Now trailing Anthropic, OpenAI officially files for an IPO

OpenAI's IPO filing and what it actually means

I enjoy watching how companies signal their intentions before they do anything formal. OpenAI filed confidentially, then immediately announced the filing themselves, reasoning that it would leak anyway. That kind of pre-emptive transparency is either very self-aware or very calculated, and with Sam Altman it's usually both.

The valuation at $852 billion post-money, with Anthropic just ahead at $965 billion, means two companies that have never turned a profit are competing to see who goes public at a higher number. That's not a criticism—it's just where the AI investment cycle is right now. OpenAI has raised more than $180 billion in total funding and is still burning cash on compute and infrastructure. The IPO isn't about needing money. It's about formalising a valuation that already exists on paper and giving employees somewhere to land their shares.

The tender offer for employees at the $852 billion valuation is arguably the more immediate story. There's real pressure internally when you have staff holding equity in a company worth nearly a trillion dollars with no liquid market for it. Handling that before or alongside a public offering removes a significant retention risk.

The "third phase" framing

Altman published a blog post the same day describing OpenAI entering what he calls its third phase—from research organisation to product company to infrastructure provider for the AI economy. That framing is unequivocally designed for an investor audience. The language around making AI "abundant, affordable, safe, useful, and easy enough for every person and organisation" reads like an S-1 preamble. Which, now that the filing exists, is exactly what it is.

What's worth noticing is that Altman is trying to position OpenAI less as a product company competing on features and more as a platform the broader economy runs on. Whether investors buy that framing depends on whether OpenAI can demonstrate revenue that looks like infrastructure revenue rather than subscription revenue. Those are very different multiples.

Competition is real and getting harder to dismiss

ChatGPT has 900 million weekly active users according to OpenAI's own figures. That's enormous. But the competitive picture has shifted considerably since that user base was built. Google has embedded AI into search at scale. Anthropic is now valued higher than OpenAI. Meta is giving away capable models for free. The unit economics of AI products are genuinely difficult right now because inference costs are still high and commoditisation of base model capability is moving faster than most people expected two years ago.

OpenAI shutting down fringe projects like Sora's short-form video app and focusing investment into enterprise is probably the right move operationally. Enterprise contracts are stickier, margins are better, and the sales cycle, while slower, produces more predictable revenue. But enterprise is also where Microsoft sits, and Microsoft is both OpenAI's biggest partner and increasingly a competitor building its own AI capabilities on top of the models it helped fund.

The structural complexity nobody is talking about

OpenAI is in the middle of converting from a capped-profit structure to a for-profit public benefit corporation. That restructuring is not complete. Going public before it's finished creates an unusual situation where prospective shareholders will be buying into a company mid-transition, with governance questions that haven't fully resolved. The S-1 will need to explain that clearly, and it's not obvious the explanation will be simple. That's definitely easier to handle while still private, which is why they want optionality on timing.

The timing question also derives from watching how SpaceX's roadshow is received. If the market absorbs SpaceX well, OpenAI and Anthropic will both have cleaner conditions to price into. If it's choppy, expect the "we haven't decided on timing" language to do a lot of work.

This filing confirms the era of AI companies as private clubs for institutional capital is ending. Public market investors will soon have direct exposure to the core companies building foundation models. That means analysts, journalists and the general public will have access to financial disclosures that don't currently exist. Whatever the numbers show, that transparency will be genuinely clarifying for an industry that has operated largely on narrative and valuation rounds rather than audited results.