OpenAI's IPO Delay: What a Listing Below $1 Trillion Would Actually Signal

OpenAI filed confidentially for an initial public offering on 8 June 2026. Within weeks it was reported to be weighing a delay into 2027 rather than list below a trillion dollars. That hesitation is more informative than the filing, because it tells you the company and its bankers do not currently agree with the market about what it is worth.

What has actually been filed, and what has not

A confidential filing is a real S-1 submitted to the regulator without public disclosure. It lets a company begin the review process while keeping its financials private until it decides to go ahead.

OpenAI made that filing on 8 June 2026. Anthropic had filed a week earlier, on 1 June. Both companies were therefore in the queue at the same moment, which matters for how each gets priced.

What has not happened is a public S-1, a roadshow or a price range. Until those appear, every valuation number in circulation is a private mark or a report of one.

The numbers underneath the valuation

Three figures decide how this prices, and they point in different directions.

Revenue is large and has stopped accelerating

The company has confirmed roughly $2 billion per month, which annualises near $25 billion. That is an extraordinary business by any historical standard. It is also broadly where it sat in February.

A flat quarter at that scale is not a crisis. It is a problem for a valuation that assumes continued tripling, and public markets price the second derivative more harshly than private ones.

The losses are structural, not one-off

Internal projections reported to the press point to cumulative losses continuing until roughly 2029 or 2030, at a revenue level far above today's. Analysts including HSBC have publicly questioned whether that timeline holds.

The private mark is very high relative to revenue

An $852 billion post-money valuation against roughly $25 billion of annualised revenue is a multiple in the mid thirties. Software companies have traded there before. They have generally not done so while losing more than they earn.

Why delay at all

A company that has already filed is not delaying because the work is incomplete. It is delaying because it does not like the price.

Listing below the last private round creates a visible down round on the most public stage available. Every employee with equity, every existing investor and every future counterparty sees it. Waiting costs nothing except time, provided the funding position holds.

That last condition is the one to watch. A company with this cost base needs continuous access to capital, and delaying a listing narrows the options for supplying it.

What a confidential filing actually buys you

The mechanism exists so companies can begin regulatory review without committing publicly. It lets a business test the process, receive comments and revise disclosure while keeping competitors and customers in the dark.

It also creates an option. A company that has filed can move quickly if a favourable window opens, and can sit indefinitely if one does not. Filing is closer to loading a weapon than firing it.

The cost is that the option decays. Financial statements in a filing go stale, requiring updated audited figures, and a company that files and then waits several quarters signals something to the market whether it intends to or not.

The employee equity problem

A large share of compensation at companies like this is equity granted at a strike price tied to prior valuations. Listing below the last round does not just embarrass investors. It puts a visible number on grants that staff have been treating as worth considerably more.

Retention risk is a real input into listing timing and it is rarely discussed publicly. A company weighing a delay is weighing this alongside the headline valuation.

Three scenarios, and what each one means for buyers

For a software buyer, only the middle column matters. A listing that forces public scrutiny of unit economics tends to be followed by price increases across the vendor's product line, because public markets reward margin in a way private investors have not.

The mechanism is straightforward. Private investors fund growth and tolerate losses because they are underwriting a terminal value years out. Public shareholders reprice a company every quarter against reported margin. That difference in time horizon changes what management optimises for, usually within two reporting cycles of a listing.

The practical form this takes is rarely a headline price rise. It is tighter usage allowances and more granular tiering. Generous legacy plans get retired quietly. Enterprise agreements renew on less favourable terms. Buyers who only watch list prices miss all of it.

The comparison problem nobody has solved

Anthropic reported an annualised run rate of about $47 billion in late May 2026, ahead of OpenAI's figure. That comparison is not clean, and it is being made constantly anyway.

The two companies report differently. Anthropic reports on a gross basis that counts end-customer spend through cloud resellers as revenue, booking partner payouts as an expense. OpenAI reports closer to net. Comparing the headline numbers directly overstates the gap.

It also understates something else. Roughly 85% of Anthropic's revenue comes from enterprise and developer customers, against a revenue base at OpenAI weighted toward consumer subscriptions. Public markets price those two mixes very differently, and the enterprise mix usually wins on multiple.

Where this reading is weak

Three honest problems with everything above.

The loss figures come from leaked financials rather than published accounts. They have been widely reported and not disputed, which is not the same as being audited in public.

A flat quarter is a single data point. Enterprise revenue is lumpy and a large contract landing in September would reframe the entire growth narrative.

And the strongest counter-argument is simply scale. A company going from nothing to $25 billion in annualised revenue in four years has done something no software company has done before. Judging it on a quarterly derivative may be the wrong frame entirely.

The context most coverage leaves out

Two structural facts sit underneath this listing and rarely appear in the same article as the valuation.

The first is that a large share of the company's costs are paid to a small number of counterparties who are also investors in it. Compute purchased from cloud providers who hold equity positions creates a relationship where the cost line and the capital structure are entangled. That arrangement is mapped in the breakdown of the AI circular deals, and it is one of the reasons a public filing would be genuinely informative rather than merely confirmatory.

The second is competitive. The strongest growth in this category over the past year has been in developer and enterprise products rather than consumer subscriptions. A company whose revenue is weighted toward consumer faces a harder argument about durability, because consumer subscriptions churn and enterprise contracts renew.

Neither of those facts is hidden. Both are simply less exciting than a valuation number, and both will do more to determine the outcome.

There is also a timing consideration that has nothing to do with either company. Listings cluster. When one large technology company prices well, the window opens for others, and when one prices badly it closes for months. With two comparable companies in the queue simultaneously, the first to move effectively sets the terms for the second, which is an argument for going early and an argument for waiting depending on how confident you are.

What to watch instead of the valuation headline

The wider funding picture behind all of this, including how much capital is being deployed by the companies OpenAI buys compute from, is covered in the breakdown of 2026 AI capital expenditure.

One more indicator is worth adding, and it is the least discussed. Watch whether compute purchase commitments are restructured before any listing. A company preparing for public scrutiny has an incentive to renegotiate long-dated obligations. Those would sit awkwardly in a risk factors section. Such renegotiations tend to surface in the counterparty's disclosures first.

None of the four signals above requires access to private information. All of them can be tracked from public reporting, and together they give a far better read on timing than any valuation rumour will.

Frequently asked questions

Where to start this week

One thing to check and one question to ask.

Check what share of your AI spend goes to a single provider. Concentration that felt efficient during a subsidised phase becomes exposure the moment that provider has public shareholders.

Then ask your account manager one question in writing. Has our pricing structure been reviewed ahead of any change in your company's ownership? You will learn as much from how quickly they answer as from what they say.

Related reading