Nvidia and AI: A $3 Trillion Bet by 2026

An Ambitious Calculation for the Future of AI

Three years ago, David Cahn, a partner at Sequoia, undertook a bold analysis of the colossal spending by Silicon Valley on artificial intelligence infrastructure. In 2023, he was struck by Nvidia's annual revenue of $50 billion generated from its graphics processing units (GPUs). Extrapolating from this figure, and considering the operating costs of data centers as well as the profit margins of operators, Cahn estimated that a revenue of $200 billion would be necessary to amortize the initial investment.

Cahn viewed this as a stimulating challenge for entrepreneurs, encouraging them to develop AI-based products and services to leverage this massive infrastructure and generate substantial revenues. Today, after three years marked by exponential growth, Cahn has reassessed AI infrastructure spending for 2026, estimating it at $1.5 trillion.

The $3 Trillion Challenge

According to Cahn, the AI industry will need to generate a total of $3 trillion to justify all investments in chips and other data center-related expenses. This figure may even be underestimated, as memory costs rise and the use of specialized chips for inference becomes increasingly common. "Recently," he writes, "the revenue required per gigawatt of CapEx has significantly increased due to these bottlenecks and rising construction costs."

Promising but Insufficient Revenues

On the other hand, some companies are beginning to show signs of promising revenues. Anthropic reportedly reached an annual recurring revenue of $60 billion, while OpenAI generated $13 billion in 2025, although this figure was revised to $20 billion in annual recurring revenue in November of the same year. However, these numbers remain far below the $3 trillion needed, leaving a significant gap to fill.

Expectations from Tech Giants

Torsten Slok, chief economist at Apollo, a major asset manager, is closely monitoring this gap. In a recent note, he highlighted that tech giants like Google, Meta, Microsoft, and Amazon all anticipate a massive increase in their free cash flow by 2028. This means they expect that massive investments in chips will start to pay off.

The Risks of Uncertain Returns on Investment

But what if these expectations are not met? Slok warns of a growing risk in the use of AI: more and more organizations are turning to cheaper open-weight models, often developed in China, rather than those designed by leading labs. Additionally, token prices are generally declining. The latest model from OpenAI, according to its CEO Sam Altman, is 54% more efficient in terms of tokens for coding tasks. This is good news for users concerned about the cost of their AI agents, but it could be detrimental for companies building token factories if users do not significantly increase their overall token consumption.

Slok expresses concern that if hyperscalers do not meet their cash flow targets, the market reaction could be severe. "With so much at stake resting on so few names," he writes, "a slower return on investment would not only be a sector issue but could risk plunging the economy into a recession and the S&P 500 into a correction."

A Necessary Vigilance

This is a crucial aspect to consider as companies and users direct their AI agents toward cheaper solutions. The industry's ability to meet these ambitious financial goals could have significant repercussions not only on the tech sector but also on the global economy as a whole.