ARTIFICIAL INTELLIGENCER-OpenAI versus Anthropic: What the revenue race means for their IPOs

By Kenrick Cai

April 8 (Reuters) - The timing is starting to crystallize for a trio of mega-IPOs, and for those hoping to get some liquidity out of other AI companies going public, there may not be enough investor appetite to go around this year.

SpaceX is set to kick it off, launching its road show as early as June, according to my colleague Echo Wang's exclusive report. OpenAI and Anthropic are expected to follow in the second half of the year.

Together, the three companies could absorb so much investor demand that even well-established names like Canva and Databricks — valued in the tens to hundreds of billions of dollars — find themselves crowded out.

Several analysts and industry experts told Reuters the SpaceX deal alone would likely claim an outsized share of demand.

The broad reopening of the IPO window that many companies have spent years waiting for could then be pushed into 2027, PitchBook analyst Kyle Stanford warned.

Some executives at pre-IPO software companies told me that bankers are already nudging them to ensure their timing doesn't conflict with SpaceX's.

In this week’s issue, we look at the revenue showdown between Anthropic and OpenAI. Plus: why projections on data center buildout costs through 2030 make for precarious math. Scroll on.

Do you think SpaceX’s IPO will affect the timing of OpenAI, Anthropic or smaller AI companies’ plans to go public? Share your thoughts by emailing me or following me on LinkedIn . Forward this newsletter to your friend who would benefit from weekly news and insights on AI. They can also subscribe here .

OUR LATEST REPORTING IN TECH AND AI

• Meta unveils first AI model from costly superintelligence team

• Intel joins Musk's Terafab AI chip project to power humanoid, data center goals

• Anthropic touts AI cybersecurity project with Big Tech partners

• OpenAI urges California, Delaware to investigate Musk's 'anti-competitive behavior’

• Citigroup says AI helps speed account openings and systems upgrades

• SpaceX’s orbital data centers could face same hurdles as Microsoft’s undersea project

OPENAI VERSUS ANTHROPIC

Watch out, OpenAI.

Anthropic reported on Monday that its annualized revenue has surpassed $30 billion.

That means it appears to have eclipsed OpenAI, which disclosed last week, as part of its mammoth fundraising round, that it was generating $2 billion per month — or at least $24 billion annualized.

It would be a remarkable feat of catch-up by Anthropic, the company behind ChatGPT competitor Claude and, more importantly, a suite of AI tools for businesses such as coding agent Claude Code.

At the start of 2025, OpenAI led Anthropic in annualized revenue by $6 billion versus $1 billion, according to self-reported figures from both companies. By early 2026, that gap had widened to $20 billion for OpenAI versus Anthropic’s $9 billion.

But the popularity of Anthropic’s coding agents has soared this year, and the company has built on the momentum with a series of software plug-ins for Claude that wiped out $1 trillion from software and services stocks globally in February. Despite the Pentagon's blacklisting of Anthropic over AI use restrictions, the company’s enterprise success has allowed its business to keep chugging along.

OpenAI’s supporters may dispute the figures. Khosla Ventures partner Ethan Choi, whose firm was an early investor in OpenAI, argued in March that comparing each company’s self-reported figures is “apples to oranges.”

Both companies host their models on third-party platforms such as Amazon Web Services and Microsoft Azure, and pay a cut to those hyperscalers as part of the arrangement. But Anthropic counts revenue on a gross basis, meaning it does not subtract the cut it pays to third-party platforms in the same way that OpenAI does, Choi said. In other words, if OpenAI followed its rival’s accounting principles, its numbers would be higher.

But regardless of the precise numbers, Anthropic’s extraordinary growth rate points to a lesson that OpenAI is now trying to apply. In terms of generating cash, the key metric is not the number of users — where ChatGPT dwarfs Claude — but rather the volume of tokens.

Tokens, the name for a unit of data, correspond to the size of a workload that a user tasks a chatbot to perform. In terms of revenue, it’s starting to show that it is better to have a small number of developers, whose coding tasks are token-intensive, than a large number of users asking a chatbot casual questions, one OpenAI investor told me.

It’s no wonder OpenAI has redrawn its product roadmap to focus on the enterprise, as my colleague Deepa Seetharaman wrote about in last week’s newsletter. The company will be hoping that, in shutting down services like video-generation app Sora and redirecting resources toward coding tool Codex, it can tap into the same stream of revenue from heavy token usage.

As both companies race toward prospective IPOs before the end of the year, their head-to-head battle for enterprise clients will likely have the greatest bearing on what financials they can present during their respective road shows.

CHART OF THE WEEK

At least 110 gigawatts of AI data center capacity is currently in the planning stage through 2030. The funding required makes for “an implausible sum to raise in such a short period of time,” my colleague Jeffrey Goldfarb argues after crunching the numbers in his Breakingviews column this week.

Nvidia CEO Jensen Huang said last year that costs would range from $60 billion to $80 billion per gigawatt. That brings the implied outlay to $6.6 trillion to $8.8 trillion, even assuming that companies announce no further data center construction.

Taking into account the full sum of projected operating cash flows for Alphabet, Amazon, Meta, Microsoft and Oracle, plus estimates of available debt and investments, the available funding totals roughly $7.5 trillion, near the midpoint of Huang’s projected range.

Of course, this estimate makes many assumptions, including that the construction of a 1 GW data center will cost the same over time and that the stock market will remain as patient with the hyperscalers as it has been so far. But the bottom line, Goldfarb argues, is that AI’s ambitions may simply outpace the available funding.

(Reporting by Kenrick Cai; Editing by Lisa Shumaker)

((Kenrick.Cai@thomsonreuters.com; 415-530-6118; Signal: @kenrick.01))