The news
The AI IPO pipeline is forcing the economics of frontier AI into the open. In September 2026, details of Anthropic's prospectus surfaced through Reuters, OpenAI ruled out a 2026 listing, and AI infrastructure firm Firmus was reported to have set dates for a multibillion-dollar listing.
Reuters, which reviewed Anthropic's prospectus, reported on September 28 that revenue grew about twelvefold in 2025 to nearly $4.6 billion, while the company posted a $42 billion net loss. About $34 billion of that loss was an accounting charge tied to convertible financing instruments, and excluding writedowns the company lost more than $8 billion on an operating basis. Compute and infrastructure spending tripled to about $7.3 billion, and the filing lists $518 billion of planned compute commitments over coming years, Reuters reported.
Reuters also reported that nearly a quarter of Anthropic's revenue comes from two customers, that many of its largest clients are not locked into long-term contracts, and that the company may seek a valuation above $2 trillion in a listing expected after the November US midterm elections. AFP reported earlier in September that both Anthropic and OpenAI had filed confidential IPO documents.
OpenAI is slowing down. Chief executive Sam Altman told Fortune, in an interview the magazine published on September 12, that "given everything happening with safety, right now would be an ill-advised moment to go public," and when asked about timing he ruled out a listing in 2026. AFP linked the reassessment to recent safety incidents involving AI models.
The infrastructure layer is moving faster. Firmus, a Sydney-founded developer of AI data centers with more than 900 megawatts under contract across seven sites, plans to open a roughly A$7 billion (about US$5 billion) Australian IPO on October 6 and list on October 22, according to a term sheet reported by Reuters and cited by Startup Daily and The Next Web. In China, National Technology News, citing The Information, reported on September 24 that DeepSeek is seeking about $7.5 billion in new funding.
The numbers
- Anthropic 2025 revenue, per Reuters
- Nearly $4.6 billion
- Anthropic 2025 net loss (incl. ~$34 billion accounting charge), per Reuters
- $42 billion
- Anthropic planned compute commitments, per Reuters
- $518 billion
- Share of Anthropic revenue from two customers, per Reuters
- Nearly 25%
- Firmus planned IPO size
- About A$7 billion (US$5 billion)
- DeepSeek funding target, per The Information via National Technology News
- About $7.5 billion
Why CEOs should care
For procurement leaders, a public listing turns a vendor's finances into a document you can read. The reported disclosure that many of Anthropic's largest clients are not locked into long-term contracts, and that two customers account for nearly a quarter of revenue, shows how much a few buyers matter to a frontier lab's growth story. Companies with meaningful AI spend should expect sales teams to push multi-year commitments that make revenue more predictable before a listing, and should trade any such commitment for price protection, service guarantees and exit rights.
CFOs should weigh the gap between revenue and compute spending. Planned compute commitments of $518 billion against 2025 revenue of about $4.6 billion mean frontier labs must grow revenue many times over. That supports aggressive pricing to win volume now, but it also raises the risk of later price increases or product changes if growth disappoints. Keep critical workloads portable across at least two model providers.
Boards should also note that safety is now an explicit factor in listing decisions. Altman tied OpenAI's timing to safety, and public filings will put AI risk factors in front of investors and regulators. Directors at companies building on these models should ask management how a vendor's safety incident, pause or disclosure could affect their own products and commitments.
The bigger picture
The listings under way show two different AI businesses reaching public markets. Model developers such as Anthropic are asking investors to fund heavy current losses for scale later, while infrastructure firms such as Firmus sell contracted capacity to those same developers. Anthropic's reported $7.33 billion of compute spending was more than half of its $12.65 billion in 2025 operating expenses, according to Reuters. Both kinds of company depend on continued demand from a small number of very large buyers, which is why the first public filings will be read closely for customer concentration and the terms of compute contracts.
What’s next
Firmus's bookbuild on October 6 and 7 and its October 22 debut, as set out in a term sheet reported by Reuters, would be an early public test of investor appetite for AI infrastructure this autumn. After that, watch for Anthropic's public filing and pricing after the midterms, whether DeepSeek's round closes, and any change in OpenAI's timeline.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
