OpenAI has told investors that its annualized revenue run rate for September was nearly $50 billion, significantly below the figure approaching $70 billion that the artificial intelligence company had previously indicated, underpinning the difficulty of comparing the finances of fast-growing AI businesses as they prepare for potential public listings.
The discrepancy arose mainly from an effort to make OpenAI’s revenue figures directly comparable with those of rival Anthropic, according to a person familiar with the matter who spoke to Reuters on Thursday. The Financial Times first reported the latest revenue figure.
OpenAI had previously indicated to investors at a separate event that its September revenue run rate was approaching $70 billion, according to media reports.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
The difference largely reflects how the two companies account for revenue generated through cloud-computing partners. Unlike Anthropic, OpenAI does not include certain revenue from sales through cloud partners such as Amazon Web Services and Google Cloud in the figure it reports on the same basis.
Both companies distribute their AI models and services through third-party cloud platforms, creating differences in how revenue flows through their businesses and how those sales are presented to investors.
Anthropic pays cloud partners approximately 16% of every dollar earned through them. Revenue generated through those arrangements accounted for half of the company’s revenue last year, according to a Reuters analysis.
The difference in reporting methods complicates attempts to compare the companies using headline revenue figures alone. A lower reported run rate does not necessarily mean OpenAI’s underlying business has deteriorated by the same amount; it can also reflect which sales are included in a particular measure and how revenue from distribution partners is treated.
The issue is becoming increasingly significant as both companies prepare for potential initial public offerings, when investors will scrutinize their financial statements, revenue recognition policies, customer economics, and ability to sustain growth.
OpenAI entered this year with an annualized revenue run rate of $20 billion, compared with $6 billion in 2024, demonstrating the rapid expansion of demand for its AI products.
However, its quarterly revenue was overtaken by Anthropic for the first time in the second quarter. OpenAI reported $6.7 billion in quarterly revenue, compared with Anthropic’s $11.5 billion.
Anthropic’s annualized revenue run rate crossed $65 billion in July and is expected to reach $100 billion by the end of the year, sources previously told Reuters.
Those figures have placed the two companies at the center of an intense contest over enterprise AI spending, coding tools and the commercial adoption of large language models. Their growth has also attracted investor attention as the AI industry absorbs enormous sums of capital to develop models, build computing infrastructure and expand distribution.
But the latest OpenAI figures show why headline growth claims require careful interpretation. Annualized revenue run rates are not the same as revenue already earned over a full year, nor do they establish whether a company is profitable.
The measure commonly involves multiplying revenue from a single month by 12 to estimate what the business would generate over a year if its current pace continued. It can be useful for rapidly expanding companies whose historical financial results do not fully capture their current scale, but it can also exaggerate the durability of recent growth.
Monthly sales can fluctuate, customer demand can change, and new products may experience rapid adoption that does not persist. Differences in accounting and distribution arrangements can further complicate comparisons between businesses.
For OpenAI, the gap between the previously signaled figure approaching $70 billion and the latest figure of nearly $50 billion makes the reporting basis particularly important. Investors will need to distinguish between the company’s underlying sales momentum and differences arising from the treatment of cloud-partner revenue.
The figures also show why comparisons between OpenAI and Anthropic cannot rely exclusively on annualized revenue estimates. A meaningful assessment requires consistent definitions of revenue, clear disclosure of third-party distribution arrangements and a closer examination of the costs associated with generating each dollar of sales.
IPO Preparations Will Bring Greater Scrutiny of AI Economics
The potential public listings of OpenAI and Anthropic are likely to make these distinctions more consequential. Public-market investors will seek a clearer understanding of how much revenue each company generates directly, how much comes through distribution partners, and how much it retains after paying for computing capacity and other operating costs.
Anthropic’s cloud partnerships illustrate the complexity. If cloud partners receive approximately 16% of revenue generated through their platforms, the gross amount associated with those sales does not translate directly into revenue retained by Anthropic. Additionally, OpenAI’s decision not to include certain cloud-partner sales in its comparable figure means that headline numbers may describe different economic measures.
Neither approach, on its own, establishes which company has the stronger business. The comparison depends on the precise accounting definitions used and on whether the figures are intended to represent total customer spending, recognized revenue, or revenue after particular partner payments.
Profitability will be another central consideration. Rapid revenue growth can demonstrate strong demand, but it does not reveal whether a company can generate sustainable earnings after accounting for the costs of training and operating AI models, purchasing computing capacity, developing new products and acquiring customers.
That backdrop is deemed consequential in an industry where demand is rising alongside substantial spending on data centers, chips and cloud infrastructure. Investors will want to know whether growing sales are translating into improving margins and whether companies can maintain their pace of expansion without proportionately increasing their costs.
OpenAI’s rise from $6 billion in annualized revenue in 2024 to $20 billion at the start of this year illustrates the scale of its expansion. Anthropic’s reported quarterly lead and its projected $100 billion year-end run rate, meanwhile, suggest that competition for enterprise AI spending is intensifying.
Yet the latest reporting discrepancy is a reminder that growth figures must be assessed on a consistent basis before they can support conclusions about market leadership or business quality.
As the companies move toward potential IPOs, the major concern for investors will not simply be which AI developer can report the largest revenue run rate. It will be which business can convert demand for AI into durable revenue, defend its margins, and generate sustainable returns on the capital required to compete.
Until comparable financial disclosures become available, headline annualized revenue figures offer an indication of momentum, but only a partial view of the economics behind the AI boom.