The Anthropic IPO was one of the most tightly controlled listings in recent Wall Street history. The AI company, whose offering was dubbed “Project Lantern” internally, limited some of its 22 underwriters to teams of four bankers, asked senior bankers to sign six-month bans on working with rivals and shared deal documents through a platform that logged every view, Reuters reported on November 22, citing several people close to the process.
The restrictions followed a leak. Nearly four weeks before Anthropic’s S-1 filing with the Securities and Exchange Commission was made public, Reuters reported details from a draft of it, revealing financials, risk factors and dealings with other tech companies. One person familiar with the matter told Reuters the four-banker limit likely came out of fear that information would leak, and that the banks found a way to work around it anyway.
Four-person teams and senior staff only
Anthropic instructed some of its 22 IPO underwriters to limit the number of bankers working on the deal to four, and asked the firms to make sure they were senior staff, several people familiar with the matter said. Goldman Sachs and Morgan Stanley, the two lead underwriters on the deal, declined to comment, as did Anthropic.
Six-month bans on advising rivals
Anthropic asked some senior officers at banks working on its IPO to sign legal documents restricting core members of the transaction team from working with rival companies, including OpenAI, Cursor, DeepSeek and Hugging Face, for six months after it went public, according to the people who spoke to Reuters.
Banks routinely manage conflicts and protect client confidentiality, but bankers and lawyers told Reuters it is rare for an IPO to impose exclusivity this broad, particularly on individual bankers. One banker who declined to sign said it was a challenge to find people who could promise not to cover or engage with Anthropic’s peers, especially as several of those companies pursue their own initial or secondary offerings.

The Anthropic IPO process in numbers. Graphic: GPTOnline, based on Reuters reporting.
A view-logging platform and a 2 a.m. filing
Anthropic used a product called NetRoadshow to share information with most of the banks underwriting the deal, instead of regular PDFs. NetRoadshow logs access details, including who has seen a document and which page they looked at, and documents carried individual bankers’ names as watermarks. Leaks continued to surface in the press, but the platform complicated bankers’ work, including because it would not let them save copies of documents, one of the people said.
The process also ran on a compressed timeline. Anthropic planned to file an updated S-1A with the price range and launch marketing on November 10. The night before, a law firm on the deal sent banks a PDF about the sale of shares in Canada, then followed with an email asking them to disregard it because it “contains incorrect price range and share information,” according to the email, which was seen by Reuters. The correct S-1A was sent through NetRoadshow around 2 a.m. New York time, and banks were told they had several hours to sign off before it was filed with the SEC.

From the leaked draft filing to the first week of trading. Graphic: GPTOnline.
A record raise, then a sharp reversal
The listing still set records. Anthropic priced its shares at $83, selling 980 million Class A shares for $81.3 billion in gross proceeds, the largest IPO on record, and the banks took home more than half a billion dollars in fees, CNN reported. The shares jumped 17% on their first day of trading on November 19, then closed down 8.5% on Friday as a sharp selloff in AI stocks showed that even a blockbuster listing could not shake the gloom hanging over the sector, Reuters reported.
Both Anthropic and SpaceX, whose June IPO also frustrated bankers with tight controls, negotiated to pay underwriters nothing on the additional shares sold through the greenshoe option. “I don’t think most companies are going to be able to demand zero spread on overallotment shares,” Jay Ritter, a University of Florida professor emeritus who researches IPOs, told Reuters.
Why It Matters
The episode points to a shift in power between technology companies and Wall Street. With demand for AI shares running hot, Anthropic and SpaceX could dictate terms that banks would normally refuse, and investors had little choice but to play along. “These companies certainly did have some market power to tell underwriters that it’s not going to be business as usual,” Ritter said. He added that the big Wall Street firms will probably keep doing business as usual for all but maybe OpenAI, after these two deals.
For Anthropic, the controls did not stop the leaks, and the market delivered its own verdict within a day of the debut. The first week of trading showed that a record raise and a 17% pop can sit next to an 8.5% slide, and that the company’s next test is the business itself, not the process that brought it to market.

