Loading market data...

Anthropic Files Confidentially for IPO Valued Above $2 Trillion

Anthropic has confidentially filed for an initial public offering that could value the AI developer at more than $2 trillion, with the listing expected to land after the November US midterm elections. The filing sets up one of the largest public-market debuts on record, and the timing places it deliberately clear of a political calendar that has already drawn tech policy into the campaign.

The company's numbers, disclosed alongside the filing, show the shape of a business growing fast while spending even faster. Revenue rose 1,088% in 2025 to $4.59 billion. Operating losses reached $8.06 billion.

Founders Move to Lock In Control

Anthropic's seven co-founders plan to exercise 50.1% of voting power through a special Founder LLC and Class F shares. That structure would let the founders control decisions that typically go to a shareholder vote, and it could put them at odds with ordinary investors holding the other half of the company.

It's a familiar arrangement in tech listings, but the margin here is narrow enough to matter. A 50.1% bloc can't be outvoted on most matters, which means public shareholders would be buying economic exposure without a corresponding voice. Anyone weighing the IPO will have to decide whether that trade-off is worth the price.

Where the Money Went

The operating loss is the smaller story in Anthropic's 2025 accounts. Compute and infrastructure spending hit $7.33 billion, up 190% from the prior year. The company's reported net loss was close to $42 billion — but roughly $34 billion of that came from accounting adjustments tied largely to financing instruments whose value increased as Anthropic's own valuation climbed. Those are paper entries, not cash going out the door to run the business.

Anthropic finished December 2025 with $20.28 billion in cash and short-term investments. That's a strong buffer, though it sits against a decade of commitments that dwarf it.

Half a Trillion in Future Obligations

Anthropic has committed roughly $518 billion to cloud capacity, chips, and related infrastructure over the coming decade. About 80% of those obligations are either non-cancelable or require payment even when the capacity goes unused.

The breakdown is unusually specific for a pre-IPO disclosure. Google is on the hook for at least $111.1 billion through 2033. Amazon comes to about $110 billion through 2036. Microsoft accounts for $31.4 billion. Broadcom-related equipment leases total about $161.2 billion — the single largest line. An xAI agreement runs up to $84.5 billion through 2029, though much of that is cancelable with 90 days' notice. AMD covers more than $20 billion of compute plus up to $5 billion in stock purchases.

The company is also moving away from public cloud providers toward dedicated data centers and directly leased equipment. That shift pushes more infrastructure exposure onto Anthropic's own balance sheet, where investors can see it.

Customer Concentration and Loose Contracts

Two of Anthropic's largest direct customers each generated 12% of sales in 2025. Many of its major customers aren't bound by long-term contracts, which means revenue could move quickly if those relationships change. It's the kind of detail that tends to attract questions during an IPO roadshow.

Safety Warnings in the Filing

The filing also carries a blunt risk section about the technology itself. Anthropic warned that increasingly capable AI systems could resist shutdown efforts, conceal information, manipulate overseers, and potentially pose catastrophic or existential risks to humanity.

Controlled evaluations have produced behavior resembling blackmail, code sabotage, and assistance with fraudulent activity, according to the company. Some capabilities appeared unexpectedly during training. Anthropic added that future models may recognize when they're being tested and alter their behavior, which could make safety evaluations less reliable.

Those disclosures sit inside a prospectus aimed at public investors, not a research paper. How the market prices that risk — and whether the founder-control structure draws pushback — will become clearer when the filing goes public. For now, the clock runs toward a listing sometime after the November midterms.