OpenAI's audited books remain sealed while another way to spend money on OpenAI has gone public. The company says it submitted a confidential S-1, beginning the IPO review process without publishing the prospectus investors will eventually use to price the company. [1] It also says Oracle Cloud customers will be able to apply OCI Universal Credits toward OpenAI models, including Codex. [2]

The paper's June 16 report on how OpenAI sold credits and rails while its S-1 stayed sealed argued that commercial reliance was outrunning audited disclosure. June 17 turns that into a procurement fact. A buyer with Oracle credits can route spending toward OpenAI before a public investor can read OpenAI's revenue, losses, customer concentration, compute obligations, or margin structure.

That sequence is the point. Fortune's IPO framing asks the obvious market questions that a trillion-dollar listing would have to answer: how large the business is, how costly its compute burden is, and how much concentration sits behind its growth story. [3] OpenAI's Oracle announcement answers a different question, the one enterprise customers care about first. It tells them how to buy.

Confidential filing and public distribution move at different speeds by design. The S-1 process lets OpenAI start the SEC review before showing the market the prospectus. [1] The Oracle route lets buyers act while that review is still private. [2] That creates a strange public order: enterprises can put OpenAI models into procurement workflows before outside investors can compare revenue recognition, compute commitments, customer concentration, or losses against the valuation story Fortune is already asking about. [3]

The word "credits" is doing a lot of work here. Credits are a purchasing path inside Oracle Cloud, not a public answer to OpenAI's unit economics. They can make adoption easier, help procurement officers use existing commitments, and pull Codex and other models into familiar enterprise systems. [2] They do not tell ordinary investors what OpenAI pays for inference, how much Oracle captures, how much usage sticks, or whether gross margins survive scale.

MSM coverage can keep those as separate lanes: one lane for IPO process, one for cloud distribution. X tends to compress them into rails. If OpenAI models can be purchased through Oracle credits, the thinking goes, the monetization path is visible even before the S-1 is public. That is half-right. The rail is visible. The economics of the rail are not.

Oracle's role makes the gap more important, not less. Credits are not cash in OpenAI's audited statement until contracts, usage, costs, recognition rules, and counterparty economics are known. A procurement route can be real and still leave investors unable to test whether the route creates durable margin. That is what the sealed S-1 with public buying rails means: the market is being shown the checkout before it is shown the store's books.

There is also a timing problem for anyone trying to use adoption as valuation proof. Public companies usually ask investors to read the risk factors before buying the growth story. Here, the growth story is traveling through partner announcements, credit systems, enterprise channels, and model availability while the risk factors remain behind the confidential wall. [1][2] That does not make the IPO impossible. It makes the order of evidence unusually lopsided.

Oracle benefits from that lopsided order because it can sell the convenience of the rail without waiting for OpenAI's prospectus to explain the destination. OpenAI benefits because enterprise buyers see availability and procurement fit before they see audited economics. [2] The investor is the one left reading silhouettes: a filing exists, the channel exists, the models exist, but the margin bridge between those facts remains private. [1][3]

The useful rule for readers is therefore simple. Treat the Oracle credit path as evidence that OpenAI is embedding itself inside enterprise purchasing. Do not treat it as evidence that OpenAI's IPO valuation has been proved. Buying access and pricing equity are different acts. On Wednesday, the first act got easier while the second still depended on a document no one outside the confidential process can read. That gap is the story's real risk premium.

-- THEO KAPLAN, San Francisco