Search
Money

Anthropic hits the Nasdaq with $200 billion in cash and $900 billion in commitments

The Claude maker's record IPO meets the market as investors weigh surging sales, steep accounting losses and local fights over the data centers it needs.

Anthropic co-founder Jack Clark gives a talk on AI to a full lecture hall at the Schwarzman Centre in Oxford.
File photo. Photo: Mvolz / Wikimedia Commons, CC0

Key Takeaways

  • Anthropic sold 980 million shares at $83, above its $70-$78 range, raising $81.3 billion, the largest IPO on record
  • Quarterly sales hit $17 billion, up more than 1,100% from a year earlier, but losses topped $104 billion through September
  • The company has committed more than $900 billion to vendors over 19 years, mostly for data centers
  • Cofounders hold a majority of shareholder voting power in the public benefit corporation

Anthropic was scheduled to begin trading on the Nasdaq on Thursday under the ticker ANT. The debut gives public investors their first chance to price a company that holds about $200 billion in cash and has promised vendors more than $900 billion over the next 19 years.

The San Francisco company closed the largest initial public offering on record on Wednesday. It sold 980 million Class A shares at $83 apiece to raise $81.3 billion, topping the $75 billion SpaceX collected when it listed in June. Anthropic first marketed the stock at $70 to $78 and raised the price as investor orders piled up. Another $26 billion pledged by AMD, Broadcom, Google, Microsoft and Nvidia brings the total raised to roughly $107 billion.

CEO Dario Amodei and his sister Daniela Amodei went to Nasdaq's New York headquarters Thursday morning to ring the opening bell. At an event marking the listing, Amodei described machine intelligence as "one of the seminal moments in human history." Enthusiasm spilled into the wider market, with the Nasdaq up 460 points, or 1.9%, in early Thursday trading.

The offering values Anthropic at about $1.5 trillion. That is more than Berkshire Hathaway, JPMorgan Chase, Visa or Walmart, though still short of the nearly $1.8 trillion SpaceX fetched. OpenAI, Anthropic's chief rival, has been valued at $1.25 trillion in a fundraising round that is still underway. WIRED noted reports that Anthropic had sought a higher valuation and failed to get it, which the magazine read as a sign of some anxiety on Wall Street.

Those two signals fit together better than they first appear. Lifting the price above the marketed range shows buyers wanted shares. Missing a richer valuation shows they still set limits on what they would pay for a company this young and this unprofitable. That kind of discipline is healthier than a pure frenzy.

Sales are soaring, and so are the bills

The growth numbers explain why demand ran hot. Anthropic's most recent quarterly sales came to $17 billion, more than 12 times the year-earlier figure. The number of customers spending six figures a year on its products has risen twelvefold over the same stretch, and operating losses are shrinking. Tomasz Tunguz, a general partner at Theory Ventures, told WIRED it was hard to imagine a stronger IPO prospectus.

The income statement looks very different. Anthropic lost more than $104 billion in the first nine months of this year, a figure inflated by how the company must account for paying employees in stock. Investors have so far looked past that number, and the record raise suggests they still do.

The figure that deserves more attention is the $900 billion in vendor commitments, most of it tied to building the data centers that run Claude and Anthropic's other models. Spread evenly, that works out to roughly $47 billion a year. The latest quarter, multiplied by four, implies annual sales near $68 billion. On paper, the business already generates enough revenue to cover an average year of those obligations. Some analysts cited by WIRED see the gap the same way, given how fast sales are climbing.

The catch is that the math only works if growth continues. Amodei and OpenAI's Sam Altman have both argued that demand for AI models is surging and will keep investors patient. Holding that demand, though, likely requires each new model to let customers automate work they could not automate before. If the gains flatten, cheaper competitors become a real alternative, and a 19-year spending plan becomes much harder to justify.

The data center fight reaches the IPO

Anthropic's most concrete operational risk may sit far from Wall Street, in the towns where its computing capacity is supposed to rise. WIRED noted that U.S. voters this month largely rejected data centers when given the chance, and local protests against new projects have continued. Delays in opening or expanding facilities could slow Anthropic's work on more powerful models.

In West Virginia, a grassroots group called Mason County United is fighting a data center under construction for Anthropic. At the site, which the developer Nscale is building, heavy rain combined with grading work coincided with flooding in neighbors' backyards. "We're frustrated, we're angry," said Elaine Matheny, a spokesperson for the group. Nscale says it has repaired the damage and bought out some homeowners at above-market prices.

In Mississippi, Yolanda Daniel of the coalition Environmental Advocates Mississippi has vowed to fight an Anthropic data center planned to open next year across the street from her. She argues the company is spreading the costs of AI development across communities while keeping the gains for itself. Environmental activists have also noted that Anthropic has made no commitment to renewable power at its data centers and has never published an environmental impact report.

The opposition has reached the offering itself. Several labor groups, including National Nurses United, urged retirement funds to stay out of the IPO.

Tunguz sees a structural problem rather than a few local disputes. "Clearly permitting a new data center development will be extremely difficult in the US," he said, adding that moving construction abroad would bring more risk, complexity and cost. For a company whose product is effectively computing capacity, that problem goes to the core of the business.

Safety pledges and founder control

Anthropic is the first frontier AI lab to go public. It is also the first company focused solely on generative AI to list, and the largest public benefit corporation to make a market debut. That legal structure lets it weigh societal effects alongside shareholder returns. The company was founded in 2021 by a group that left OpenAI over strategic differences, and the two have been rivals ever since. For a time, OpenAI was expected to reach the public markets first.

Amodei has made safety central to how the company sells itself, and he has acknowledged it carries a cost. At a livestreamed investor meeting last week, he called building safeguards the limiting factor in deploying Anthropic's models, ranking it above access to electricity. "We don't want these agents to run wild," he said. He has said the company held back a new model earlier this year despite the risk to sales, and that it screens its 5,300 employees for a public-benefit orientation, which means passing over some qualified candidates.

That stance has commercial value. Anthropic's clash with the Defense Department this year over limits on how Claude can be used does not appear to have slowed growth much. But public shareholders are now buying into a company that has said openly it will sometimes give up revenue for caution. That is a deliberate trade-off, and it is better disclosed up front than discovered later.

Governance is the less flattering part of the picture. Amodei and his fellow cofounders together hold a majority of shareholder voting power. Employees who donate shares to the philanthropic funds the company offers also hand their voting rights to Amodei. Anthropic's IPO filing argued that part of its public benefit will come as employees and early investors become major philanthropists. Nothing guarantees that outcome, and some investor watchdogs have warned about the concentration of control. Buyers of ANT get exposure to the company's growth, but very little say over how it is run.

A test for the wider AI trade

The listing lands in a nervous stretch for AI stocks. Investors are worried about the debt financing the data center buildout, about the possibility of new regulation on emerging technologies and about Jensen Huang's abrupt departure from Nvidia last week.

Sam Grelck, an equity strategy analyst at Truist Advisory Services, cautioned against reading too much into a single day. "I wouldn't view Anthropic's IPO as a verdict on AI," he told CBS News, though he said the reception will offer another read on investors' appetite for the theme. Matthew Kennedy, a senior market strategist at Renaissance Capital, has forecast that the shares could reach $100 by the end of their first day.

History offers a more sobering reference point. In a Truist review of 32 of the largest IPOs ever, more than half of the companies saw their shares fall by at least 50% at some point in their first year of trading. Anthropic's first day will show how much investors want in. Its first year will show whether it can build the data centers and models needed to justify that price.

Sources

Benjamin Hayes

Benjamin Hayes writes about banking, consumer finance, and the services people use to manage money. His coverage follows account fees, payment systems, borrowing costs, and changes to financial products. He focuses on explaining published terms and industry announcements, giving readers context on how banking developments may affect everyday transactions and access to financial services.

View all stories