According to Sina Finance, some investors suspect Anthropic favored heavy users of its Claude model in allocating IPO shares, though the company also distributed large allocations to major institutions including Citadel. More than half of the shares reportedly went to existing investors, including Altimeter Capital, Lightspeed Venture Partners and Menlo Ventures. About 60% of institutions that ordered shares received none, while BlackRock reportedly sought about $10 billion worth and received roughly half that amount.

Anthropic shares rose more than 30% above the offering price when trading began Thursday, then closed up 17%. On Friday, they fell as much as 15%, briefly trading just 5 cents above the offer price, before closing down 8.5% at $88.91, still 7% above the offering price. The company’s market value fell by about $109 billion. Nvidia, Meta and Oracle also declined, and the Nasdaq market lost a combined $4 trillion in value. Anthropic’s IPO and concurrent placement raised more than $100 billion; the IPO allocation was valued at $81 billion.

The allocation drew criticism from investors who said they received fewer shares than requested or none at all. One hedge fund manager also questioned the underwriters’ decision on the opening time and price. Some current and former employees were permitted to sell about 300 million shares immediately, worth roughly $26 billion at the offering price, rather than face the usual lockup. Anthropic said Friday morning that the underwriters had fully exercised the IPO’s 15% greenshoe option. About 300 million more shares are due to become tradable on December 15.

The selloff also weighed on plans by rival OpenAI, which plans to go public next year and is pursuing a large funding round at a post-money valuation of $1.25 trillion. Meanwhile, some San Francisco AI workers focused on a claim by an Anthropic mathematician that Claude had solved a 167-year-old Millennium Prize problem.