Humanity can’t afford not to spend $800B on AI, Anthropic insists
Anthropic’s planned stock market listing has brought attention to the scale of its commitments to computing infrastructure. The company says access to ever greater computing capacity is essential to developing its AI models, but the spending creates a major financial risk if revenue does not grow quickly enough.
Its prospectus reports $417 billion in future minimum payments under agreements entered into or amended after 31 December 2025, followed by a further $245 billion in agreements since 30 June 2026, some subject to conditions. The article estimates that listed commitments could total $782 billion to $807 billion, including deals with Microsoft, Google, Amazon, Broadcom and others; some contracts require payment regardless of usage. Anthropic’s first-half 2026 revenue exceeded compute and infrastructure costs by nearly two to one, while its operating loss fell to $693 million from $2.14 billion a year earlier.
In short
- Anthropic’s infrastructure commitments may reach $807 billion.
- The company says expanding compute is vital to model development.
- Revenue is growing, but long-term spending poses a financial risk.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Anthropic’s case is that computing capacity is a prerequisite for developing more capable AI, and securing it in advance may be necessary to compete and meet future demand. If AI produces substantial benefits for science, productivity and public services, large infrastructure commitments could be a worthwhile investment; its reported revenue growth and narrowing operating loss offer some support for expecting the business to scale.
The case against
The commitments could expose Anthropic to very large fixed costs before future demand or revenue is assured, especially where contracts require payment regardless of usage. If AI progress, adoption or pricing falls short of expectations, the spending could weaken the company and divert capital from other priorities; recent revenue growth and a smaller operating loss do not by themselves establish that commitments approaching $800 billion will pay off.
Explained simply
Anthropic is a company that develops artificial intelligence systems, including the Claude chatbot. Training and running these systems requires large amounts of computing power, supplied by specialist data centres and technology firms.
The company has made long-term agreements for computing capacity with providers including Microsoft, Google and Amazon. Some payments may be due even if Anthropic does not use all the capacity, so these commitments could become a financial strain if its income fails to keep pace.
The debate is over whether securing far more computing power is necessary to build better AI systems and compete in the market. It also matters to investors because Anthropic is preparing for a possible stock market listing, where its future costs and ability to earn revenue will be closely examined.