The tech sector's record-breaking rally continues, yet the foundation supporting this surge is facing intense scrutiny. Altimeter Capital founder Brad Gerstner argued at the All-In Podcast summit that the market's ascent is earnings-driven rather than a speculative bubble, but the conditions needed to sustain the "AI trade" are narrowing rapidly. The monthly revenue figures from frontier labs could very well be the pivotal factor that determines whether the rally breaks out to new highs this year.

Gerstner noted that Anthropic, OpenAI, and SpaceX together generate roughly $100 billion in annualized revenue today, which must climb to at least $180 billion by the end of this year for the AI investment story to hold together. If the massive capital expenditure wave from cloud giants cannot be matched by sufficient rental demand for computing power, this aggressive investment cycle would run out of steam.

At the moment, Gerstner is maintaining an "average" equity exposure, steering clear of high-leverage bets. Should monthly revenue at frontier labs stay near the $8 billion mark, while lower oil prices keep long-end yields in check, he might raise his stake. Conversely, he reserves the right to trim. He cautioned that 2026 will be fundamentally different from the last three years, noting, "Everyone knows about AI now; the expectations are already baked into prices. It's time to follow the facts and stay flexible."

Earnings Growth, Not Bubbles, But the Risk Lies in Concentration

Tech stocks are up 15% this year and 39% since January of last year, yet Gerstner stresses this is a world away from the dot-com crash of 2000. The primary driver today is improving corporate earnings, not multiple expansion. With an anticipated earnings growth rate of around 26% this year, valuations for the Nasdaq and the S&P 500 have actually contracted slightly. For instance, Nvidia (NASDAQ: NVDA) trades at just 14 times next year's fully taxed GAAP profit, and he points out that "the Nasdaq, S&P 500, the SOX index, and Nvidia are all trading significantly below their historical averages."

However, the market's extreme concentration is a red flag. Semiconductors alone account for roughly 70% of the Nasdaq's gains this year, while massive sectors like consumer staples, software, and financials have barely budged. Gerstner calls this "both good news and bad news."

The biggest winners are the chipmakers and infrastructure providers—what he calls "token makers." Cloud capital expenditure translates almost one-for-one into free cash flow for these companies, producing venture-capital-like outcomes. Dell Technologies (NYSE: DELL) has surged fivefold over five years, and SK Hynix has multiplied ninefold in just 18 months.

Frontier Lab Revenue Surge Ignites the Spring Rally

The cornerstone of this spring's historic tech rebound was a single monthly revenue print from Anthropic. Gerstner recalls that Anthropic's monthly revenue run-rate jumped from $2 billion in January to $4 billion in February, then rocketed to $11 billion in March following the release of Claude Opus 4.5 and Claude Code. "The market got an exclamation point on the question of whether AI revenue would materialize," he said. "Not only is it here, but it's huge, and that's what drove the historic gains in April and May."

After that, the market entered a sideways pattern in June and July. Gerstner attributes this to Anthropic's disclosed annualized run-rate of roughly $65 billion—coming in below the $75 billion whisper number—along with concerns about rapid progress from open-source models. Today, the combined annualized revenue for Anthropic, OpenAI, and SpaceX stands near $100 billion. Gerstner believes this figure must rise to between $180 billion and $200 billion by year-end to rationalize current valuations. He defines the monthly revenue data from frontier labs—whether it's $4 billion or $8 billion—as "the single most important data point in the market today."

Who Foots the Bill for Trillion-Dollar Capex?

Gerstner highlighted a core contradiction he sees in the AI thesis, referencing his conversations with OpenAI CEO Sam Altman: How can a company like OpenAI, with around $13 billion in current revenue, commit to absorbing $1 trillion in capital expenditure?

His framework is straightforward. Microsoft, Google, and Amazon are building data centers to lease out, so the genuine question is who will rent all that computing power. "There must be offtake revenue to pay the rent." Even if AI-related revenue hits $200 billion annually by the end of this year, it still falls far short of what is needed. To justify the ongoing capital spending, that number must scale to $450 billion, then $800 billion, and eventually approach $1 trillion.

Still, Gerstner doesn't believe demand is the bottleneck. He sees the total addressable market from knowledge workers, advertising, software code, and enterprise workflows as "the largest potential market in world history." Penetrating just 4% of it—about $1.2 trillion—would be enough to absorb all current AI capital expenditure.

He also cited Nvidia (NASDAQ: NVDA) CEO Jensen Huang's two-year-old prediction that inference demand would grow by a billion-fold, noting it is materializing in the agent era. This year, 47 quadrillion inference tokens are expected to be generated, median enterprise spending on AI has spiked 17-fold in 18 months, and Codex users have grown 40-fold in just eight months.

Margin Expansion Becomes the New Corporate Playbook

AI's permeation into the enterprise is rewriting the speed of margin expansion. Gerstner calculates that between 2015 and 2025, Nasdaq companies grew EPS by about 10% annually, with margin expansion contributing roughly 38 basis points each year. He now believes AI is pushing that contribution above 100 basis points.

He uses Uber Technologies (NYSE: UBER) and Snowflake (NYSE: SNOW) as examples. Uber guides to roughly 20% revenue growth while pledging not to add headcount. Snowflake expects about 30% revenue growth while having a hiring freeze. "For these companies, the biggest line item is employees and engineers. They're not firing everyone; they're just not hiring at the same pace they used to."

Gerstner also expects consumer-grade AI agents to unlock another trillion-dollar market while consuming vast amounts of inference tokens.

Three Key Risks: Regulation, Power, and Rates

Gerstner categorizes the major risks facing AI expansion into three areas. On regulation, he believes policymaking will remain chaotic for some time, but the final solutions won't land at any extreme. He agrees with Elon Musk's proposal for a peer-review mechanism, while warning against repeating past policy mistakes.

On power and compute infrastructure, he is skeptical of overly bullish forecasts. SemiAnalysis analyst Dylan Patel's projection of 43 gigawatts of new global compute next year—roughly equal to America's existing total—seems too aggressive to Gerstner. Due to permitting bottlenecks, grid interconnection queues, equipment shortages, and a lack of skilled workers, he thinks the real number will land closer to 25 gigawatts, with about half going to Anthropic and OpenAI.

He adds that 25 gigawatts is still sufficient to hit next year's revenue targets for frontier labs. Anthropic reportedly generates around $10 to $11 billion in revenue this year using just 1.5 gigawatts, meaning an additional 4 to 5 gigawatts could theoretically support another $100 billion in revenue growth.

On interest rates, Gerstner notes that the Fed's rate hike probability is priced at over 90% for tomorrow, and higher borrowing costs will directly raise the cost of financing data centers. He invokes Warren Buffett's famous line: "Interest rates are to asset prices what gravity is to matter." If the 10-year Treasury yield climbs to 5.5%, equity valuations would face significant downward pressure, given that investors could earn 5.5% to 6% without taking any stock risk.

2026 Strategy: Stick to the Data, Avoid Leverage

Gerstner frames his strategy through a "flight path" metaphor, currently maintaining a "medium" positioning, and flatly rejects high-leverage approaches. He outlines two scenarios: if frontier labs hold monthly revenue at $8 billion while marquee AI companies complete their IPOs, the rally could extend to new heights. However, if the 10-year Treasury yield spikes toward 5.5% due to oil prices, or regulators interfere with the IPO pipeline, the market would face downside pressure.

He specifically warns against the use of 4x leveraged speculative strategies. In a high-volatility, high-rate environment, "using leverage in this market is extremely dangerous."

Regarding the shift in investment logic, Gerstner is direct and concise: From 2023 to 2025, investors merely had to bet on one thing—that AI would be the biggest super-cycle in tech history—and then sit back for the gains. But "2026 is not that kind of environment. Everyone knows AI; those expectations are already priced in. Now it's about making decisions case by case, staying mentally flexible, and following the facts."