OpenAI Just Crossed $25 Billion in Revenue — Here Is How the AI Industry's Economics Have Changed in 12 Months

Introduction: The Numbers That Rewrote the History Books

In mid-2024, OpenAI was generating approximately $3.4 billion in annualized revenue. It was widely cited as one of the fastest-growing companies in the history of software — and with good reason. Eighteen months earlier, it had been generating almost nothing. The trajectory was extraordinary, but it existed within a frame that analysts could at least compare to prior technology cycles.

That frame no longer applies.

By February 2026, OpenAI's annualized revenue had crossed $25 billion — a figure confirmed to Reuters by a person familiar with the company's finances, representing a 17 percent increase from the $21.4 billion the company was generating at the end of 2025. OpenAI filed a confidential S-1 registration statement with the SEC on June 8, 2026, setting the stage for what could be the largest IPO in technology history. Its rival Anthropic, which started 2024 with an annualized revenue run rate of $87 million, crossed $19 billion in March 2026 and $30 billion in April — a trajectory that has no comparison in the recorded history of enterprise software. Salesforce took twenty years to reach $30 billion in annual revenue. Anthropic did it in under three years from a standing start.

These numbers are not projections. They are disclosed figures from active companies with real customers paying real money. Understanding how the economics got here, what is driving the growth, and what these trajectories imply for the next two years is the most important business story in the technology industry right now.

How OpenAI Got From $3.4 Billion to $25 Billion in 18 Months

The trajectory of OpenAI's revenue growth is, on its own terms, the fastest sustained commercial scaling in technology history. The company generated $1 billion in revenue in 2023. That figure tripled to roughly $6 billion in 2024. By August 2025, the annualized run rate had reached $10 billion. It crossed $21.4 billion at year-end 2025, and $25 billion by February 2026. The company is now generating approximately $2 billion per month.

The product mix driving that growth has shifted substantially over the period. ChatGPT subscriptions — the consumer product that most people interact with — drove the early growth, capitalizing on a user base that has now reached more than 900 million weekly active users. OpenAI reports more than 50 million consumer subscribers at a range of price points, from the $20-per-month Plus tier to the $200-per-month Pro tier. That consumer base is both OpenAI's most visible revenue source and, increasingly, its smaller growth engine.

Enterprise is where the acceleration is coming from. Sacra's analysis of OpenAI's revenue mix found that enterprise represented more than 40 percent of total revenue by early 2026, up from approximately 30 percent a year earlier, and is on track to reach parity with consumer by the end of 2026. The company crossed one million business customers in November 2025. ChatGPT for Work reached seven million seats, growing 40 percent in just two months. OpenAI's API business — which powers third-party applications built on its models — is processing more than 15 billion tokens per minute. Each of those tokens represents a small payment from a developer, company, or enterprise customer building on OpenAI's infrastructure.

The enterprise push has been deliberate and structured. OpenAI has partnered with four of the world's largest consulting firms to help corporate clients move beyond pilot projects to full-scale AI deployments — precisely the conversion problem the AI pilot data discussed elsewhere this week identified as the key bottleneck in enterprise adoption. The partnerships give OpenAI a distribution channel into large organizations that do not buy software by clicking a website button.

The Anthropic Story Is Even More Extreme

If OpenAI's growth is the fastest in technology history, Anthropic's growth since mid-2025 is in a category that existing financial vocabulary struggles to describe.

The trajectory runs as follows: $87 million annualized in January 2024. $1 billion by December 2024. $9 billion by the end of 2025. $14 billion in February 2026. $19 billion in March 2026. $30 billion in April 2026. Meritech's Alex Clayton — who has reviewed the IPO trajectories of more than 200 public software companies — said he had never seen a growth rate like this before Anthropic's 2025 performance. It has accelerated since he made that statement.

The product driving Anthropic's acceleration is Claude Code, the company's agentic coding tool launched publicly in mid-2025. Claude Code reached $1 billion in annualized revenue within six months of launch. By February 2026 it was generating over $2.5 billion in run-rate revenue. Business subscriptions for Claude Code quadrupled between January 1 and mid-February 2026 alone. Claude Code is not a consumer chatbot product. It is professional infrastructure that software engineers and development teams use as a core part of their workflow. That distinction matters commercially: users who integrate a tool into their daily professional workflow do not cancel subscriptions the way casual consumers do.

Anthropic's revenue is overwhelmingly enterprise in character. More than 1,000 enterprise customers spend over $1 million per year on Claude — a figure that doubled in under two months following the company's $30 billion Series G raise in February 2026 at a $380 billion valuation. Claude is the only frontier AI model available across all three major cloud platforms simultaneously — AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry — which means enterprise customers using any of those platforms can access Claude without leaving their existing cloud environment.

A 79 percent overlap exists between OpenAI's and Anthropic's enterprise customer bases. The AI industry is not a winner-take-all market at the enterprise level. Organizations are not choosing between ChatGPT and Claude. They are using both, for different tasks, at different price points, routing between them based on specific requirements. That multi-model behavior is one of the defining structural features of how enterprise AI spending is actually distributed.

The Revenue Split and What It Signals About the Market

The consumer-versus-enterprise revenue split at both companies tells an important story about where AI money is actually coming from and where it is going.

OpenAI's consumer base — 900 million weekly active users, 50 million paying subscribers — is the largest in the industry by a significant margin. Consumer revenue has scale advantages: the unit economics of a $20 monthly subscription are simple, churn is manageable, and word-of-mouth growth is free. The constraint is that consumer AI is increasingly a commodity. When Claude, Gemini, GPT, Meta AI, and a dozen open-source alternatives are all competitive on most everyday tasks, the price a consumer will pay for a subscription converges toward the most accessible option.

Enterprise AI behaves differently. An enterprise customer spending $1 million per year on Claude has integrated the model into workflows, trained staff on it, and built internal processes around its outputs. Switching costs are high. The relationship is stickier. Enterprise revenue also carries different growth dynamics: a single large account can expand from $1 million to $5 million per year as deployment broadens, without requiring any new customer acquisition spending. This is why Anthropic's $30 billion run rate, despite its much smaller consumer base, positions it as a genuinely formidable competitor to OpenAI rather than a permanent second-place finisher.

The most revealing metric from the enterprise side is token consumption. OpenAI's API is processing more than 15 billion tokens per minute across all applications. At even a conservative average price per token, that volume generates substantial revenue continuously, twenty-four hours a day, without any human sales effort. Token consumption grows with the model's utility — better models that companies use for more tasks naturally consume more tokens. The model capability improvements OpenAI and Anthropic are shipping every six to eight weeks are therefore directly revenue-accretive, which creates a compounding dynamic where technical progress and financial performance reinforce each other.

What $25 Billion Means for OpenAI's IPO

OpenAI filed its confidential S-1 with the SEC on June 8, 2026 — the formal first step toward a public listing. The company has explicitly said it has "not decided on timing yet" and that a listing "may be a while." But the filing puts the IPO machinery in motion, and the valuation math it anchors is significant.

Pre-IPO reporting places OpenAI's valuation in a range from $730 billion to $852 billion, with some analysts arguing that a $1 trillion valuation is achievable if revenue growth continues at its current pace. The $122 billion private fundraising round that OpenAI confirmed alongside the revenue disclosures was priced at an $852 billion post-money valuation — giving the IPO a clear floor to work from.

At $25 billion in annualized revenue and an $852 billion valuation, OpenAI trades at approximately 34 times revenue in the private market. That multiple is high by traditional software standards but defensible given the growth rate. A company growing from $3.4 billion to $25 billion in revenue over eighteen months, with a model suggesting continued acceleration, does not get valued on current earnings. It gets valued on the credibility of its future trajectory.

The critical caveat is the cash burn. OpenAI's inference costs reached $8.4 billion in 2025 and are projected to rise to $14.1 billion in 2026. Projected total cash burn for 2026 is approximately $27 billion, rising to around $63 billion in 2027. At $25 billion in revenue and $27 billion in projected cash burn, OpenAI is not yet profitable — and will not be for at least two to three years under current projections. Public market investors will price the IPO against the credibility of the path to profitability, which is the same calculation that justified losses for Amazon, Uber, and every other large-scale platform company that went public before its unit economics turned positive.

What the Growth Rate Signals for 2028

The most important number in this analysis is not $25 billion. It is the number that $25 billion implies about where the market is going.

Anthropic's internal projections — filed as part of its Series G materials and reported by The Information — show the company expecting to reach $70 billion in revenue and $17 billion in cash flow by 2028. OpenAI's own projections target $100 billion in revenue by 2027. Both figures are internal forecasts made by companies with obvious incentives to project optimistically. Both are also grounded in a growth trajectory that has consistently exceeded external analyst forecasts by substantial margins.

The more structurally significant question is who else benefits at this scale. Every dollar of AI revenue that OpenAI and Anthropic generate translates into compute spending that flows to Nvidia, SK Hynix, AMD, TSMC, and the data center operators building the infrastructure. It flows into the consulting firms helping enterprises deploy AI at scale. It flows into the software companies building on top of AI APIs rather than building models themselves. The $25 billion that OpenAI crossed in February 2026 is not the ceiling of the AI economy. It is the current floor of the most consequential technology companies in that economy, with a trajectory that has shown no sign of flattening.

Conclusion: The Economics Have Not Stabilized. They Are Still Accelerating.

The AI industry in June 2026 looks nothing like the AI industry in June 2024. Two years ago, $25 billion in annualized AI revenue was a speculative projection for 2028 at the optimistic end of analyst range. Today it is a confirmed figure for February 2026, already surpassed by a company that had $87 million in run-rate revenue eighteen months before that same February.

What changed is not the technology, which has been on a consistent improvement trajectory for years. What changed is the commercial infrastructure: the enterprise distribution channels, the API ecosystems, the developer tools, and the MCP-based integration standards that make it possible for AI capabilities to generate revenue at scale without requiring manual sales effort for each individual deployment. The technology became a platform. Platforms compound. And the compounding in this industry is moving faster than any prior technology cycle has moved at equivalent scale.

OpenAI's S-1 filing will eventually produce a prospectus that forces the company to disclose what all of this looks like with audited precision. When that document arrives, it will be one of the most studied financial filings in the history of technology. The revenue trajectory it will describe is already visible in outline. What it will add is the detail of how $25 billion in revenue is actually distributed, and what the realistic path to profitability looks like against a cost structure that is currently growing as fast as the revenue.

Until then, the number that matters most is the one that keeps moving: revenue that doubled in the first eight months of 2025, doubled again in the next four months, and is currently growing at a rate that the history of software does not have a useful precedent for.