Bloomberg reported Thursday evening that OpenAI’s annualized revenue run rate has surpassed $40 billion, roughly doubling from the $20 billion pace CFO Sarah Friar cited at the end of 2025. Greg Brockman, co-founder and president, told staff in an internal announcement that monthly revenue expanded by more than 20% in July alone. The number landed clean, and the market treated it as a milestone.
It is a milestone. It is also not the most important number in the filing.
What’s Driving the Market
The revenue acceleration has three engines. Codex, OpenAI’s agentic coding product, grew roughly 6x inside ChatGPT Business and Enterprise between January and April, and OpenAI has said Codex had more than 3 million weekly active users by early April. Enterprise as a whole now accounts for more than 40% of total company revenue, up from near zero in 2023, and is on track to reach parity with consumer revenue by year-end. The nascent advertising business reached $100 million in annualized revenue in under six weeks after launch. ChatGPT crossed more than 900 million weekly users by April. These are not speculative numbers. They reflect real commercial adoption at a scale most software companies never approach.
The coding layer deserves particular attention from investors weighing the IPO. Codex users skew heavily toward paying customers, making the product a disproportionate revenue contributor relative to its user count. Knowledge workers now represent roughly 20% of all Codex users, the fastest-growing segment. That is not a developer tool story anymore. It is an enterprise workflow story, and enterprise workflows carry the kind of contract durability that subscription revenue alone does not.
Still, the revenue figure answers only half the question public investors will ask.
OpenAI burned $3.7 billion in Q1 2026 against $5.7 billion in revenue, spending more than half of every dollar it earned on cash costs. Internal projections cited by The Information show a $14 billion net loss for the full year, with cumulative losses through 2029 reaching roughly $115 billion before the company turns cash-flow positive. For context, Uber burned approximately $25 billion before GAAP profitability. OpenAI is on pace to exceed that figure in 2026 and 2027 combined.
The compute obligation is the core constraint. OpenAI has contracted an incremental $250 billion of Azure services from Microsoft, and the Stargate joint venture with SoftBank, Oracle, and MGX is in the early stages of a $500 billion multi-year infrastructure buildout. Inference costs per query have fallen sharply since GPT-4 launched in early 2023, and OpenAI has emphasized ongoing efficiency gains. The problem is that total spend is rising faster than the per-query efficiency gains, driven by the sheer volume of tokens processed. OpenAI has said its APIs process more than 15 billion tokens per minute.
The Investment Opportunity
The most direct way to hold a position on the OpenAI IPO outcome while the company remains private is through Microsoft. Microsoft holds roughly 27% of OpenAI on an as-converted diluted basis, and Microsoft has said its investment was valued at approximately $135 billion following the OpenAI recapitalization. The companies reworked their commercial agreement in late April, allowing OpenAI to serve products through cloud providers other than Azure while keeping Azure as its primary cloud partner. In the nine months ending March 31, 2026, Microsoft recorded $5.9 billion in net gains from its OpenAI investment, a sharp reversal from $2.7 billion in losses during the same period a year earlier.
That accounting swing matters for how investors read both companies. Microsoft absorbs its proportional share of OpenAI’s operating results under equity-method accounting. When OpenAI was losing more money, that drag showed up directly on Microsoft’s income statement. As OpenAI’s revenue accelerates and gross margins improve, the gain recognition reverses the pain. A public listing would convert the equity-method treatment to a marked position, changing the financial statement presentation entirely.
For investors who want commodity exposure to the AI infrastructure that makes this revenue possible, the Codex growth figures point toward the compute stack beneath them. More than 15 billion tokens per minute is a GPU utilization number as much as it is a product adoption number. The companies supplying the hardware and the power to run those workloads are the silent beneficiaries of every percentage point of OpenAI’s run-rate acceleration.
Risks to Monitor
The enterprise market share picture is more complicated than the headline revenue suggests. The Ramp AI Index, which tracks spending across tens of thousands of U.S. businesses, showed Anthropic crossing OpenAI in business adoption share in April 2026, reaching 34.4% versus OpenAI’s 32.3%. In the coding-specific layer, where Codex competes most directly, market share estimates vary widely and are not consistently auditable across vendors, which makes any single percentage claim worth discounting. Anthropic has been widely reported to be exploring a public offering, and its own annualized revenue figures, while calculated differently, have been cited at levels that will complicate the premium investors assign OpenAI’s consumer advantage.
The IPO timeline carries its own uncertainty. Reports have said OpenAI has explored filing confidential IPO paperwork, but OpenAI has not publicly confirmed a filing date, exchange, or offering price. Separately, OpenAI closed a March 31, 2026 fundraising round at a post-money valuation of $852 billion, and that private price already implies a forward revenue multiple that prices in dominant market share the enterprise spending data does not yet fully support.
The Microsoft profit-share structure also remains a live variable for prospective public investors. Revenue-share payments continue through 2030 at the same percentage, subject to a total cap the companies have not publicly disclosed, which means a meaningful portion of every incremental dollar OpenAI earns flows to a partner before it reaches OpenAI’s bottom line. The S-1, whenever it goes public, will be the first document to quantify this arrangement with audited precision.
Bottom Line
The $40 billion revenue run rate is the number OpenAI needed ahead of a public offering. It validates the commercial model, anchors the enterprise story, and gives underwriters a revenue line to defend in roadshow conversations. What it does not do is resolve the more consequential question: at what point does revenue growth outrun a burn rate that is currently projected to widen, not narrow, before 2029.
Investors who focus only on the top line are pricing a software company. The S-1, whenever it goes public, will force a harder conversation about compute obligations, partner profit-sharing, and a profitability timeline that requires everything to go right for several years in a row. The revenue is not in doubt. The path from $40 billion in revenue to a profitable public company is where the actual pricing work begins.
