OpenAI Files Confidential IPO S-1 Amid Surging Costs
OpenAI Confirms Confidential IPO Filing as Spending and Revenue Both Hit Record Highs
OpenAI, the San Francisco-based creator of ChatGPT, confirmed on June 8, 2026 that it has submitted a confidential draft S-1 registration statement to the U.S. Securities and Exchange Commission, formally launching the process toward what could become one of the most significant public offerings in technology history. The filing comes as newly surfaced financial figures reveal the full scale of the company's aggressive spending on compute infrastructure — and an equally aggressive revenue trajectory that has outpaced its own internal forecasts.
The company wasted little time making the news public. In a statement posted to its website, OpenAI said: "We recently submitted a confidential S-1. We expect it to leak so we're just announcing it." Goldman Sachs and Morgan Stanley are advising on the offering, according to CNBC and The Wall Street Journal, with a listing potentially targeting as early as September 2026.
A Company Spending at Scale — and Growing Into It
The financial picture emerging ahead of the IPO is one of extraordinary scale on both sides of the ledger. According to CNBC and the New York Times, OpenAI recorded approximately $3.7 billion in revenue for full-year 2024, while posting a net loss of approximately $5 billion — meaning it spent roughly $1.69 for every dollar it earned that year. Estimates from The Information, reported by Deep Learning AI's The Batch in August 2024, put OpenAI's total operating expenses at approximately $8.5 billion for 2024, including roughly $4 billion on Microsoft Azure inference costs and $3 billion on model training.
A more granular breakdown from Epoch AI, citing The Information and the New York Times, suggests OpenAI spent approximately $3 billion on training compute, $1.8 billion on inference compute, and $1 billion on research compute (amortized) during 2024 alone.
By 2025, the numbers had grown substantially on both sides. According to Reuters, reporting in February 2026 and citing a source familiar with the matter, OpenAI's full-year 2025 revenue reached $13 billion — beating its own $10 billion internal target. Cash burn for the year came in at $8 billion, under its $9 billion target. However, that improvement in cash discipline masks a dramatic surge in one specific cost category: inference compute.
According to internal Microsoft financial documents obtained by tech blogger Ed Zitron and reported by The Register in November 2025, OpenAI spent $8.7 billion on inference workloads on Microsoft Azure alone in just the first three quarters of 2025 — more than double the $3.7 billion spent on inference across all of 2024. The same documents show OpenAI paid Microsoft more than $12 billion in compute costs since 2024, and made $493.8 million in revenue share payments to Microsoft in 2024, a figure that rose to $865.8 million in only the first three quarters of 2025.
According to Reuters and The Information, OpenAI's inference expenses increased fourfold in 2025, causing its adjusted gross margin to fall to 33%, down from 40% in 2024. OpenAI CFO Sarah Friar confirmed that the company ended 2025 with 1.9 gigawatts of compute capacity — a 9.5x increase from 200 megawatts in 2023, according to Data Center Dynamics.
Revenue Growth Has Been Exceptional — But Margins Remain Pressured
Despite the mounting costs, OpenAI's revenue growth has been among the fastest ever recorded for a private technology company. According to OpenAI's own disclosures, as cited by tech-insider.org, the company's annual recurring revenue scaled from approximately $2 billion in 2023 to $6 billion in 2024 to over $20 billion annualized at the end of 2025. By February 2026, its annualized revenue run rate had crossed $25 billion, according to multiple sources including Sacra and Investing.com.
Looking further ahead, OpenAI has projected revenues exceeding $280 billion by 2030, according to Reuters and CNBC reporting in February 2026 that cited a source familiar with the company's investor presentations. The company has also projected cumulative losses of $44 billion between 2023 and 2028, before turning a profit of approximately $14 billion in 2029, according to Yahoo Finance citing The Information.
On the cost side, OpenAI is targeting approximately $600 billion in total compute spend through 2030, according to Reuters. That figure is a significant revision downward from the $1.4 trillion compute spend CEO Sam Altman had previously promoted.
The IPO Filing in Context: Valuation, Rivals, and What Comes Next
OpenAI's confidential S-1 filing follows a March 2026 funding round that closed $122 billion in committed capital at an $852 billion post-money valuation, according to CBS News and Fortune. That valuation positions the company for a public debut that could see it cross the $1 trillion mark — though the path there runs directly through its ability to improve gross margins, which remain constrained by surging inference costs.
Wedbush Securities analysts, as quoted by CBS News, described OpenAI's forthcoming financial disclosures as "highly anticipated given the speculation around the company burning through cash to secure compute and build out infrastructure to train and run AI models." The S-1, when it becomes public, will mark the first time investors and the broader public receive a comprehensive, audited view of OpenAI's financials.
OpenAI is not alone in moving toward a public listing. According to Fortune and Reuters, rival AI company Anthropic submitted its own confidential S-1 filing on June 1, 2026 — one week before OpenAI's announcement. SpaceX has separately begun an IPO roadshow. The wave of major technology listings anticipated for the second half of 2026 reflects both investor appetite for AI exposure and the capital demands these companies face as they scale infrastructure.
The Microsoft relationship will likely receive particular scrutiny from public market investors. OpenAI receives compute from Microsoft at discounted rates as part of their ongoing partnership, but the leaked documents reported by The Register suggest the cost of that relationship — and OpenAI's dependence on Azure infrastructure — has grown substantially. Revenue share payments to Microsoft nearly doubled between full-year 2024 and just the first three quarters of 2025.
Why This Matters Beyond Wall Street
The financial figures now emerging from OpenAI's pre-IPO disclosures offer the clearest picture yet of what it actually costs to build and run frontier AI systems at scale. The surge in inference costs — spending associated with running AI models for end users, not just training them — is particularly significant. It suggests that as AI adoption grows, the economics of serving those users at scale remain deeply challenging, even for the market leader.
OpenAI's 33% adjusted gross margin, while an improvement on earlier projections, still lags far behind the margins typical of established software companies, which often operate above 70%. Closing that gap will likely be one of the central narratives of OpenAI's public market story. The company's path to profitability — projected around 2029 — depends on either a dramatic reduction in per-query inference costs, a continued steep rise in revenue, or both.
The scale of the compute buildout also has implications far beyond OpenAI itself. The company's 9.5x increase in compute capacity between 2023 and 2025, combined with the projected $600 billion in total compute spend through 2030, signals the kind of capital intensity that is reshaping energy grids, data center markets, and semiconductor supply chains globally.
For everyday users of AI tools — whether for work, health management, or personal productivity — the financial pressures on companies like OpenAI will shape which products get built, how they are priced, and how sustainable access to them will be over the long term.
For more tech news, visit our news section.
What to Watch
OpenAI has not announced a specific IPO date. Goldman Sachs and Morgan Stanley are advising on the offering, with a listing targeted as early as September 2026, according to CNBC and The Wall Street Journal. The confidential S-1 will become public at some point before any roadshow begins, at which point audited financial statements will provide a more complete picture of the company's cost structure, revenue breakdown, and forward guidance than has previously been available.
Key metrics to watch when the S-1 becomes public will include: gross margin trends and whether the fourfold increase in inference costs in 2025 has stabilized; the structure of OpenAI's Microsoft relationship and the terms of compute pricing; revenue diversification beyond ChatGPT's consumer subscriptions into enterprise; and the timeline and conditions attached to the projected 2029 profitability milestone.
Anthropic's parallel S-1 process will also provide a comparative data point for investors trying to assess how OpenAI's unit economics stack up against its closest rival.
Stay ahead of the AI economy. The financial forces shaping companies like OpenAI will directly influence the tools available for health tracking, workplace productivity, and personal optimization. Understanding where AI is headed — and what it costs — matters for anyone building habits and workflows around these technologies. Join the Moccet waitlist to stay ahead of the curve.