The listing time of the three major IPOs has gradually become clear, and for those investors who hope to realize funds through the listing of other artificial intelligence companies, the market may not have enough investment demand to undertake all new shares this year. Space Exploration Technology Company (SpaceX) will be the first to launch the listing process, starting a road show as early as June. OpenAI and Anthropic are expected to follow suit and land in the capital market in the second half of the year.
These three companies may attract massive investment demand. Even mature companies such as design platform Canva and data platform Databricks with valuations of tens to hundreds of billions of dollars may face being squeezed out of the market.
Multiple analysts and industry experts told Reuters that SpaceX's IPO alone may account for a large share of investment demand.
Kyle Stanford, an analyst at PitchBook, a private equity market research institution, warned that the IPO market, which many companies have been waiting for for many years, may be delayed until 2027.
Some executives of software companies planning to go public revealed that investment bankers have reminded them to avoid conflicts with SpaceX's listing time.
Will the listing of SpaceX affect the listing plans of OpenAI, Anthropic or small and medium-sized artificial intelligence companies?
OpenAI vs. Anthropic, OpenAI, be careful. Anthropic announced on Monday that its annual revenue has exceeded $30 billion.
This means that its revenue scale has surpassed OpenAI. OpenAI disclosed in a huge financing last week that its monthly revenue reached US$2 billion and its annualized revenue was at least US$24 billion.
For Anthropic, this is a stunning comeback. The company's product Claude is ChatGPT's core competitor. More importantly, it has also launched enterprise-oriented artificial intelligence tools such as code assistant ClaudeCode.
According to public data from both parties, OpenAI’s annual revenue will be US$6 billion in early 2025, ahead of Anthropic’s US$1 billion. By early 2026, this gap had widened to $20 billion for OpenAI and $9 billion for Anthropic.
But since this year, Anthropic's coding products have soared in popularity, and the company has also launched a series of software plug-ins for Claude. In February this year, the impact of related products caused the value of global software and service stocks to evaporate by US$1 trillion. Although the U.S. Department of Defense blacklisted Anthropic over restrictions on the use of artificial intelligence, its enterprise business has continued to grow steadily.
Supporters of OpenAI may dispute this data. Ethan Cui, a partner at venture capital firm Khosla Ventures (whose institution is an early investor in OpenAI), said in March that comparing the revenue data self-reported by the two companies is an "inappropriate comparison of apples and oranges."
He pointed out that both companies deployed their models on third-party platforms such as Amazon Cloud Technology and Microsoft Azure, and paid royalties to these supercomputing service providers. However, Anthropic uses the total revenue caliber and does not deduct fees paid to third-party platforms like OpenAI does. In other words, if OpenAI adopted its rival’s accounting methods, its revenue figures would be higher.
But regardless of the specific value, Anthropic’s astonishing growth rate reveals a key logic, and OpenAI is beginning to follow this idea. In terms of cash income generation, the core indicator is not the number of users (the user scale of ChatGPT far exceeds Claude), but the amount of Token usage.
Token is a unit of data measurement, corresponding to the amount of calculations required by users to assign tasks to the chatbot. An OpenAI investor said that from a revenue perspective, a small number of developer users with high token consumption (such as executing code tasks) are far more than a large number of ordinary users who only chat.
OpenAI has re-planned its product route and fully focused on the enterprise market. The company hopes to cut into the revenue track brought by high token consumption by shutting down services such as the video generation tool Sora and shifting resources to the coding tool Codex.
Both companies are striving to complete their IPOs before the end of the year, and the head-to-head competition for corporate customers will directly determine the financial performance they can show during their road shows.
By 2030, the planned artificial intelligence data center computing power capacity will reach at least 110 gigawatts.
Nvidia CEO Jensen Huang said last year that the cost of building a data center per gigawatt of computing power is between $60 billion and $80 billion. Even if companies no longer plan to add new data centers, the total investment will be as high as 6.6 trillion to 8.8 trillion US dollars.
If the estimated operating cash flows of Alphabet, Amazon, Meta, Microsoft, and Oracle are combined with available debt and investment funds, the total available funds are approximately US$7.5 trillion, which is only in the middle of Huang's estimated range.
Of course, this calculation is based on many assumptions, including that the construction cost of a 1 GW data center remains unchanged for a long time and that the capital market continues to be patient with supercomputing companies. The core conclusion is that the development ambitions in the field of artificial intelligence may have far exceeded the carrying capacity of funds that can be raised.