All podcasts / BG2 Pod / Summary

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

2025-09-26 - 104 min - source - Read full transcript
Brad Gerstner (host)Clark Tang (host)Jensen Huang

Key insights

Nvidia's OpenAI investment and its OpenAI compute deal are separate decisions: the equity stake is an opportunistic bet on OpenAI becoming a multi-trillion-dollar hyperscaler, unrelated to the compute contracts.
Huang says he believes OpenAI is likely to become the next multi-trillion-dollar hyperscale company alongside Meta and Google, calling the chance to invest before it gets there one of the smartest investments imaginable, and stressing Nvidia is not obligated to invest, it was simply offered the opportunity.
openai-nvidia-partnership
AI compute demand is now compounding across three distinct scaling laws rather than one, which is why Huang says he underestimated last year's already-aggressive forecast.
He describes pre-training scaling, post-training scaling (reinforcement learning as the model 'practices' a skill through inference), and inference-time scaling (models 'thinking' via research and reasoning before answering) as three separate exponentials now stacking on top of each other, each independently driving more compute per query.
ai-compute-economics
Wall Street's sell-side consensus has Nvidia's growth flatlining to about 8% from 2027-2030, a forecast Huang says he is comfortable with because it reflects a persistent gap between what AI-industry leaders are actually building toward and what the 25 analysts who forecast Nvidia's growth for a living believe.
Gerstner frames this as a 'massive divergence of belief' between what Sam Altman, Sundar Pichai, and Satya Nadella say versus consensus Street models, attributing the gap to skepticism that today's compute shortage will flip into an oversupply glut.
ai-compute-economics
Huang argues fears of circular revenue and roundtripping between Nvidia and OpenAI misunderstand the deal structure: OpenAI's roughly $400B, 10-gigawatt buildout is funded by its own growing offtake revenue plus separately raised equity and debt, not by Nvidia's investment.
He distinguishes the compute purchase (funded by OpenAI's revenue, equity raises, and debt, subject to normal lender/investor scrutiny of OpenAI's own numbers) from Nvidia's equity stake (a standalone opportunistic bet), and notes OpenAI has no obligation to use Nvidia chips if a competitor's chip proves better.
openai-nvidia-partnership
Huang's total-cost-of-ownership argument is that competing ASIC chips could be given away for free and Nvidia would still win, because performance-per-watt directly converts scarce gigawatts of power into more customer revenue.
He walks through the math: if Nvidia's tokens-per-watt is roughly 30x a rival chip (using Hopper-to-Blackwell as the proxy gap), a customer with a fixed power allocation gives up 30x the revenue by choosing the cheaper chip, an opportunity cost far larger than any gross-margin discount a competitor could offer.
competitive-moat-strategy
Nvidia's annual release cadence (Hopper, Blackwell, Rubin, Rubin Ultra, Feynman) combined with 'extreme codesign' - co-optimizing chip, networking, and software across six to seven chips simultaneously - is what let Blackwell hit a 30x performance jump over Hopper despite the end of Moore's law-driven transistor gains.
Huang says this pace would be impossible without using AI internally to design Nvidia's own chips, and that the annual cadence also locks in years of supply-chain visibility (wafer starts, HBM, packaging) that competitors and ASIC challengers cannot match.
competitive-moat-strategy
Huang lays out a three-tier taxonomy of chip businesses - architectural platforms (x86, ARM, Nvidia GPU), ASICs, and customer-owned tooling (COT) - and argues most current AI ASIC projects (Google TPU aside) will stay stuck building one component of a much larger, fast-changing AI factory system.
He credits Google's TPU advantage to starting TPU1 years before the AI market existed, the classic startup pattern of building before a market is large; he argues new ASIC entrants face a market that has already evolved from a single GPU chip into a complex, disaggregated AI factory (citing Nvidia's new CPX chip for context processing/video), making it far harder to catch up now than three to five years ago.
competitive-moat-strategy
Huang argues that cutting Nvidia out of the China market via export restrictions was a unilateral disarmament that let Huawei build monopoly profits and a stated three-year plan to catch up to Nvidia.
He notes Nvidia previously held roughly 95% market share in China before being forced out, and that Huawei and Alibaba are now announcing plans to build data centers around the world funded by that monopoly position; he frames continued competition in China as good for both US influence and for keeping Chinese engineers inside a US-aligned technology ecosystem.
us-china-ai-race
Both host and guest treat the US-China AI contest as an existential, Manhattan Project-scale race that is being funded by private companies rather than government, with a sharp drop in Chinese AI PhDs choosing to stay in the US treated as an early warning indicator.
Gerstner cites a source at a leading US AI lab estimating that roughly 90% of top Chinese AI researchers wanted to stay in the US three years ago, versus 10-15% today, with many now considering Europe instead; both frame this immigration pipeline as a leading KPI for whether the US keeps its innovation edge.
us-china-ai-race
Huang supports the newly passed 'Invest America' program giving every child born in the US from 2026 onward a $1,000 investment account at birth, funded partly by corporate contributions including Nvidia's, as a way to spread AI-era wealth creation broadly.
Gerstner, who says he helped drive the policy, frames it as updating the social contract so every child becomes a shareholder in the country's future growth rather than being purely left out as AI-driven wealth concentrates; Huang calls it a genius idea and ties it to the broader theme of the American dream and 'the right to rise.'
american-dream-policy
Huang rejects the premise of mass AI-driven unemployment, arguing intelligence is not zero-sum: Nvidia's own productivity gains from internal AI use have led it to hire more people, not fewer, because higher productivity funds pursuit of more ideas.
He argues the fear of AI eliminating jobs assumes humanity has run out of ideas and things left to do; instead he expects AI to change tasks within jobs (eliminating some, creating others) while overall economic growth and hiring continue, framing this as the reason to build AI systems and companies that also reinvest in re-industrializing American manufacturing.
american-dream-policy
Huang sizes today's AI infrastructure market at roughly $400B a year and argues a GDP-based bottoms-up estimate implies a runway to roughly $5 trillion a year in annual capex if AI eventually augments a meaningful share of the world's roughly $50 trillion in labor-driven GDP.
His math: if AI augments about $10 trillion of that $50 trillion in human-intelligence-driven GDP, and roughly half of the resulting token-generation revenue at 50% gross margins needs to run through AI infrastructure, that implies close to $5 trillion in annual capex, a 4-5x increase over today's roughly $400B market.
ai-compute-economics

Companies

Techniques and frameworks

Summary

Jensen Huang joins Brad Gerstner and Clark Tang about a year after their last conversation, opening with the newly announced OpenAI-Nvidia partnership: Nvidia will invest up to $100B in OpenAI and become a preferred compute partner as OpenAI self-builds its own AI infrastructure for the first time, layered on top of existing OpenAI buildouts through Microsoft Azure, Oracle Cloud, and CoreWeave. Huang frames the equity investment and the compute contracts as two separate decisions: he believes OpenAI is likely to become the next multi-trillion-dollar hyperscale company, making the equity stake an attractive opportunistic bet independent of whatever chips OpenAI ultimately buys. Much of the early conversation works through why Nvidia's own compute demand keeps compounding: Huang describes three now-simultaneous scaling laws (pre-training, post-training reinforcement learning, and inference-time "thinking") that together are driving inference compute up by what he calls a billion times versus the old one-shot approach, and he walks through a GDP-based model implying the addressable AI infrastructure market could grow from roughly $400B today to as much as $5 trillion a year.

A recurring thread is Huang's defense against skeptics on two fronts: Wall Street's flat 2027-2030 growth consensus, and financial-media concerns about a compute "glut" or "circular revenue" roundtripping between Nvidia, OpenAI, and other partners. On the glut question, Huang argues the industry is still years away from having fully converted general-purpose computing to accelerated computing, so oversupply risk stays low until that conversion completes. On circular revenue, he distinguishes OpenAI's compute buildout (funded by OpenAI's own growing revenue, equity raises, and debt) from Nvidia's separate equity investment, arguing there is no obligation for OpenAI to keep buying Nvidia chips if a competitor's architecture proves better. He extends this into a broader total-cost-of-ownership argument: because Nvidia's performance-per-watt advantage is so large (roughly 30x Hopper-to-Blackwell), a rival chip would have to be given away for free before it beat the revenue a customer could generate from a fixed, power-constrained gigawatt of Nvidia hardware.

The conversation turns to competitive moat and system design. Huang lays out Nvidia's shift to an annual release cadence (Hopper, Blackwell, Rubin, Rubin Ultra, Feynman) enabled by what he calls "extreme codesign," simultaneously optimizing chip, networking, and software across six to seven chips a year, something he says would be impossible without using AI internally to help design Nvidia's own products. He offers a three-tier taxonomy of chip businesses (architectural platforms, ASICs, and customer-owned tooling reserved for Apple-iPhone-scale volume) to argue that most current AI ASIC efforts, Google's TPU program aside, will remain stuck building one component of a much larger, fast-changing AI factory system rather than displacing Nvidia's full-stack platform.

The second half shifts to geopolitics and policy. Huang argues that cutting Nvidia out of China (from roughly 95% market share previously) amounted to unilateral disarmament, letting Huawei build monopoly profits it is now using to fund a stated three-year plan to catch up; he frames continued competition in the Chinese market as good for both US economic interests and for keeping Chinese engineers inside the US-aligned technology ecosystem. Both hosts and Huang treat the broader US-China AI contest as an existential, Manhattan Project-scale race funded by private companies rather than government, citing a sharp reported drop (from roughly 90% to 10-15%) in the share of top Chinese AI PhDs choosing to stay and work in the US as a leading warning indicator worth watching closely. The episode closes on immigration and the "American dream": Huang gives qualified support to the $100,000 H-1B visa fee as "a good start" against abuse while cautioning it should not be the final answer, and both speakers discuss the newly passed "Invest America" program, which gives every US-born child a $1,000 investment account at birth, as a way to spread AI-era wealth broadly. Huang closes by rejecting mass AI unemployment fears, arguing that because AI has made Nvidia itself more productive and led it to hire more people rather than fewer, intelligence is not a zero-sum game.

Notable Quotes

"I think that OpenAI is likely going to be the next multi-trillion dollar hyperscale company." - Jensen Huang

"Nobody needs atomic bombs. Everybody needs AI." - Jensen Huang

"Even if they gave it to you for free, your opportunity cost is so insanely high." - Jensen Huang

"Fundamental to the American dream is the right to rise." - Brad Gerstner

"Just get on it while it's going kind of slowly, and go exponential along the way." - Jensen Huang