Anthropic's S-1: $16.3 Billion in Six Months and the Real Cost of Frontier AI

Anthropic's S-1: $16.3 Billion in Six Months and the Real Cost of Frontier AI

 

 

Article

Anthropic has built its brand on the idea that AI should be built carefully. On October 26 it started selling something more concrete: shares. The Claude maker publicly filed its S-1 with the SEC, and the document is the clearest look yet at what it costs to stay at the frontier of AI.

What happened

Anthropic filed a registration statement for an initial public offering of its Class A common stock. It plans to list on the Nasdaq Global Select Market under the ticker "ANT." The number of shares and the price range haven't been set. The public filing follows a confidential draft submitted in June, and Goldman Sachs and Morgan Stanley are named as contacts for the prospectus.

The numbers

From the consolidated statements in the filing:

Period

Revenue

Compute and infrastructure

First half 2026

$16.29B

$8.95B

First half 2025

$1.16B

$2.33B

Full year 2025

$4.59B

$7.33B

Full year 2024

$386M

$2.52B

That's roughly 14x growth year over year for the half. In every period the filing reports, this is the first where revenue clearly exceeds compute and infrastructure spending.

Then there's the loss. Anthropic reported an $80.7 billion net loss for the first half. According to The Wall Street Journal's reading of the draft, about $72 billion of that came from accounting charges tied to the changing value of convertible notes and other financing instruments, not cash spent running the business. It will still dominate headlines, so readers deserve to know why most of it is a paper loss.

 

 

The part founders should read twice

The S-1 is a map of dependencies. CNBC reports the filing describes compute capacity ordered for delivery between 2028 and 2031, with pricing not yet final, that's expected to add $125 billion to $150 billion in payment obligations. It also says Anthropic will pay SpaceX about $3.1 billion a month for Nvidia-powered compute once all promised capacity arrives.

Reporting on the draft flags concentration too: two unnamed customers accounted for roughly a fifth of first-half revenue.

Why does that matter if you build on Claude, GPT or Gemini? Because your supplier's economics become your economics. A lab carrying hundreds of billions in compute commitments has every reason to chase volume, cut prices for big committed customers and lock in enterprise contracts. Public shareholders will also expect growth every quarter, which can reshape pricing and product priorities in ways a private company could avoid.

What remains uncertain

  • Valuation and raise size are blank in the filing.
  • Third-quarter revenue of $16.8 billion has been reported by Reuters and the Financial Times from an investor meeting. It isn't in the S-1's financial statements and should be labeled as reported.
  • The listing date depends on SEC review and market conditions.

Context

The filing puts Anthropic ahead of OpenAI in the race to go public. It landed the same day SpaceX disclosed a $40 billion GPU lease, and as chip stocks rallied ahead of big tech earnings. The pattern is hard to miss: frontier AI is now as much a capital-markets business as a research one.

What builders should do

  1. Avoid single-vendor dependence. Keep your prompts and evaluation tests portable across at least two model providers.
  2. Negotiate now. Labs racing to grow revenue around an IPO are motivated sellers for committed enterprise spend.
  3. Read the risk factors section of any S-1 from a company you depend on. It's the most candid document a company will ever publish about what could go wrong.

This article is information, not investment advice.

Leave a comment

Please note, comments must be approved before they are published