Anthropic's S-1 Raises Big Questions on Growth, Compute and Control

Anthropic booked $16.29 billion in revenue in the first half of 2026 against $2.38 billion in cost of revenue — an implied gross margin of 85% — according to the company's S-1 filing, the most comprehensive public statement to date on how a pure-play AI lab makes its money.

The filing, which runs to more than 300 pages, went live on Monday. Analysts who parsed it found a trove of financial detail alongside several unanswered questions that will shape the IPO and the broader AI market.

The document also disclosed that CEO Dario Amodei and his six co-founders each own more than 200 million shares, a stake that could be worth billions if Anthropic reaches the valuation above $1 trillion that analysts expect. The lab holds more cash than almost every company in the U.S., and as CNBC's Leslie Picker writes, no company has ever gone public with anywhere near that much.

How fast is revenue actually growing?

The headline numbers show explosive growth. Violeta Todorova, senior research analyst at Leverage Shares, told CNBC that first-half 2026 revenue of $16.29 billion compared with just $1.16 billion in the first half of 2025 — a fourteenfold increase year over year. She asked whether the company can grow fast enough to "keep pace with expectations."

"It is unclear how much it costs to generate each additional dollar of revenue as computing demand increases," Todorova said.

The sequential trend is the sharper concern. Anthropic's revenue grew approximately 144% in the second quarter of 2026 compared with the first quarter, then slowed to roughly 47% growth in the third quarter.

"What the prospectus didn't tell us is management's expectations around the sequential slowdown in the growth rate," said Michael Field, chief equity strategist at Morningstar.

The stakes are direct. A company widely expected to price at $1 trillion or more must show investors that deceleration is a function of scale, not of demand fading.

What does the compute bill hide?

The filing states that "substantially all of cost of revenue" is associated with compute and infrastructure costs. It also warns that operating margin and adjusted operating margin are expected to "fluctuate over time, particularly in periods when substantial incremental compute is delivered to us and we incur the associated expenses."

That language leaves a critical split invisible to investors, according to Nick Patience, AI lead at the Futurum Group.

"Investors can see compute falling as a share of revenue, but not how much of the forward commitment is earmarked for training the next generation of models versus serving customers," Patience told CNBC. "That split dictates future margins."

He added that questions remain around whether compute costs can keep falling faster than commitments grow, and whether Anthropic can secure enough capacity to serve its customers.

The prospectus itself acknowledges that supply has been a binding constraint. Anthropic wrote that it had previously sought "quantities of compute from [cloud service providers] in excess of what those partners were willing to provide."

It states plainly: "We have historically operated in a compute-constrained environment in which demand for compute has exceeded the supply available to us."

That admission matters beyond Anthropic. The filing confirms what hyperscalers' capital expenditure figures have implied all year: demand for AI compute across the sector is outrunning supply, and even the best-funded labs cannot simply buy their way out of the bottleneck.

Who actually controls the company?

Anthropic's structure as a public benefit corporation raises governance questions with no clear parallel. Seven co-founders, including Dario Amodei, hold special voting power through a single Class F share equal to 50.1% of total voting power over matters brought to a shareholder vote.

Anthropic describes its capital and corporate governance structure as having "no direct public company precedent" in the filing.

"We also know now that management plans a series of different class share issues," said Morningstar's Field. "But what we don't know is whether investors will be fine with the lack of control implied, or will this impact the share price. Something we won't know until the shares start trading."

The question is not academic. Multi-class structures have drawn investor resistance at other tech listings, and Anthropic's arrangement goes further: minority shareholders will hold essentially no vote on matters the founders choose to decide.

What valuation will Anthropic get?

The trillion-dollar question — the IPO valuation — remains open. Anthropic is staying quiet about the number for now. The filing says only that it will be "determined by negotiations among us, the selling stockholders, and the representatives."

Analyst estimates range between $1 trillion and $2 trillion, which would make the listing a major market event either way. A pricing at the higher end could restore some of the confidence the AI sector has lost in recent weeks. A pricing at the lower end could pressure valuations across the industry.

The wider week in AI: chips, cash and buyouts

The Anthropic filing landed in a week crowded with AI infrastructure news:

• Apple on Wednesday announced its first laptop with a touchscreen, which it calls MacBook Pro with ProLED XDR Display.
• SK Hynix's shares rose after the company reported record quarterly profit that sharply beat estimates, and announced over $25 billion in dividends and buybacks.
• Nvidia and Advanced Micro Devices said they will invest a total of $4 billion in Meta, after the social media giant announced plans to raise $95 billion to support its investments in AI infrastructure.
• Microsoft said Monday that it is offering buyouts to a wide swath of employees as it strives to balance a costly AI infrastructure build-out with capital efficiency.
• SpaceX has closed $40 billion in debt financing to buy graphics processing units, or GPUs, used to power AI systems, Elon Musk said on Sunday.

Taken together, the week's announcements describe an industry still pouring capital into compute at record scale — Meta raising $95 billion, SpaceX borrowing $40 billion for GPUs, Microsoft cutting payroll to fund infrastructure — while the market begins to ask harder questions about returns.

That is the context for the gaps in Anthropic's prospectus. The S-1 answers how much revenue a leading AI lab generates and what it costs to generate it. It does not answer how fast growth continues, how margins hold as compute commitments scale, or how investors will price a governance structure that leaves them without control. Those answers arrive when the shares start trading, and the $1 trillion-to-$2 trillion range suggests the market's verdict will move far more than one stock.