OpenAI’s IPO Filing Starts a Compute Cash Race With Anthropic

OpenAI’s initial public offering (IPO) filing has put the ChatGPT maker into Wall Street’s queue without forcing it to name a date. The company said on June 8, 2026 that it confidentially submitted a draft S-1 registration statement to the U.S. Securities and Exchange Commission (SEC), leaving share count, price and timing unsettled.

The move follows Anthropic’s June 1 confidential filing and pulls the two artificial intelligence (AI) labs into the same public review track. The race now runs through cash needs: chips, data centers, power contracts and the accounting detail that private investors have accepted behind closed doors.

The Private Review Starts Now

A confidential S-1 starts SEC staff review before the full prospectus is visible to investors. Under SEC draft registration guidance, an issuer conducting an IPO must publicly file its registration statement and earlier nonpublic drafts at least 15 days before a road show, or before effectiveness when there is no road show.

We recently submitted a confidential S-1. We expect it to leak so we’re just announcing it.

That was the company’s public rationale on Monday. The same notice said timing has not been decided and that some work may be easier while the business remains private. It was published under Rule 135 of the Securities Act of 1933, the safe-harbor notice used before a registered securities offering.

The quiet review still changes the rhythm. Bankers, auditors and lawyers can start resolving SEC comments while the public sees none of the revenue tables, margin history, customer concentration, legal risks or governance mechanics that will shape the deal.

Anthropic Set the Near-Term Clock

Anthropic, maker of the Claude chatbot, submitted its own draft S-1 on June 1. Its announcement said the number of shares and price had not been set, and that any offering would depend on market conditions and other factors.

One week is a small gap in IPO processing. It is a large one in a market where valuation and growth stories can move with every enterprise contract, model launch and capacity deal.

The Cash Stack Is Already Huge

OpenAI’s March fundraising release reads like a preview of the questions public investors will ask. In the March 31 funding announcement, the company said it had closed its latest round and expanded a revolving credit facility while naming a long list of cloud, chip and data center partners.

The investor roster included Amazon, NVIDIA, SoftBank and Microsoft alongside funds such as a16z, TPG, T. Rowe Price, BlackRock affiliates, Fidelity, Sequoia Capital, Temasek and Thrive Capital. The same announcement said more than $3 billion came through bank channels for individual investors and that some shares would be included in ARK Invest exchange-traded funds.

Compute Sets the Capital Requirement

The largest numbers are attached to the physical buildout. In September, OpenAI, Oracle and SoftBank said five new U.S. data center sites put Stargate at nearly 7 gigawatts of planned capacity and over $400 billion of investment over three years. The same Stargate data center announcement said the program was on a path to the full $500 billion, 10 gigawatts commitment.

The sites named or described in that update included Shackelford County, Texas; Dona Ana County, New Mexico; Lordstown, Ohio; Milam County, Texas; a Midwest site later updated as Wisconsin; and additional potential expansion near Abilene, Texas. Oracle’s separate agreement covered up to 4.5 gigawatts of additional Stargate capacity, with the companies describing it as a partnership exceeding $300 billion over five years.

Anthropic’s own funding release puts the same pressure on a smaller corporate history. Its May 28 Series H announcement said the funding would expand compute for Claude and named Micron, Samsung and SK hynix as strategic infrastructure partners tied to memory, storage and logic chips.

What the S-1 Will Have to Show

The confidential filing keeps the financial model hidden for now. When the draft becomes public, the first read will be less about chatbot popularity and more about how much cash each dollar of AI revenue consumes.

• Revenue mix – the split among ChatGPT subscriptions, enterprise contracts, application programming interface (API, the software channel developers use to call AI models), coding tools and any advertising pilots.
• Gross margin – the cost of serving prompts, running inference and reserving cloud capacity against each line of revenue.
• Capital commitments – lease obligations, purchase commitments, data center contracts and chip supply arrangements that sit beyond ordinary operating costs.
• Governance – voting control, related-party arrangements, nonprofit-linked constraints and the board structure public shareholders would inherit.
• Customer concentration – how much enterprise revenue comes from the largest buyers and whether any reseller or cloud partner drives a major share of sales.

The S-1 also gives competitors a map. Pricing, retention, customer mix and risk language will travel through every boardroom trying to decide whether to buy AI capacity, back a rival lab or wait for cheaper models.

CoreWeave Gave Investors a Live Rehearsal

CoreWeave’s market debut gives the AI trade a public comp, even though its business is narrower than a frontier model lab. The company sells cloud capacity built around graphics processing units (GPUs, chips used for parallel AI computation), and it began trading on Nasdaq after pricing 37.5 million shares at $40 each.

That listing gave investors audited documents, public debt discussion and a daily share price attached to AI infrastructure demand. OpenAI’s March release named CoreWeave among its cloud providers, placing the newly public supplier inside the same chain of companies trying to finance capacity before usage patterns settle.

A lab IPO would bring a different kind of document to market. It would pair consumer adoption, enterprise contracts and model development with the infrastructure obligations that sit behind each query. The result will be a prospectus investors can compare against software companies, cloud providers and semiconductor customers, with no perfect match in any group.

Public Markets Will Price the Delay

The gap between a confidential submission and a priced IPO can run for months. The SEC review has to move, markets have to hold, and the issuer has to accept the disclosure burden that comes with quarterly reports, public risk factors and investor calls.

OpenAI has already shown public-market access through bank channels, ARK Invest funds and a credit line supported by major banks. Anthropic has already shown a higher last-round valuation and a disclosed revenue run rate above its main rival’s monthly figure. The first public S-1 to reach EDGAR will set the working benchmark for the other.

The next price tag comes when one of the drafts finally reaches EDGAR.

Disclaimer: This article is for informational purposes only and is not investment advice. IPOs and private-company valuations involve market, liquidity and disclosure risks. Readers should consult a qualified financial professional before acting on securities-related information. Figures are accurate as of publication on June 9, 2026.