TL;DR

  • On June 8, 2026, OpenAI filed a confidential S-1 with the SEC, targeting a September 2026 public listing with Goldman Sachs and Morgan Stanley as underwriters
  • The private valuation sits at $852 billion, with analysts projecting a debut above $1 trillion - one of the five largest IPOs in US history
  • The same week, ChatGPT’s market share fell below 50% for the first time - to 46.4%, with Gemini at 27.7% and Claude at 10.3%
  • OpenAI’s Q1 2026 non-GAAP operating margin was negative 122%: it spends $2.22 for every dollar it earns
  • Noam Shazeer - co-author of Attention Is All You Need and the AI talent Google paid $2.7 billion to retain in 2024 - just left Google to join OpenAI
  • Anthropic filed its own S-1 a week earlier, on June 1, targeting October, at a $965 billion valuation - the two biggest AI labs are racing to Wall Street simultaneously

The timing is almost too perfect to be coincidence - and yet it is. On June 8, 2026, OpenAI submitted a confidential S-1 registration with the SEC, beginning the legal process toward a public listing. The same week, for the first time since ChatGPT launched in November 2022, OpenAI’s flagship product held less than half of the global AI assistant market. The company is going to Wall Street at the precise moment it is no longer the only name in the room.

This doesn’t make the IPO a bad bet. It makes it a fascinating one - and a revealing test of what public markets actually think they are buying when they buy into the AI era.

The Filing

OpenAI’s confidential S-1, filed June 8, is the formal starting gun for a public listing process that has been years in discussion. Goldman Sachs and Morgan Stanley are leading the deal. The roadshow is expected to begin in late August, with a September debut as the working target - pending SEC review and whatever the public filing reveals about the financials.

The valuation picture is straightforward in some ways and murky in others. The most recent private round pegged OpenAI at $852 billion, on the back of a $122 billion raise closed in March 2026. Analysts covering the deal are projecting a first-day market cap above $1 trillion, which would place it alongside the largest IPOs in history. The gap between $852 billion and $1 trillion-plus reflects underwriter optimism that public markets will re-rate the company upward on debut day - a bet that has worked for some tech listings and failed badly for others.

Revenue gives the bull case something to work with. OpenAI hit approximately $25 billion in annualised revenue as of February 2026, up from $20 billion at year-end 2025. That is genuinely impressive growth. The question is what sits underneath it.

The $2.22 Problem

The number that will be at the centre of every analyst note and investor call once the public filing drops is the operating margin. OpenAI’s Q1 2026 non-GAAP operating margin was negative 122%. For every dollar the company earns, it spends $2.22.

That is not a rounding error. It is a structural fact about the economics of running frontier models at scale. The compute costs to train and serve GPT-5, GPT-5.5, and whatever comes next are not declining fast enough to catch up with the revenue line - especially as OpenAI’s own Codex and o-series reasoning products push average token consumption per session into territory that makes flat subscription pricing look economically absurd. (I wrote about the token economics behind this earlier this year, and the math has not improved.)

The bull argument is that this is a scale problem, not a structural one. As revenue doubles, inference costs are supposed to scale sublinearly. The bet public markets are being asked to make is that the negative margin compresses to something defensible before the cash runway runs out - and that $1 trillion in market cap is a reasonable price to pay for that option.

I am a hobbyist watching from the outside, not a financial analyst, so I hold this with the appropriate uncertainty. But the $2.22 figure is the number that will determine whether this IPO looks like Amazon in its loss-making years, or something more concerning.

The Market Share Story

The week of the IPO filing coincided with a milestone that would have been unimaginable three years ago: ChatGPT’s share of the global AI assistant market fell below 50%.

The numbers, from Momentic’s June 2026 AI chatbot market data: ChatGPT at 46.4%, Google Gemini at 27.7%, Claude at 10.3%.

In absolute terms, ChatGPT still has 1.11 billion monthly users - up from 1.05 billion in December 2025. This is not a story about OpenAI losing users. It is a story about the market growing faster than any single provider can capture. Gemini added 129 million users in five months, largely on the back of Android integration - Google has replaced Google Assistant with Gemini at the operating system level across the world’s most widely used mobile platform, which is the kind of structural distribution advantage that does not come from better prompting. Claude went from 60 million monthly users in December 2025 to 245 million by May 2026 - roughly a fourfold surge, partly attributable to a measurable spike in ChatGPT uninstalls in February when OpenAI announced its Department of Defense partnership.

What the 46.4% figure represents is the end of something. ChatGPT was, for most of its existence, a synonym for AI assistant in the same way that Google was a synonym for search. That era is over. The market has fragmented, and it has fragmented permanently - no single provider is going to recapture monopoly share from here.

Whether this matters to the IPO is a question of framing. A company with 1.1 billion monthly users and $25 billion in annualised revenue going public is still an enormous event, regardless of what percentage of the market it represents. But “the ChatGPT era” as a shorthand for undisputed AI dominance is now historical, and public investors who bought into the IPO narrative three years ago were buying a story about monopoly infrastructure. That story requires updating.

The Talent Signal

On June 18 - ten days after the IPO filing - Noam Shazeer announced he was leaving Google to join OpenAI.

The name requires some context if you follow AI by products rather than papers. In 2017, Shazeer was one of the eight co-authors of Attention Is All You Need - the research that introduced the transformer architecture and underpins essentially every major language model built since, including GPT-4, Gemini, and Claude. He subsequently co-founded Character.AI, was bought back by Google in 2024 in a deal valued at $2.7 billion, became VP of Engineering and co-lead of the Gemini models, and has now left for the company whose flagship product runs on his architecture.

This matters in two ways. The obvious one is what it signals about OpenAI’s ambition and resources in the months before a public listing - recruiting at this level, at this cost, is not accident. The subtler one is what it signals about Gemini. Shazeer co-led Gemini to 662 million monthly users and a 27.7% market share gain. He is leaving anyway. What he sees at OpenAI that is worth walking away from that position is, at minimum, a question worth asking.

I want to be careful not to over-index on one personnel move. Talent transitions at this level involve personal, professional, and financial factors that are rarely reducible to a single narrative. But the fact that the person who helped write the foundational paper for modern AI is now at OpenAI - and specifically at OpenAI as it goes public - is the kind of detail that tends to stick.

The Anthropic Race

A detail worth noting: OpenAI was not the first to file. Anthropic submitted a draft S-1 to the SEC on June 1 - one week before OpenAI. It is targeting an October 2026 listing on Nasdaq, with its roadshow expected to begin in August or September.

Anthropic’s current valuation is $965 billion, following its Series H. I covered the context around that figure - and the Bloomberg documentary that went inside the company as it eclipsed OpenAI’s private valuation for the first time - in an earlier post. The short version: Anthropic’s revenue surged 80-fold in Q1 2026, it has a structural partnership with Amazon, and it spent the spring litigating with the Trump administration over safety guardrails in a Pentagon contract.

Two companies valued close to a trillion dollars, both filing for IPO within weeks of each other, both racing toward a public listing window in late 2026. The AI sector has not seen anything like this before. Goldman Sachs projected earlier this year that 2026 IPO proceeds could reach $160 billion - almost entirely on the back of SpaceX, Anthropic, and OpenAI. Whether that holds depends on what investors actually see when the public filings land.

What Public Markets Are Actually Being Asked to Buy

Here is the version of this story that I keep coming back to.

The IPO narrative for both OpenAI and Anthropic is fundamentally a bet on AI becoming infrastructure - on the kind of durable, compounding returns that come from owning the platform layer of a transformational technology. The analogy drawn is usually to Microsoft in the nineties, or AWS in the late 2000s: companies that looked expensive relative to current earnings and cheap relative to what the earnings eventually became.

What makes the OpenAI version of this argument interesting to stress-test is the $2.22 figure. AWS’s path to margin was largely an engineering and scale problem - more customers, same data centres, better software. OpenAI’s path to margin requires either inference costs to fall faster than competition can commoditise revenue, or a differentiated product position that lets it hold pricing even as Gemini and Claude and open-weight models push the price per token toward zero. Those are not the same kind of problem. One is a curve that bends. The other requires winning a market that is actively fragmenting.

None of which means the IPO is a mistake. It may be the right moment to take capital from public markets precisely because the story is still compelling and the competitive landscape has not yet fully resolved. But the version of OpenAI going public that the pre-IPO narrative implies - a monopolist with unassailable reach and improving economics - is not quite the company that filed on June 8.

The version that actually filed has 46.4% market share and falling, $2.22 in costs per dollar earned, and the co-author of Attention Is All You Need just walking in the door. That is still an extraordinary company. It is just a different one.

What I’m Watching

The public S-1. The confidential filing starts the clock. SEC review typically takes a few weeks, after which OpenAI must publish the prospectus at least 15 days before beginning a roadshow. When that document lands, the $2.22 figure will either be contextualised with a path to improvement, or the absence of one will be notable.

The Anthropic filing. October is the target. If it goes first, the comparative valuation math gets interesting very fast - two companies with different margin profiles, different revenue growth rates, and different relationships with the US government, priced against each other in real time.

ChatGPT’s market share trajectory. 46.4% is a floor, not necessarily a collapse. Whether it stabilises at this level or continues to fall as Gemini’s Android integration compounds will matter a great deal to the public listing story. A company entering public markets with declining share is a different pitch than one entering with a floor.

Noam Shazeer’s first public signal from OpenAI. He is a researcher who thinks in architectures. Whatever he is working on will eventually become visible. I am curious what shape it takes.

I’m a hobbyist watching this from the outside, not a financial analyst. Nothing here is investment advice. As always, read the primary sources and make your own call.

Sources