Anthropic versus OpenAI: the IPO race has already diverged

For a few weeks it looked like a straight race: Anthropic versus OpenAI, two S-1s filed in June, two private companies nearing a trillion dollars and the same question for Wall Street: who will get there first?

By mid-September, however, the race had already changed shape.
Sam Altman ruled out an OpenAI listing in 2026, citing safety and alignment issues as priorities. That same weekend, Anthropic continued to move towards the public market: Reuters reported that the company had chosen the Nasdaq for its possible debut.

For Anthropic the step comes after months in which financial growth, access to capital and the clash with the Pentagon have become parts of the same dossier. The relevant question is what story the public market will be willing to price and the conditions it will impose on each company.

We analyse both cases now that Anthropic is preparing to go public while OpenAI tries to fund privately the phase that will precede its own listing.

In June they seemed to be starting from the same point

Anthropic confidentially filed its draft S-1 on 1 June 2026. OpenAI did the same on 8 June. Both companies made clear that the filing offered flexibility and that the timing, size and terms of the offering still depended on the market.

That, however, is where the symmetry ends.
In March OpenAI closed a $122 billion round at a post-money valuation of $852 billion. At the end of May Anthropic raised $65 billion at a post-money valuation of $965 billion, overtaking its rival on the value of the latest private round.

A difference to be read with caution: a private valuation incorporates round terms, investor rights and expectations that the public market can recalibrate within hours. It remains, though, a clear signal of how quickly Anthropic has turned commercial growth and financial narrative into fundraising power. Beyond that, however, everything is still to be decided.

Anthropic’s numbers

In May, Anthropic stated that it had exceeded $47 billion in run-rate revenue. Reuters later reported that the annualised figure had passed $65 billion by the end of July, up from around $9 billion at the end of 2025.

The term “run rate” matters because it annualises a recent pace and should be kept distinct from audited twelve-month revenue. It can amplify accelerations that have yet to prove they can hold over time. For an investor, however, the jump remains significant because it measures how quickly Claude, and above all enterprise adoption, are making their way into corporate budgets.

On profitability, the picture is equally delicate. The Financial Times reported that Anthropic expects a second consecutive quarter of positive adjusted operating income. The metric excludes significant items, including stock-based compensation, and should therefore be kept distinct from accounting profit. The useful signal lies elsewhere: the company wants to present itself to the market as a platform that is demonstrating a credible economic trajectory ahead of its listing.

OpenAI and a different kind of scale

Reducing OpenAI to its capital requirements would be a misreading.
The company comes to the contest with ChatGPT’s consumer distribution, a far broader developer ecosystem, deep infrastructure relationships and an ability to turn a general-purpose product into a platform.

The March round also shows how much capital the private market is still willing to put on the table. Choosing to wait, however, does not mean OpenAI has stopped laying the financial groundwork for a listing. Quite the opposite.

On 15 September Reuters, citing the Financial Times, reported that the company had held preliminary talks with major investors about a new private round that could lift its valuation to around $1.2 trillion. The discussions are said to be at an early stage and to have been initiated by the investors, while OpenAI has not commented. Three days later another figure emerged: according to the Financial Times, OpenAI expects around $278 billion in cumulative negative free cash flow between 2026 and 2030, as it continues to increase investment in compute and infrastructure.

For OpenAI, staying private for longer therefore takes on a very clear meaning: preserving freedom to invest while it funds an industrial scale that requires amounts of capital hard to compare with those of an ordinary software company.

In short, Anthropic may be the first to be judged on public-company numbers; OpenAI, meanwhile, is buying time to consolidate infrastructure, revenue and governance without the immediate constraint of the market.

The White House is now priced into the risk

The financial comparison is intertwined with Washington in an unusual way. According to the Financial Times, OpenAI has discussed the possibility of granting the federal government a 5% stake as part of a broader proposal on public participation in the value created by AI. The proposal is still preliminary and does not set out an already approved corporate structure. The gap between political discussion and actual ownership structure remains substantial.

Anthropic has experienced the opposite relationship. Its refusal to loosen certain safeguards on the use of its models for autonomous weapons and domestic surveillance led to a clash with the Department of War and to its designation as a supply chain risk. On 27 August a federal court then ruled in Anthropic’s favour on significant constitutional and administrative grounds, including the First Amendment, due process and several claims under the Administrative Procedure Act.

For investors, the lesson is less simple than “political proximity is good” or “political independence is good”. Public contracts, regulation, procurement, national security and ownership structure have become factors capable of affecting value. In a company that could be worth more than a trillion, governance stops being a footnote.

Anthropic is ahead (for now)

The calendar keeps shifting. And time keeps passing.
After indications of a possible roadshow launch in October, the Wall Street Journal reported that Anthropic is now leaning towards a November listing, partly so it can show investors its third-quarter results. Reuters, meanwhile, reports that the company could hold off until after the midterms.

The direction is unchanged: barring further delays, Anthropic should go public before OpenAI.

This gives it a narrative advantage and an operational risk. Whoever arrives first sets the multiples, the questions and the benchmarks by which the sector will be read. It also means being the first to discover how far investors are willing to tolerate capex, dependence on cloud providers, training costs, regulatory pressure and revenue growing at rates that are hard to model.

Who will have the better IPO?

Today the answer is: it will depend on what the market decides to buy.
Obvious, we know, but genuinely true.

Anthropic offers extraordinary growth, a higher private valuation and a financial trajectory that is becoming easier to read.
OpenAI offers scale, distribution, brand and an ecosystem that no other lab possesses in the same form.

The listing will turn these differences into very concrete questions. How defensible are the margins? How much of the growth depends on the cost of compute? How concentrated is the customer base? What is the relationship between safety, product speed and legal risk? How much more capital will be needed? And what governance will be acceptable when technical decisions with geopolitical consequences have to coexist with public shareholders?

The contest between Anthropic and OpenAI looked like a race to ring the Wall Street bell. In September it became something more useful to watch: two different ways of arriving at the same problem. Anthropic is trying to prove it is ready for the market. OpenAI has chosen to prove it can afford to keep the market waiting.