The question hanging over every investment committee meeting this week is not whether Anthropic is a remarkable business. It plainly is. The question is what any institution can honestly pay for it given what the amended S-1 actually says.
Third-quarter revenue increased around 47%, a steep drop from 144% growth in Q2 and 132% in Q1. That is not a rounding error. That is the kind of deceleration that changes how investors value a growth asset by an order of magnitude, and it arrived alongside a cost structure moving in the opposite direction. Compute and infrastructure expenses rose faster than revenue, pushing the company to a $567 million operating loss. Anthropic had posted operating income in the prior quarter. It is now loss-making again.
The scale of the forward commitments is harder to dismiss than any single quarterly number. Anthropic said it has entered into purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion, “substantially all of which is related to compute and infrastructure.” What is verifiable in the amended S-1 is that this $895.4 billion figure is the committed spend; the filing does not support a separate claim that additional estimated payments could push the total compute tab beyond $1 trillion. Against nine months of revenue totaling $33.2 billion, that figure is not a moat. It is a decade-plus obligation that has to be financed, serviced, and eventually justified by revenue that does not yet exist at anything like the required scale.
Anthropic’s prospectus discusses why results can swing: capacity constraints when demand exceeded available compute supply, reduced usage around summer vacation periods, and enforcement actions against customers who violated usage policies. That is a plausible explanation. It is also, by construction, impossible to verify before pricing. Institutional buyers are being asked to assign a multiple to a recovery that has not happened yet.
The market has already told Anthropic what it thinks. Reuters reported in October that some current and prospective investors were discussing valuations around $1.5 trillion, down from roughly $2 trillion previously expected. Other investors told bankers they viewed $1.4 trillion to $1.5 trillion as fair value, although a valuation closer to $1.75 trillion could still be possible. Even $1.5 trillion prices the business at roughly 22 times its current annualized run rate. That is a multiple that requires both a Q3 reacceleration and a belief that $895 billion of compute commitments are an asset, not a liability.
The bull case is real and should not be caricatured. Amazon and Google have committed capital to Anthropic at scale precisely because the alternative, ceding the frontier model market entirely, is worse. Nvidia and Microsoft said last year they would invest up to $10 billion and up to $5 billion respectively in Anthropic, with Anthropic committing to buy $30 billion of Azure compute capacity powered by Nvidia. That web of strategic dependencies means the compute bill is partly underwritten by the same companies that need Anthropic to succeed. It is not purely speculative debt.
But here is what the investment committee conversation actually turns on: growth rebounded before, to 132% in Q1 2026 from 48% in Q4 2025, after the Claude Opus 4.5 release in late November 2025. Whether Q3’s ~47% marks a trend or a pause will shape what investors pay. The entire valuation argument collapses into that one empirical question. Reuters reported that Anthropic plans to begin formal investor marketing during the week of November 9. By then, the roadshow will have to answer it.
Stocks to Watch
- Amazon (AMZN): A major outside investor in Anthropic, with its cloud unit AWS serving as a primary compute partner. A successful IPO at premium valuation crystallizes gains on its stake; a stumble at $1.4 trillion changes how investors talk about the AI spending story that Amazon has used to justify its own capital expenditure cycle.
- Alphabet (GOOGL): Anthropic’s IPO valuation shift has direct implications for Amazon’s and Alphabet’s reported earnings, as the companies hold stakes whose paper gains have inflated their financials. A below-target pricing is not just reputational.
- Nvidia (NVDA): The compute commitments that make institutions nervous are, almost entirely, orders for AI infrastructure that runs on Nvidia systems. If Anthropic proves that frontier AI can generate sufficient revenue to service this kind of capital structure, Nvidia’s long-term demand story gets materially stronger. The risk runs the other direction too.
