Model

Anthropic shows CFOs what nine-fold spend buys, two days after its S-1 revealed a first $974 million profit

Two days after its IPO filing disclosed a first quarterly operating profit and net revenue retention near 900%, Anthropic shipped reporting tools that show enterprise customers what their token spend returns. The tools land days before pricing in a listing that could reportedly value the lab at up to $2 trillion.

Vincent JiangVincent Jiang · 3 min read
Share
Dario Amodei, Anthropic's co-founder and chief executive, at a 2023 meeting at 10 Downing Street in London
1 / 6Slide 1 of 6
Dario Amodei, Anthropic's co-founder and chief executive. Two days after its S-1 filing went public, his company shipped reports that show customers what their token spend returns.

Anthropic$350B — Anthropic, private, latest valuation $350B launched a reporting suite on Wednesday October 28, 2026, that shows enterprise customers what their token budget buys: a team's output over any date range, what it cost, where the workflow broke, plus pooled group budgets to cap spending 1.

The timing is the argument. The S-1, the registration filing for its Nasdaq listing, went public two days earlier, and The Inference sized up its 93 pages of risk factors earlier today 5. What is new inside it is unit economics: first-half revenue of $16.3 billion, a first operating profit of $974 million in the second quarter, and net revenue retention near 900%, meaning the same customers spent nine times more within a year 2.

Customers paying at least $100,000 a year grew from about 950 to 11,900 2. TechCrunch reads the investor financials as Claude Code's work, the coding agent built on Claude, Anthropic's model family 1.

The reports answer the CFO, and deepen the lock-in

Return on investment is the loudest question in enterprise AI: most of the Fortune 500 still sees no major gain from deploying it 1. Jellyfish, an engineering-analytics firm, studied 276,000 customers and found spend on Anthropic-based coding tools up 10x since January, now about 2% of total engineering cost, "growth that most CFOs didn't plan for," chief scientist Nicholas Arcolano says 1.

Anthropic is separately spending $100 million to train 10,000 frontier-deployed engineers, AI consultants embedded inside customer organizations, by the end of 2027 3. TechCrunch calls the new console reports "a kind of FDE-lite," that consulting motion rendered in software 1. The reports answer the CFO's question, and they make the same spend harder to leave: "once you're in, it's going to be harder and harder to switch," says Jellyfish AI product lead Nikolas Albarran 1.

A third of the code, and no reason to leave

Jellyfish's data puts Anthropic-based tools behind nearly a third of proposed code changes 1, and the firm's homepage banner says AI spend per engineer tripled in six months 4. Ramp, an expense-software firm, counted 43.8% of US businesses paying Anthropic in September against 40.6% for OpenAI$500B — OpenAI, private, latest valuation $500B 5. OpenAI's Codex coding app and its summer Astra model have not slowed adoption: "there's not a big enough reason to switch," Albarran says 1.

What the subscribers actually buy

The filing warns that 5% to 7% of revenue in the nine months to September 30 came from customers who broke its usage policy or terms 2. Control stays inside: a founders' entity holding a single Class F share keeps 50.1% of the vote, and the Long-Term Benefit Trust elects the majority of the board 6.

Alex Morris, chief executive of the investment firm F/m Investments, expects the offering "substantially oversubscribed" 7. He still asks what happens when seven people can appoint the board of "what could be one of the top five, if not the largest company on the planet" 7. Chipmakers funding their own customers he calls a financing risk "probably more for Nvidia−1.97% — Nvidia, down 1.97 percent today"; Nvidia is putting $3 billion into Meta's $35 billion stock sale, with $60 billion more planned 78.

Subscribers also inherit compute already committed, a total the filing leaves blank 2.

Anthropic's committed compute already runs past $430 billion

$0B$50B$100B$150B$200BBroadcom equipment leases$161.2BGoogle$114.5BAmazon$110BMicrosoft$44.9B
Data
Value
Broadcom equipment leases$161.2B
Google$114.5B
Amazon$110B
Microsoft$44.9B
Stated long-term cloud commitments to Google, Amazon and Microsoft (contract minimums) plus Broadcom equipment leases, as listed in the S-1 filed October 26, 2026; the filing leaves the combined total blank. Billions of US dollars.2

Pricing could come as soon as November 16 5. OpenAI, which delayed its own listing to next year 5, remains the flip risk. The reading that matters: whether 900% retention holds once CFOs can read the receipts 5.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Loading NVDA

Reader comments

0 comments

    Model

    Anthropic prices Claude for financial advisors at $4,800 per 20 seats; FINNY charges $50 a month

    A day after Anthropic filed to list on Nasdaq as ANT, its Claude for Financial Advisors product picked up six connectors and its first public price: $4,800 for 20 seats. Startup FINNY went the other way, charging $50 a month plus an undisclosed share of the assets it helps advisors win.

    Model

    Google DeepMind's Genie 4 ships unpriced, and the last launch cost Unity 24% in a day

    DeepMind's new world model fixes the gaps Unity's CEO cited in January, yet ships as an unpriced research preview, so the milestone that would actually reprice the game engine trade is still ahead.

    Model

    Universal and Sony matched 60,202 songs in Suno's training files, at up to $150,000 each

    Reservoir, the Presley estate and Universal and Sony are filing priced song lists at $150,000 per work, backed by Suno's own admissions. Licensing, the road Udio took, now reads as cheap insurance.

    Model

    OpenAI's new model runs tasks about nine times cheaper than Claude Opus 5.5, the model in Anthropic's S-1

    OpenAI published Terra's price sheet the same day Anthropic's S-1 went public: frontier capability at a tenth of the flagship cost, against a prospectus priced for a $1.5 trillion valuation.

    Model

    Mistral's €3 billion round funds a pivot from open models to selling compute, with Nvidia on both sides

    Six days, three disclosures: a gigawatt target, a promised sixfold capacity increase, and weights that missed their own deadline. Mistral is building an infrastructure business, and Nvidia funds both sides of it.

    Model

    Forbes says Edwin Chen is worth $18 billion. No one has ever priced Surge AI.

    Forbes' 2026 list values Edwin Chen's 75% of Surge AI at $18 billion, implying the bootstrapped data-labeler is worth about $24 billion. Rival marks at Scale AI and Mercor were set by cash; Surge's traces to funding talks that never closed.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.