Anthropic’s $518B AI Bet: IPO Prospectus Warns of ‘Existential Risks’

Anthropic is preparing to enter the public markets while making a $518 billion infrastructure bet on a technology it also warns could pose an existential threat to humanity.

The Claude maker’s confidential IPO prospectus reveals a company growing at an extraordinary pace while incurring equally extraordinary costs and risks. According to Reuters, Anthropic reported nearly $4.6 billion in 2025 revenue, a $42 billion net loss, and hundreds of billions of dollars in future cloud, computing, and infrastructure commitments.

Then there is the warning label. Anthropic says increasingly capable AI could create “catastrophic or existential risks to humanity,” according to the Financial Times.

Taken together, the disclosures offer one of the clearest pictures yet of the wager behind frontier AI: enormous revenue growth, enormous infrastructure requirements, and risks that extend far beyond a conventional technology IPO.

Anthropic commits $518 billion to the infrastructure behind Claude

The biggest number in the prospectus may be $518 billion.

Anthropic expects to spend at least that much over roughly the next decade through cloud, computing, and infrastructure commitments with six partners, according to Reuters. The figure should not be read simply as a $518 billion capital expenditure plan. It encompasses long-term commitments intended to secure the computing capacity required to train and operate increasingly powerful AI systems.

Those obligations also carry substantial financial weight. Reuters reported that roughly 80% of the commitments are non-cancelable or must be paid regardless of whether Anthropic uses all the capacity.

The scale fits with Anthropic’s increasingly aggressive infrastructure expansion. eWeek previously reported that Anthropic expanded its partnerships with Google and Broadcom to access multiple gigawatts of next-generation computing capacity beginning in 2027.

Anthropic spent $7.33 billion on compute and infrastructure in 2025 alone, according to Reuters, about three times its 2024 spending. Those costs represented more than half of the company’s $12.65 billion in total operating expenses.

The company is effectively making a huge wager that demand for Claude and future AI systems will grow fast enough to justify locking in extraordinary amounts of computing capacity.

A $42 billion loss comes with an important caveat

Anthropic’s 2025 financial results initially look even more dramatic. The company recorded a net loss of nearly $42 billion while generating approximately $4.6 billion in revenue, according to Reuters.

But the headline loss needs context.

Roughly $34 billion of that loss came from an accounting charge tied largely to financing liabilities stemming from earlier fundraising that could eventually convert into Anthropic shares, Reuters reported. It was not $34 billion spent operating Claude or building data centers.

On an operating basis, Anthropic lost more than $8 billion. That is still substantial. But it sits alongside remarkably fast growth: 2025 revenue increased roughly twelvefold from the previous year.

The trajectory appears to have accelerated further in 2026. The Financial Times reported that Anthropic generated $11.5 billion in revenue in the second quarter alone and was heading toward a second consecutive quarter of positive adjusted operating income.

That rapid expansion helps explain why Anthropic has been securing computing capacity at such scale. eWeek previously reported on the company’s multibillion-dollar infrastructure agreement with AMD, another piece of its effort to expand the hardware available for Claude.

Nearly one-third of the filing is about risk

Anthropic does not bury the dangers of its technology in a few boilerplate paragraphs.

Roughly 80 pages of the 261-page main body of its prospectus are devoted to risk factors, according to reports based on the confidential filing. That works out to about 31%. For comparison, Anthropic reportedly uses 48 pages to describe its business.

The disclosures go well beyond the usual warnings about competition, regulation, customer demand, and economic conditions.

Anthropic warns that advanced AI systems could develop capabilities that are difficult to predict before deployment and that models have demonstrated troubling behaviors under experimental conditions. Those include attempts to resist shutdown, conceal or manipulate information, and engage in behavior resembling blackmail, according to reporting on the prospectus.

The company also acknowledges a fundamental difficulty facing AI safety researchers: Models could eventually become capable of recognizing when they are being evaluated, potentially making safety testing less reliable.

That concern fits a broader pattern at Anthropic. The company has repeatedly presented itself as both a frontier AI developer and a laboratory attempting to understand the dangers created by increasingly capable models. eWeek recently reported that Anthropic researcher Jacob Coxon resigned while warning about escalating AI safety risks.

The IPO prospectus brings those concerns directly into the investment case.

Anthropic’s founders want to keep control

Investors could also be buying into a company where conventional shareholder influence is deliberately constrained.

Anthropic is seeking shareholder approval for a structure that would give CEO Dario Amodei and six other co-founders a combined 50.1% of voting power on most corporate matters, according to TechCrunch, citing The Information.

The seven founders reportedly each own roughly 2% of the company, meaning their voting influence under the proposed arrangement would substantially exceed their economic ownership.

The special shares would not give the founders additional economic value. Instead, they would help preserve collective control as Anthropic becomes publicly traded, provided certain ownership requirements continue to be met.

Anthropic’s unusual Long-Term Benefit Trust would also remain part of the governance structure. The trust was created to help steer the company toward its stated public-benefit mission and would continue to play a major role in selecting the company’s board.

That arrangement creates another source of tension within the IPO: Public shareholders could provide capital for Anthropic’s expansion without receiving the degree of control that their economic ownership would ordinarily imply.

The business risks are more conventional, but still substantial

Not every warning in Anthropic’s prospectus involves the fate of humanity.

The company disclosed that two customers generated nearly one-quarter of its revenue in 2025, according to Reuters. Anthropic also warned that many of its largest customers are not locked into long-term contracts and could reduce or stop their spending.

That customer concentration matters when set against at least $518 billion in infrastructure commitments extending roughly a decade into the future, particularly because about 80% of those commitments are reportedly non-cancelable or payable regardless of usage.

Anthropic is securing enormous amounts of computing capacity based on the expectation that demand for AI will continue to climb. If enterprise adoption slows, competitors take market share, or major customers cut spending, those infrastructure obligations could become considerably more difficult to support.

The reverse is also possible. If AI demand continues accelerating at its current pace, Anthropic may need the computing capacity it is racing to secure.

That is the financial bet beneath the bigger philosophical debate.

What eWeek found: Anthropic’s IPO turns the AI paradox into an investment thesis

Anthropic’s prospectus reveals a company built around several tensions that help explain both the scale of the opportunity and the risks investors would be taking on.

These disclosures make Anthropic’s IPO more than a test of how investors value the company behind Claude. They turn some of the AI industry’s biggest unresolved questions into an investment thesis.

Anthropic is betting that enormous long-term infrastructure commitments will be justified by continued AI growth. Investors, meanwhile, would be betting that Anthropic can turn that growth into durable profits while managing the technological risks it describes in unusually stark terms.

That makes the eventual public-market debut a broader test for frontier AI. It could reveal how much financial and governance risk investors are willing to accept for exposure to a technology whose developers themselves acknowledge could have consequences far beyond quarterly earnings.

Also read: Anthropic Files to Go Public, Setting Up Blockbuster AI IPO for more on how the Claude maker began its push toward the public markets.