: When Donald Jr. becomes CEO, then we’ll really know who was behind this from the beginning.
Jordan Nanos: Maybe this is too far gone for you guys, but I think that’s exactly what’s happening right now. Anyway, cut this part out.
Jordan Schneider: It’s aligning the country with the laws on the books.
Dylan Patel: Leave it in, Jordan.
Jordan Schneider: All right, we’ll leave it in.
Jordan Nanos: All right — you guys are on camera pushing back on my wild conspiracy theory. Nothing I say is financial advice. Don’t come after me, please.