Three years ago, Sequoia partner David Cahn set out to quantify Silicon Valley’s gargantuan AI‑infrastructure spend. In 2023 he reacted to Nvidia’s reported annual GPU revenue of $50 billion, and by adding implied data‑center operating costs and operator margins, he concluded that $200 billion in revenue would be needed to recoup the upfront outlay.

Rising Costs and a New Figure

Fast‑forward three years of hyperscaling, and Cahn now projects AI‑infrastructure spending for 2026 at $1.5 trillion. His calculations suggest the industry must generate a total of $3 trillion in revenue to justify the chips, data‑center hardware, and associated expenses. He warns that rising memory prices and the growing use of exotic or inference‑specific chips could push the number even higher, noting, “the required revenue per GW of CapEx has sharply increased due to these bottleneck dynamics and rising construction costs.”

Revenue Benchmarks from the Front‑Runners

Meanwhile, Anthropic is believed to have crossed $60 billion in annual recurring revenue (ARR), and OpenAI reportedly earned $13 billion in 2025 (it announced $20 billion ARR in November 2025). These figures, while impressive, still leave a sizable gap to the $3 trillion target.

Hyperscalers’ Cash‑Flow Forecasts

According to Apollo’s chief economist Torsten Slok, the four hyperscalers—Google, Meta, Microsoft and Amazon—are all projecting massive accelerations in free‑cash‑flow by 2028, effectively betting on the payoff from their massive chip purchases.

Emerging Risks

Slok highlights a growing risk: more enterprises are turning to cheaper, open‑weight models—often Chinese—rather than proprietary frontier‑lab offerings, driving token prices down. OpenAI’s latest model, according to CEO Sam Altman, is 54 % more token‑efficient on coding tasks, which eases user costs but could hurt firms that monetize token usage if overall demand does not surge.

Macroeconomic Implications

Slok warns that if the hyperscalers miss their cash‑flow goals, the fallout could extend beyond the AI sector. “With so much riding on so few names,” he writes, “a slower payoff wouldn’t just be a sector problem—it would risk tipping the economy into recession and the S&P 500 into a correction.” Investors, therefore, must weigh the cost‑efficiency of AI agents against the broader financial stability of the ecosystem.

In sum, the AI boom’s staggering capital outlays demand not just technical breakthroughs but also robust, revenue‑generating business models. Without that balance, the $3 trillion benchmark may remain an elusive headline rather than a sustainable reality.