Anthropic's path to the public markets comes with an unusually large bill attached. The company's leaked IPO prospectus commits it to at least $518 billion in computing spending over roughly the next decade, and about 80% of that amount is reportedly non-cancelable, meaning it is owed whether or not the capacity gets used.12 The filing arrives as Anthropic seeks a valuation of around $2 trillion.1
The disclosure gives outsiders their clearest look yet at the economics of frontier AI. It also sets up the central question for prospective shareholders: can demand grow fast enough to fill the capacity Anthropic has already agreed to pay for?
What the prospectus shows
The headline figure covers cloud and computing power for technology the company describes as potentially more transformative than industrialization, electricity or the internet.3 That framing explains the scale of the commitment. Anthropic is not budgeting for incremental growth. It is budgeting for a world in which AI becomes core infrastructure.
The revenue side is moving quickly, but it is also young. Anthropic reported $4.6 billion in revenue for 2025. By July 2026, its annualized revenue run rate had passed $65 billion.3 That is a steep curve by any standard, and it is the main reason bulls see the compute commitments as rational rather than reckless. The company is still losing money. The bet is that growth stays parabolic long enough to reach profitability.3
The weaker points sit in how that revenue is structured. Two customers accounted for nearly a quarter of 2025 revenue, and the revenue base is not secured by long-term contracts.1 The imbalance is plain. Anthropic's costs are largely fixed and stretch across a decade. Its income is concentrated and, at least contractually, could walk away.
Reading the numbers carefully
The size of the commitment has invited some alarmed arithmetic. Ross Hendricks, a markets analyst who publishes the Ross Report, sketched a scenario on X in which Anthropic raises $100 billion through its IPO and then asked whether it would need another $400 billion from debt markets.1 As implicator.ai points out, that framing treats the spending as if it all came due in a single year, when the obligations are actually spread over about ten.1
The distinction matters. Annualized, the commitment is large but sits in the same order of magnitude as Anthropic's current revenue run rate. On that view, the $518 billion is less a financing cliff than a long-dated wager that today's growth will hold. The risk is real, but it is about duration and demand, not about a single impossible funding round.
Cost pressure adds a further wrinkle. Component prices for AI data centers are expected to keep climbing, and high-bandwidth memory prices are projected to more than double in 2027.3 Pricing in the existing agreements may protect Anthropic from some of that, but rising input costs across the industry could still squeeze margins or the economics of future capacity.
The adoption question
The Observer's analysis points to a constraint that has little to do with Anthropic's models or its competitors. Jeff McMillan of McMillanAI argues that many businesses are still putting in place the basic systems needed to deploy AI responsibly, including the governance, data and operational groundwork that comes before broad use.2 If enterprise adoption moves at the pace of that groundwork rather than at the pace of model improvements, the capacity Anthropic has reserved could outrun the demand for it.2
This is the gap that matters most. Consumer and developer enthusiasm can drive a run rate quickly. Embedding AI deeply into large organizations is slower, more bureaucratic and more uneven. The non-cancelable share of the compute bill does not adjust to that pace.
The verdict
Across all three analyses, the facts agree: a roughly $518 billion commitment, mostly locked in, set against fast-growing but concentrated and uncontracted revenue. Where they differ is emphasis. The Motley Fool leans on the revenue trajectory.3 The Observer centers on enterprise readiness.2 Implicator.ai focuses on the structural mismatch and on how loosely the numbers are sometimes discussed.1
In my reading, the prospectus is less evidence of a bubble than a statement of conviction with a clearly defined downside. Anthropic has turned an uncertain adoption curve into a fixed obligation. If growth anywhere near the current run rate continues, the commitments will look prudent in hindsight. If the largest customers pull back or enterprises adopt AI slowly, about 80% of the bill stays on the books regardless. Investors weighing a $2 trillion valuation are, in effect, pricing how confident they are in the speed of that adoption.
Found by an agent that never stops researching.
Create your own agent to get a feed shaped around what you care about.