North American AI big model giant steps on IPO brake pedal
[North American AI large model giant steps on the IPO brake pedal] In the past six months, Wall Street has been brewing a narrative script of "large model giant IPO, capital market carnival again". Investment bankers generally believe that as the revenue explosion of large model companies is confirmed in early 2026, these companies will take advantage of the trend to promote the IPO process in the second half of the year to seize the best market window period. However, after entering July, this expectation obviously turned cold. Recently, media reports stated that OpenAI CEO Sam Altman rejected the “price reduction to ensure listing” plan proposed by the advisory team and reiterated that the market value of trillions is its bottom line.
In the past six months, Wall Street has been brewing a narrative script of "big model giant IPO, capital market carnival again". Investment bankers generally believe that as the revenue explosion of large model companies is confirmed in early 2026, these companies will take advantage of the trend to promote the IPO process in the second half of the year to seize the best market window period. However, after entering July, this expectation obviously turned cold. Recently, media reports stated that OpenAI CEO Sam Altman rejected the “price reduction to ensure listing” plan proposed by the advisory team and reiterated that the market value of trillions is its bottom line. Prior to this, Anthropic's valuation reached US$965 billion after completing financing at the end of May, which inadvertently brought certain pricing pressure to the powerful Ultraman. In this regard, Wall Street investment banks generally judge that the IPO of large model companies originally scheduled to be launched in the second half of 2026 may be postponed to the first half of 2027 due to the decline in market risk appetite and the uncertainty of the liquidity environment. In addition to pricing pressure, industry events such as key lawsuits and the suspension of data center construction have followed one after another, which seems to imply that it is not the best time for capital market candidate superstars such as OpenAI and Anthropic to test the waters for IPOs. The IPO boom of big models The news that OpenAI postponed its IPO coincided with the completion of the "largest IPO in history" SpaceX's IPO debut. On June 12, SpaceX successfully rang the bell on Nasdaq. The market value on the first day reached US$1.77 trillion. The scale of fundraising set a record for the largest IPO in the history of the US stock market. Retail investors subscribed for more than US$100 billion. However, the revelry only lasted a short time. On June 23, SpaceX fell about 16% in a single day. After rebounding for several trading days, on July 15, SpaceX once fell below the issue price of US$135 per share, closing at US$135.27 per share. SpaceX, OpenAI and Anthropic were once recognized by the market as the top three potential U.S. stock IPOs. Nowadays, the SpaceX IPO has opened high and gone low, setting a not-so-pretty anchor point for the IPOs of OpenAI and Anthropic. OpenAI's anchor valuation is "closer to $700 billion to $800 billion than $1 trillion," Fidelity Securities wrote in a recent IPO valuation report on top AI companies. Judging from more straightforward financing data, the liquidity squeeze caused by the IPO of AI giants may become a constraint for the next stage of IPO, and even the entire index and market. Data shows that SpaceX’s total raised funds (including oversubscription) reached approximately US$85.7 billion. According to current market expectations, excluding the over-subscription potential of OpenAI and Anthropic, the two companies may each raise US$60 billion from the IPO market. This also means that the three companies’ combined fundraising scale of more than US$200 billion will be approximately twice the total amount of funds raised by IPOs in the United States from 2024 to 2025 (approximately US$110 billion). Imbalance between supply and demand of new AI stocks This is not over yet. The second-tier AI large model companies and the leading AI cloud computing power companies are all preparing to go public, and they may continue to increase the IPO market supply. For example, Perplexity, a large first-tier model company, has announced that it will enter the U.S. stock market in 2027. The company completed US$900 million in Series D financing in May 2026, with a post-money valuation of US$14 billion. The "Secure Super Intelligence" founded by former OpenAI chief scientist Ilya has also been exposed as having an IPO plan. The company's valuation in 2025 will reach 32 billion U.S. dollars. Rumors of a new round of financing have raised the target to 60 billion U.S. dollars. The IPO fundraising scale is expected to easily reach more than 10 billion U.S. dollars. Databricks, which focuses on cloud computing, has seen its valuation rise to $62 billion after completing a $10 billion Series J round of financing in December 2025. It has hired Goldman Sachs and Morgan Stanley to prepare for an IPO in the second half of 2026. The IPO financing plans of these companies will be of nuclear bomb scale before 2025. But in front of the "Big Three" of SpaceX, OpenAI, and Anthropic, their total financing is only a fraction of the former's expected issuance size.
In the face of the peak on the supply side, funds on the undertaking side will also be diverted by the capital expenditures of very large cloud providers. Goldman Sachs released a latest report in June this year predicting that the four cloud giants Google, Amazon, Microsoft, and Meta will have a total capital expenditure of US$725 billion in 2026, an increase of 77% from US$410 billion in 2025. These capital expenditures are not entirely supported by the cloud provider's cash flow, but come more from bond financing, bank credit, private debt, etc. And these investment directions share a certain degree of capital pool with U.S. stock IPOs. Under the "dark cloud" of the Federal Reserve's interest rate hike, the contradiction between supply and demand in the IPO market has become even more acute. Bank of America Securities released a report in June predicting that the Federal Reserve will raise interest rates three times in September, October, and December, by 25 basis points each time, for a total of 75 basis points. Goldman Sachs, Morgan Stanley, Deutsche Bank and other institutions have also changed their opinions one after another, turning their interest rate cut expectations into interest rate hike expectations. The high risk-free interest rate will push up the financing interest rates of large cloud providers, which will in turn affect the motivation of funds to subscribe for new IPO shares of large model companies. Amid the craze on the AI track, market participants began to examine the rationality of pricing for large AI models. Previously exposed OpenAI prospectus financial report data showed that the company achieved operating income of US$5.7 billion in the first quarter, but incurred an operating loss of US$9.3 billion and a GAAP loss of US$21.3 billion included in equity incentive expenses. At the same time, although Anthropic claimed to have achieved book profits recently, due to the ban on overseas users from using its Claude model in June, the market generally believed that the book profits it had just achieved would be difficult to sustain. In addition, data center expansion and regional power grid expansion are slowing down, high-bandwidth memory prices are rising - as well as the impact of China's DeepSeek, Zhipu and other companies on API prices. These series of factors will further compress the subsequent gross profit margin of AI model companies. The global technology fund manager of Puxin Group wrote in an internal memo in early July: "We are selective in allocating new AI shares issuance - only companies that see a clear positive free cash flow path to 2028 will consider investing." In addition to operating data, legal risks are capitalization obstacles that large AI model companies have never faced. At the same time as the news of OpenAI’s delayed IPO came out, on July 10, Apple submitted a 41-page complaint in the Northern California Federal Court, taking OpenAI and its hardware joint venture and several former employees who switched from Apple to OpenAI to court, accusing these former employees of systematically stealing trade secrets of unreleased products. In addition to cases, the market has become increasingly worried that under the background of the "natural" talent flow in Silicon Valley in the past three years, employees from major companies such as Apple and Meta have switched jobs to AI startups, which will turn into the risk of trade secret litigation for large model companies before their IPOs. (
In the face of the peak on the supply side, funds on the undertaking side will also be diverted by the capital expenditures of very large cloud providers. Goldman Sachs released a latest report in June this year predicting that the four cloud giants Google, Amazon, Microsoft, and Meta will have a total capital expenditure of US$725 billion in 2026, an increase of 77% from US$410 billion in 2025. These capital expenditures are not entirely supported by the cloud provider's cash flow, but come more from bond financing, bank credit, private debt, etc. And these investment directions share a certain degree of capital pool with U.S. stock IPOs. Under the "dark cloud" of the Federal Reserve's interest rate hike, the contradiction between supply and demand in the IPO market has become even more acute. Bank of America Securities released a report in June predicting that the Federal Reserve will raise interest rates three times in September, October, and December, by 25 basis points each time, for a total of 75 basis points. Goldman Sachs, Morgan Stanley, Deutsche Bank and other institutions have also changed their opinions one after another, turning their interest rate cut expectations into interest rate hike expectations. The high risk-free interest rate will push up the financing interest rates of large cloud providers, which will in turn affect the motivation of funds to subscribe for new IPO shares of large model companies. Amid the craze on the AI track, market participants began to examine the rationality of pricing for large AI models. Previously exposed OpenAI prospectus financial report data showed that the company achieved operating income of US$5.7 billion in the first quarter, but incurred an operating loss of US$9.3 billion and a GAAP loss of US$21.3 billion included in equity incentive expenses. At the same time, although Anthropic claimed to have achieved book profits recently, due to the ban on overseas users from using its Claude model in June, the market generally believed that the book profits it had just achieved would be difficult to sustain. In addition, data center expansion and regional power grid expansion are slowing down, high-bandwidth memory prices are rising - as well as the impact of China's DeepSeek, Zhipu and other companies on API prices. These series of factors will further compress the subsequent gross profit margin of AI model companies. The global technology fund manager of Puxin Group wrote in an internal memo in early July: "We are selective in allocating new AI shares issuance - only companies that see a clear positive free cash flow path to 2028 will consider investing." In addition to operating data, legal risks are capitalization obstacles that large AI model companies have never faced. At the same time as the news of OpenAI’s delayed IPO came out, on July 10, Apple submitted a 41-page complaint in the Northern California Federal Court, taking OpenAI and its hardware joint venture and several former employees who switched from Apple to OpenAI to court, accusing these former employees of systematically stealing trade secrets of unreleased products. In addition to cases, the market has become increasingly worried that under the background of the "natural" talent flow in Silicon Valley in the past three years, employees from major companies such as Apple and Meta have switched jobs to AI startups, which will turn into the risk of trade secret litigation for large model companies before their IPOs. (