Video · 2025-09-26 · 1h 44m · 6 moments

NVIDIA: OpenAI, Future of Compute, and the American Dream | BG2 w/ Bill Gurley and Brad Gerstner

✦ AI generated

timeline · colored by role

Mechanism

AI scaling now runs on three compounding scaling laws — pre-training, post-training (reinforcement learning practice), and inference-time thinking — rather than the single pre-training scaling law people previously assumed drove progress.

Jensen lays out the mechanism behind the current AI compute boom: pre-training, post-training (RL 'practicing'), and inference-time 'thinking' are three separate, stacking scaling laws, which is why compute demand keeps compounding.

transcript

Jensen Huang: I estimated we now have three scaling laws, right? We have pre-training scaling law. We have post-training scaling law. Post-training is basically like AI practicing... in order to do that, you've got to do inference. So now training and inference are now integrated in reinforcement learning... And then the third is inference... the new way of doing inference, which we appreciate, is thinking. So think before you answer.

Claim◆

OpenAI is likely going to be the next multi-trillion-dollar hyperscale company, so investing in it early — before it reaches that scale — is one of the smartest investments Nvidia can make.

Jensen explains why Nvidia chose to invest up to $100B in OpenAI: he believes it will become the next multi-trillion-dollar hyperscaler, so getting in early is a rare, high-return opportunity, not a requirement of the partnership.

transcript

Jensen Huang: If that's the case, the opportunity to invest before they get there, this is some of the smartest investments we can possibly imagine. And you got to invest in things, you know, right? And it turns out we happen to know this space. And so the opportunity to invest in that, the return on that money is going to be fantastic. So we love the opportunity to invest.

supports · 1

Claim

Claims that China cannot build AI chips, cannot manufacture at scale, or is years behind the US are all false — China's chip and manufacturing capabilities are essentially caught up, not meaningfully behind.

Jensen dismisses the narrative that China can't build advanced AI chips or manufacture at scale, or that it lags the US by years, saying the gap is essentially gone ('nanoseconds') and the US must compete accordingly.

transcript

Jensen Huang: Some of the things I heard, they could never build AI chips. That just sounded insane. Two, that China can't manufacture. China can't manufacture. If there's one thing they could do is manufacture. And three, they're years behind us. Is it two years, three years? Come on. They're nanoseconds behind us. And so we've got to go compete.

provides context · 1supports · 6

Prediction

Human intelligence represents roughly $50 trillion of world GDP, and as AI augments that labor at scale, it implies a market of roughly $5 trillion a year in AI infrastructure capex — a four-to-five-times increase over today's ~$400 billion AI infrastructure market.

Jensen sizes the long-term AI infrastructure opportunity by starting from human intelligence's ~$50T share of global GDP, arguing that even partial augmentation implies roughly $5T of annual AI-factory capex versus ~$400B today.

transcript

Jensen Huang: Human intelligence represents what, 55, 65% of the world's GDP. Let's call it $50 trillion. And that $50 trillion is going to get augmented by something... what's likely to happen is that that $50 trillion is augmented by, let's pick a number, 10 trillion... and 5 trillion of it needs a factory, needs an AI infrastructure. So if you told me that on an annual basis the capex of the world was about $5 trillion, I would say the math seems to make sense.

provides context · 3

Claim

OpenAI's roughly $400 billion of buildout is funded by its own revenue offtake plus equity and debt it raises — not by circular financial engineering — and Nvidia's investment decision is entirely separate from and unconnected to that revenue relationship.

Responding to 'circular revenue' and roundtripping accusations, Jensen argues OpenAI's massive buildout is financed by real offtake revenue, equity, and debt — vetted by smart investors and lenders — and is entirely separate from Nvidia's unrelated decision to invest.

transcript

Jensen Huang: 10 gigawatts is like $400 billion, right? Something like that. And that $400 billion dollars will have to be largely funded by their offtake, right? Their revenue, which is growing exponentially. It has to be funded by their capital, the money they've raised through equity and whatever debt they can raise. Those are the three vehicles... there's the revenue side of it and has nothing to do with the investment side of it. The investment side of it is not tied to anything.

rebuts · 1

Mechanism◆

Even if rival AI chips (ASICs/accelerators) were given away for free, a customer would still choose Nvidia because Nvidia's superior performance-per-watt generates far more revenue from the same fixed power budget, making the opportunity cost of using a lower-performance free chip too high.

Jensen argues Nvidia's performance-per-watt advantage (e.g., 30x Blackwell-vs-Hopper) is so large that under a fixed power budget, a competitor's chip would need to be free — and still wouldn't be worth it, because the revenue lost from lower tokens-per-watt vastly exceeds any price discount.

transcript

Jensen Huang: Let's say you were able to secure two more gigawatts of power... your performance or tokens per watt was twice as high as somebody else's token per watt because you did deep and extreme code design, and my performance was much higher per unit energy, then my customer can produce twice as much revenues from their data center. And who doesn't want twice as much revenues?... Black Wall's 30 times. So you've got to give up 30x revenues in that one gigawatt... even if they gave it to you for free, your opportunity cost is so insanely high. You would always choose the best perf per watt.

rebuts · 1

Highlight slides

Nvidia's $100B OpenAI Bet: Investing Before the Hyperscale Inflection✦ from: OpenAI is likely going to be the next multi-trillion-dollar hyperscale company, so investing in it early — before it reaches that scale — is one of the smartest investments Nvidia can make.Performance-per-Watt > Free Hardware✦ from: Even if rival AI chips (ASICs/accelerators) were given away for free, a customer would still choose Nvidia because Nvidia's superior performance-per-watt generates far more revenue from the same fixed power budget, making the opportunity cost of using a lower-performance free chip too high.Blackwell's 30x Gap Makes the Math Decisive✦ from: Even if rival AI chips (ASICs/accelerators) were given away for free, a customer would still choose Nvidia because Nvidia's superior performance-per-watt generates far more revenue from the same fixed power budget, making the opportunity cost of using a lower-performance free chip too high.

Related episodes

Video · The Cognitive Revolution

Fable's Back, AI Engineer Recap, & SambaNova

→

Article · Interconnects

Kimi K3: The open-weights escalation

→

Audio · All-In Podcast

OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze

→