Brad Gerstner: No AI Bubble, Semis Eat the Nasdaq & AI's Take Off Problem
In a Nutshell
Brad Gerstner argues there's no AI bubble because semiconductors now drive 70% of NASDAQ returns through earnings growth, not multiple expansion, with Nvidia trading at just 14x forward GAAP earnings. The AI capex super-cycle is sustainable only if the top three labs reach $180B in collective revenue by year-end, requiring monthly run rates of $8B versus the current $4B trajectory. Key risks include power infrastructure bottlenecks, potential rate hikes, and regulation, but the knowledge work TAM is massive enough that 4% penetration would justify the buildout.
These notes were generated by AI and may contain inaccuracies.
Brad Gerstner from Altimeter Capital is introduced as an entrepreneur who has started five companies, bringing a different perspective than typical hedge fund managers. His legislative work led to the creation of accounts making every child in America a direct owner in the country, described as the largest unlock of direct philanthropy in U.S. history. Gerstner advocates that the antidote to socialism is more capitalism, emphasizing making every child a capitalist.
Gerstner promotes CAC heart scans as the highest ROI healthcare intervention, costing $100 for a 15-minute procedure. He states this could save 50,000 lives annually if implemented as a standard screening like mammograms. He references the Center for Heart Attack Prevention's work and encourages attendees to get CAC scans.
Markets are up 15% year-to-date and 39% since January of the previous year, despite concerns about tariffs, geopolitics, and AI regulation. Gold is flat and Bitcoin is down 10% year-to-date. Nvidia revenue has doubled, hyperscaler capex has doubled, OpenAI and Anthropic valuations have doubled, and SpaceX is up 2.5x. The market expansion is earnings-driven rather than multiple expansion, with multiples contracting on both NASDAQ and S&P.
Nvidia trades at 14 times next year's fully taxed GAAP earnings. NASDAQ, S&P, SOX, and Nvidia all trade below their average multiples. MAG 7 companies trade roughly in line with historical averages. Consumer discretionary, software, and financials sectors have barely moved, indicating a market driven by the largest capex buildout and super cycle in technology history.
Semiconductors account for 70% of NASDAQ's return. Token makers are capturing value while token buyers participate due to infrastructure market tightness. Dell has risen 5x overall and 9x in 18 months. Hyperscaler capex nearly equals dollar-for-dollar free cash flow to semiconductor companies.
Reference to a podcast with Sam Altman in October where Altman was asked how OpenAI could commit to a trillion dollars in capex with $13 billion in GAAP revenue. Following the launch of Opus 4.5 and Claude Code in December, Anthropic's revenue grew from $2 billion in January to $4 billion in February and $11 billion in March. This revenue growth ignited the April-May market rally. In June-July, consolidation occurred when Anthropic announced a $65 billion annual run rate versus market expectations of $75 billion.
The collective run rate of the top three labs (Anthropic, OpenAI, and xAI) is estimated at approximately $100 billion as of July. These labs need to reach $180 billion collectively by year-end, requiring an additional $80 billion in run rate revenue to sustain the AI trade. The critical data point is whether monthly revenues from Anthropic and OpenAI will reach $4 billion or $8 billion.
A trillion-and-a-half dollars in annual capex requires corresponding offtake revenue. Microsoft, Google, and Amazon are building capacity to rent, not for internal use. Current run rate revenue of approximately $200 billion needs to grow to $450 billion, then $800 billion, or $1 trillion to sustain the capex trajectory. The MAG 5 expected capex must be matched by gap offtake revenue.
The market anticipates adding 19 gigawatts of compute in 2026, with 7 gigawatts going to the two leading labs. For 2027, SemiAnalysis forecasts 43 gigawatts added, with 14 gigawatts allocated to leading labs. This would equal the total cumulative compute currently in the United States. By 2028, over half of total U.S. compute would be controlled by two labs.
The total addressable market for knowledge work (consumer, ads, coding, enterprise workflows) is considered the largest TAM in history. Only 4% penetration ($1.2 trillion) is needed to support the capex. Inference has grown exponentially, with 47 quadrillion tokens projected to be produced. Codex users have grown 40x in eight months. Enterprise knowledge work spending has increased 17x over 18 months.
From 2015-2025, NASDAQ EPS growth averaged 10% (6% revenue growth plus 38 basis points of annual margin expansion). AI could potentially expand margins by 100 basis points annually. Companies like Uber plan 20% growth without headcount increases, and Snowflake plans 30% growth without headcount increases. The primary cost input for companies is human capital, particularly engineers.
Regulation presents a risk of excessive oversight that could stifle development, referencing the historical precedent of shutting down 67 fission reactors. Power infrastructure challenges include permitting delays, local opposition, grid interconnection issues, skilled labor shortages, and sold-out power equipment. Interest rate increases (over 90% probability of rate hikes) raise hurdle rates for borrowed capital used in data center construction.
Current U.S. compute is less than 40 gigawatts. Standing up 43 gigawatts in one year is considered too aggressive. A more realistic estimate is 25 gigawatts, with half allocated to Anthropic and OpenAI. This capacity is sufficient to generate the required revenue targets, as Anthropic reportedly generated $100-110 billion in revenue with 1.5 gigawatts of compute.
Current NASDAQ position is up 15%. Three primary variables determine outcomes: monthly AI lab revenues (targeting $8 billion for takeoff scenario), interest rates and oil prices (5.5% on 10-year would burden equities), and regulation developments around potential Anthropic IPO delays. The 2023-2025 period required only believing AI was the biggest super cycle in technology history. The 2026 environment requires following specific facts and circumstances rather than broad thematic investing.
Position is currently medium-sized with mental flexibility maintained. If AI lab revenues come in strongly and oil prices retreat, position size will increase. If these conditions are not met, position size may decrease further. Emphasis on avoiding leveraged positions and maintaining flexibility based on incoming data rather than making all-or-nothing directional bets.
Keep All-In Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
