You are about to build a product on top of an AI model. You have a budget in naira, a bill in dollars, and one question you rarely get to ask: is the company behind this model on solid ground?
This week, one of those companies opened its books. Anthropic, maker of Claude, filed its IPO paperwork with the US Securities and Exchange Commission on 26 October. It is the closest thing the industry has had to an audited look inside a frontier AI lab, and the headline numbers are easy to misread.
The headline numbers
According to the S-1 and TechCrunch's reading of it, Anthropic booked $16.3 billion in revenue in the first half of 2026, up from $1.2 billion in the same period a year earlier. Full-year 2025 revenue was about $4.6 billion, and the company's first commercial year, 2023, brought in $41 million. Anthropic says it will list on the Nasdaq Global Select Market under the ticker ANT, with Goldman Sachs and Morgan Stanley leading the underwriters. The share count and price range are still blank.
The $80.7 billion loss that is not really a loss
The filing also reports a net loss of $80.7 billion for the half-year. Read alone, that looks catastrophic. It is not what it seems. TechCrunch notes the figure is largely driven by non-cash accounting adjustments tied to convertible debt. The debt-market research firm 9fin puts a number on it: $72.4 billion of the loss is a non-cash charge from revaluing convertible notes as the company's valuation climbed. Commons Capital, an investment adviser, makes the same point in plainer words: it is not cash leaving the building.
The operating picture is far less dramatic. Anthropic's operating loss was $693 million in the half, against $2.1 billion a year earlier, per TechCrunch. 9fin reports operating cash flow of $3.9 billion for the period. Investor Tomasz Tunguz goes further, calling the business cash-flow positive. 9fin disagrees on how long that lasts, arguing that positive cash flow would not survive the spending still ahead.
The number that should worry you: compute
The real story sits in the commitments. Anthropic has signed enormous long-term deals for chips, data centres and cloud capacity. But the sources do not agree on the total, and the filing itself leaves the aggregate blank in this draft.
- 9fin counts up to $807 billion in commitments, of which at least $430 billion cannot be exited.
- Tunguz adds seven named agreements and arrives at $637 billion, noting that the filing gives no aggregate.
- Commons Capital notes the obligations stretch as far as 2045.
Treat both totals as analyst estimates, not company disclosures. The point survives either way: 9fin estimates annual compute spend near $100 billion once spread evenly, against a company whose annualised first-half revenue is around $33 billion. Growth has to keep up with the bill.
Who is actually paying
The filing says more than 11,900 customers spend $100,000 or more a year, about 1,900 spend $1 million or more, and roughly 250 spend $10 million or more. Both 9fin and Tunguz report that no single customer exceeds 10 percent of revenue, with the top five near 25 percent. 9fin adds that about 80 percent of revenue is consumption-based, with subscriptions making up the rest. In the filing's own language, customers pay for intelligence by the token.
The credit question nobody expected
Here is where outside voices matter. The Financial Times reported on 16 October that ratings agency KBRA assigned Anthropic an issuer rating of BB+ in June, one notch below investment grade, based on limited company-supplied data. On 30 October, the FT reported that Blackstone is telling investors Anthropic is likely to receive only a double-B rating after its IPO, reversing earlier hopes of investment grade by year-end. A lower rating means pricier borrowing for a company that must keep financing chips. We have not read the FT originals; both reports reach us through Dealroom and AI Weekly respectively.
One more thing in the fine print
Governance is unusual. A Founder LLC of the seven co-founders will hold a single share worth 50.1 percent of the general vote, and a Long-Term Benefit Trust elects a board majority. Public shareholders get economics without control. Press reports value the company at $1.5 to $2 trillion, but that figure comes from the press, not the filing.
What this means for your business in Nigeria
First, expect pricing pressure to matter. Anthropic's model is metered usage in dollars. A company with hundreds of billions in fixed obligations has strong reasons to protect margins. That is our inference, not a stated plan, but budget as if token prices will not keep falling forever.
Second, do not build a single-vendor product. Today's providers are competing hard. Google's Gemini 4 Argon is on introductory pricing, and open-weight models are catching up. Design your stack so you can swap models without a rewrite.
Third, price in the naira. Every usage-based bill converts at the exchange rate of the day. Model your unit economics at a weaker naira, not today's.
Fourth, ask your suppliers. If an agency or vendor builds on one model, ask what happens if prices move or terms change.
So the question for your own roadmap is simple: if your main AI provider doubled its prices tomorrow, would your product still make money?
Sources
- U.S. Securities and Exchange Commission (EDGAR) — Anthropic, PBC — Form S-1 registration statement (opens in a new tab) (26 Oct 2026, primary source)
- Anthropic — Anthropic files S-1 with the SEC for proposed initial public offering (opens in a new tab) (26 Oct 2026)
- TechCrunch — Anthropic officially files for its public debut (opens in a new tab) (26 Oct 2026)
- 9fin — Anthropic S-1 analysis: Revenue gains mask looming $100bn+ annual infrastructure bill (opens in a new tab) (27 Oct 2026)
- Tomasz Tunguz (Theory Ventures) — Anthropic S-1 Analysis: How 7 Key Metrics Stack Up (opens in a new tab) (26 Oct 2026 (updated 2 Nov 2026))
- Commons Capital — Anthropic's IPO Filing, in Plain English (opens in a new tab) (27 Oct 2026)
- Dealroom (reporting the Financial Times) — Anthropic secured highest junk credit rating in June (opens in a new tab) (16 Oct 2026)
- AI Weekly (reporting the Financial Times) — FT: Blackstone Warns Anthropic May Only Clear Double-B Junk Rating After IPO (opens in a new tab) (30 Oct 2026)
- Google — Gemini 4 Carbon: our next frontier model (opens in a new tab) (28 Oct 2026)
Published by Inventrium Magazine
Comments