The Signal #124 — Dakota’s read on the AI news that actually matters to people running a business.

Most startup IPO stories follow a familiar arc. Revenue goes up, the company files, investors pile in. The Anthropic story has that arc on the surface. Underneath, the numbers are worth reading carefully.

What happened

Anthropic filed a registration statement for its initial public offering, with a planned debut in November 2026. On the surface, the financials look strong for a startup. But a piece published in Tech Policy Press by Matt Scherer, co-author of a report called “Brace for Impact: Protecting People and the Economy When the AI Bubble Bursts,” argues the SEC should halt the IPO entirely based on what is buried in the footnotes.

Here is what the footnotes say. Anthropic has $96 billion in cash reserves. It also has at least $582 billion in non-cancelable spending commitments to just five companies: Amazon, AMD, Broadcom, Google, and Microsoft. Those commitments cover access to the computational power required to train and run its models. Add the remaining obligations disclosed elsewhere, and the total reaches approximately $895 billion in contractual commitments, most stretching into the 2030s. Scherer notes that with additional infrastructure contracts still being negotiated, Anthropic may end the year with a trillion dollars or more in total commitments.

These obligations do not appear on Anthropic’s balance sheet as traditional debt. But Scherer argues they function as debts in a practical and legal sense because Anthropic is contractually required to pay them regardless of whether it can generate the revenue to do so.

The revenue picture adds pressure. Anthropic’s 2026 financials have not been formally audited. Third-quarter results suggest revenue growth slowed significantly over the summer. When the company released its Fable 5 frontier model this year, the commercial reception was described as tepid. One study cited in the piece found that the percentage of Americans using AI at work fell from 46 percent in the summer of 2025 to 33 percent in the summer of 2026.

Why it matters

If you are an operator making decisions about which AI tools to build workflows around, vendor financial stability is not a secondary concern. It is a core procurement question.

Think about a legal services firm that has spent three months training staff on a specific AI platform, or an e-commerce company that has integrated a particular model into its customer communication stack. If the underlying model provider hits severe financial stress, the downstream effects are real. Pricing can change. API access can be restructured or discontinued. Support quality drops. A company fighting for survival starts making decisions that are not in your interest as a customer.

None of that means Anthropic is going to fail. It means the financial picture is complicated in ways that matter when you are deciding how deeply to integrate any single vendor into your operations.

There is also a broader signal here about AI industry economics. The piece identifies four margin pressures hitting model developers simultaneously: demand that may not be large enough to justify current spending levels, fierce competition that makes it hard to raise prices, high infrastructure costs that make it equally hard to cut prices, and supply-chain and permitting challenges that delay when revenue can actually be generated. Those four forces affect every major model provider to some degree, not just Anthropic.

What most people get wrong

The common assumption is that if a major AI company has billions in cash and big-name investors, the underlying business is sound. Cash reserves are real. But a company with $96 billion in reserves and $895 billion in spending commitments is not in the same position as a company with $96 billion in reserves and no such obligations. The gap between those two situations is enormous, and it does not show up in the headline number.

Operators also tend to treat AI vendor selection the way they treat picking a software tool: find something that works, deploy it, move on. The actual calculus for AI infrastructure is closer to a supplier relationship in manufacturing. You want to know whether your supplier will be around, whether their pricing will stay predictable, and whether they have incentives aligned with your long-term needs. That requires reading past the press release.

The other thing people get wrong is treating this as a story only about Anthropic. Scherer’s piece frames it as a window into the AI industry’s financial structure broadly. Circular business dealings, non-standard accounting treatments, and spending commitments kept off the main balance sheet are not unique to one company.

The takeaway

Nothing in this post is investment advice, and whether the SEC should block the IPO is a regulatory question above the operator’s pay grade. But the underlying financial mechanics described here are directly relevant to anyone choosing AI tools for their business.

Diversify your AI dependencies where you can. Understand the contractual structure of what you are signing up for. And when a vendor’s numbers look good on the surface, spend five minutes in the footnotes.

If you want help thinking through how to evaluate AI tools and vendors without getting locked into the wrong relationships, xovionlabs.com is a good place to start.