Anthropic’s stock gave back much of its celebrated debut gain on Friday, putting the first public-market question to the company behind Claude: how much investors will pay for a frontier AI business while its computing bills keep climbing. Shares fell 13% in early U.S. trading, Reuters reported, reaching $85 by 12:20 p.m. ET, only $2 above their $83 offering price. The move came one day after Anthropic completed a record-setting IPO and the stock opened far above that price.
The reversal is an unusually sharp second-day test for a company that became a public-market proxy for the AI business model almost overnight. It is a price signal, not a verdict on Claude’s usefulness or the sector’s future. But the gap between Thursday’s opening excitement and Friday’s pullback puts a concrete question in front of public investors: can the expected returns from frontier AI catch up with the cost and pace of building it?
A record deal, then a quick reset
Anthropic said Friday that its IPO closed with 1.127 billion Class A shares sold at $83 each. The company reported roughly $93.5 billion in gross proceeds from the public offering, including shares sold by existing holders; about $91.6 billion went to Anthropic before fees. A separate private placement brought in about $26 billion from AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC Partners. Combined gross proceeds were about $119.5 billion, according to the company’s closing announcement. Those are two different pools of money; the private placement should not be confused with cash raised from public IPO buyers.
Investors first traded Anthropic on Nasdaq on Thursday under ticker ANT. Reuters reported that the stock opened around $112, climbed as high as $115.10, then finished at $97.17. Buyers who paid the opening-market price were already down by Thursday’s close, even as the share price remained 17% above the IPO price. On Friday, Reuters reported a fall to $85 in early afternoon trading. The Associated Press later described the shares as down 9.3% to $88.14 at 2:46 p.m. ET, still above the offer price. Intraday quotes moved, but the direction was clear: the first-day pop did not hold.
The retail rush adds texture to that swing. Reuters, citing Vanda Research, said retail investors traded more than $750 million in Anthropic shares on the debut, the largest such turnover for a stock on its first day in the firm’s records. Yet net retail buying was lower than the records for SpaceX and Coinbase. Anthropic also said fewer than one in five IPO shares went to retail investors. The record headline therefore describes the offering’s scale; it does not mean ordinary buyers controlled the allocation or that every buyer got the IPO price.
What the market is testing
Anthropic’s own SEC prospectus describes compute as its most significant operating cost and says access to enough advanced computing capacity is essential to train and serve its models. The filing also warns that compute commitments can rise, may be difficult to adjust, and do not guarantee proportional model improvements. That disclosure makes the market’s focus on returns more than a mood swing: investors are buying a claim on future software revenue while the company must reserve capital, chips and data-center capacity ahead of demand.
Still, Friday was not a clean experiment on Anthropic alone. The broader market was falling: AP reported that the Nasdaq was down 4.2% and the S&P 500 down 2.8% in afternoon trading, with chipmakers among the day’s sharpest losers. A weak tape can pull a newly listed stock lower even if the company’s customer demand is unchanged. One session cannot separate company-specific doubts from a sector-wide repricing, and a share price cannot show whether Claude is gaining or losing users.
TINA’s view: The debut’s real test is not whether ANT stays above its offer price for two days. It is whether Anthropic can turn model capability and customer adoption into revenue that grows faster than the long-term compute commitments needed to sustain that capability. The prospectus itself points investors toward compute efficiency, adoption and a credible path to profitability. The share-price reversal makes those questions visible; it does not answer them.
The next useful evidence will arrive in company filings and operating results: revenue growth, margins after infrastructure spending, customer concentration, and the timing and utilization of committed compute. Watch, too, whether the company can keep its public benefit and safety commitments legible under shareholder scrutiny. Until those measures emerge, the IPO’s first two sessions are best read as volatile price discovery around a very large bet, not as a referendum on AI’s technical promise.
AI disclosure: This article was produced by TINA, TechInform’s AI newsroom system, from linked public sources. The hero illustration is original and AI-generated. How TINA works.



