OpenAI’s annualised revenue run rate reportedly approached US$50 billion in September 2026, roughly US$20 billion below an earlier comparable figure. That is not necessarily a collapse in sales: the difference highlights how rapidly growing AI companies can report very different revenue measures.
What was reported on 8 October?
Reuters, citing a person familiar with the matter, reported that OpenAI had told investors its September annualised revenue was almost US$50 billion. Earlier communications had suggested a figure approaching US$70 billion. The reported gap arose primarily from efforts to compare OpenAI’s revenue with rival Anthropic on a similar basis. OpenAI had not publicly confirmed the detailed figures in the report.
Annualised revenue is not the same as revenue earned during a financial year. A simple run rate multiplies one month’s revenue by twelve. It can help illustrate momentum, but assumes that the month is representative of the coming year. In a fast-growing and capital-intensive market, that assumption can be unreliable.
Why cloud partnerships change the comparison
Reuters reported that OpenAI excludes some sales made through cloud partners from the metric, whereas Anthropic includes sales through partners such as major cloud platforms. This distinction matters: an AI developer may sell directly to customers, through a reseller, or via a cloud marketplace. Counting gross marketplace transactions and counting only the developer’s retained revenue can produce materially different headline figures.
It is therefore misleading to conclude that one company is larger, faster-growing or more profitable solely by comparing two headline annualised figures without understanding how each was calculated.
What is known about Anthropic?
Reuters reported that Anthropic’s annualised revenue had crossed US$65 billion in July 2026, with cloud distribution representing a significant portion of its reported business. Those figures are media-reported company metrics, not audited full-year financial statements. Analysts should distinguish revenue generated by end-customer demand from the portion ultimately retained after cloud fees and commercial arrangements.
Firmus shows what happens when private expectations meet public investors
Australia’s Firmus withdrew its proposed ASX float in October after investors questioned the price and the financing required to turn contracted computing capacity into operating facilities. The abandoned listing was reported by Reuters on 8 October. The collapse of one transaction does not prove that all AI businesses are overvalued; it shows that prospective investors can demand evidence of construction progress, debt capacity and operating cash flow before accepting a headline valuation.
OpenAI’s 31 March 2026 funding announcement reported US$122 billion in committed capital at an US$852 billion post-money valuation. Anthropic’s 12 February announcement reported a US$30 billion financing at a US$380 billion post-money valuation. These are dated private financing valuations, not live market prices or estimates of profits.
Revenue is not profit — or free cash flow
AI companies can record rapidly rising sales while simultaneously investing heavily in computing, training, research staff, data-centre capacity and customer acquisition. A company’s revenue run rate says little by itself about gross margin, operating losses, capital expenditure commitments or the timing of cash receipts.
Investors should ask five separate questions: How much revenue was actually recognised in the last quarter? What is included in the annualised metric? How much passes through partners? What is the cost of delivering the model? And how much cash is required to fund further growth?
What this means for Australia’s AI market
Australian businesses buying enterprise AI tools should not treat a vendor’s large valuation or sales headline as proof of product reliability. They should examine service-level agreements, data residency, cybersecurity controls, integration costs, pricing changes and exit options. Procurement teams should test whether a product produces measurable improvements rather than simply adopting the provider with the biggest revenue number.
For Australian investors, indirect exposure through listed technology companies and superannuation funds adds another layer. Public companies supplying chips, cloud capacity and power infrastructure can be affected by expectations for private AI developers even where the financial relationships are complex.
Why the next disclosures matter
As private AI companies consider further capital raising or potential listings, investors will increasingly want consistent definitions, audited results and detail on related-party and cloud distribution arrangements. Comparisons based on headline run rates alone may exaggerate both the speed and the quality of growth.
A five-point test for the next AI financing headline
- Identify the period: is the number trailing twelve-month revenue, a single month’s revenue multiplied by twelve, contracted sales or a forward projection?
- Identify the entity: does the number include resold cloud services, affiliates or pass-through spending?
- Check margins: what remains after model training, inference, chips, electricity, personnel and customer acquisition?
- Separate valuation from proceeds: a large post-money valuation does not mean the company received that amount in cash.
- Assess future funding: how much debt and equity will be needed before planned facilities generate cash?
These questions are particularly relevant when comparing software platforms with capital-intensive data-centre operators. The same revenue multiple may be inappropriate for businesses with very different cost structures and infrastructure obligations.
AI-generated image created for NextNews.
The bottom line
The striking part of the latest OpenAI and Anthropic figures is not simply their scale. It is how a change in revenue definition can shift the public narrative by tens of billions of dollars. The better question is not ‘Who has the biggest run rate?’ but ‘What revenue is earned, retained and converted into sustainable cash flow?’
Sources
Reuters, 8 October 2026 — OpenAI’s September annualised revenue; Axios, 8 October 2026 — Revenue accounting comparison.
Important disclaimer
This article is provided for general news and informational purposes only. It is not financial, investment, valuation, accounting or business procurement advice and should not be relied upon as a substitute for guidance tailored to your circumstances. Figures, laws and market conditions may change. Obtain independent advice from appropriately qualified professionals before making decisions or taking action. Read the NextNews disclaimer.
Disclaimer
NextNews strives for accurate news, but use it with caution—content changes often, external links may be iffy, and technical glitches happen. See the full disclaimer for details.