OpenAI Revenue $20 Billion Lower Than Reported: Why AI Run-Rate Numbers Need Scrutiny

In the race to build the most valuable artificial intelligence company, revenue headlines are as much a competitive weapon as a financial disclosure. This week, one of the biggest headline numbers in the sector shrank by $20 billion. Here is what was reported, why the figures differ, and what it means for anyone trying to read the AI market.

What Was Reported

On 29 September 2026, Axios reported that OpenAI’s annualised revenue was “approaching $70 billion”. On 8 October, TechCrunch relayed a Financial Times report that OpenAI has told investors its annualised revenue is in fact “approaching $50 billion”, roughly $20 billion lower.

According to the FT, the higher number originated in information shared with OpenAI’s investors and was devised through their attempts to compare OpenAI with Anthropic’s annualised revenue. TechCrunch said it had reached out to OpenAI for comment; its article did not include a response.

Why the Numbers Differ: A Question of Method

The core of the discrepancy is methodology. Anthropic counts sales made through its cloud partners as part of its revenue, whereas OpenAI does not. Make OpenAI’s figures comparable to Anthropic’s, and a larger number comes out; present OpenAI’s own accounting, and a smaller one does.

This is not unusual in technology markets. Companies that sell through marketplaces and resellers can choose to report gross sales (the full amount paid by the end customer) or net revenue (only what they keep after partner fees). Neither is inherently wrong, but they are not interchangeable, and comparisons across companies are only meaningful when the definitions line up.

What “annualised revenue” actually means

Annualised revenue, often called run rate, takes a recent period (a month, sometimes a week) and multiplies it up to a full year. It is a snapshot of momentum, not recognised revenue. For fast-growing businesses it can look much larger than trailing twelve-month sales, and it can be distorted by one-off spikes, usage-based pricing and the treatment of partner sales.

• Run rate: a forward-looking extrapolation of recent revenue.
• Recognised revenue: what appears in audited accounts for a completed period.
• Gross vs net: whether partner-channel sales are counted in full or after fees.

The Competitive Backdrop: OpenAI and Anthropic

The reason investors were making the comparison at all is that the two companies are seen as the leading independent frontier-model developers. TechCrunch notes that Anthropic’s reported run rate made a $70 billion OpenAI figure look competitive. A CNBC headline linked from the piece cites Anthropic at $65 billion annualised in July 2026, although the TechCrunch article body does not state that figure itself, so it should be treated as a secondary reference.

Whatever the exact figures, the direction of travel is clear: both companies are growing quickly, and both are under pressure to demonstrate that enormous spending on compute can be turned into durable revenue.

The Money Behind the Story

Several data points help put the revenue discussion in context.

• Funding: OpenAI raised $122 billion in a funding round in March 2026.
• 2025 financials: leaked figures cited by Inc. showed about $13 billion in 2025 revenue, with spending significantly higher.
• IPO timing: according to CNBC (19 August 2026), OpenAI’s IPO, previously expected in 2026, has been pushed to early 2027.

Seen together, these show a company that has raised capital on a scale rarely seen, is growing revenue rapidly from a much smaller base, and is preparing to face public-market scrutiny. In that setting, the gap between $50 billion and $70 billion is not a rounding error. It affects valuation multiples, the story told to prospective IPO investors and the narrative about who is winning.

A Wider Wobble in AI Financing?

The revenue clarification arrives alongside other signs that investors are asking tougher questions about AI economics. The BBC reported this week that an Nvidia-backed AI data centre firm scrapped a landmark stock market listing over market fears, and that AI chip demand pushed Samsung’s profits to a record. The two stories point in different directions: chip and infrastructure suppliers are booming, while some companies further down the chain are finding public investors less willing to pay up.

The same week also saw bond markets under pressure, with yields at multi-year highs in the UK, France and the United States. Higher borrowing costs matter for an industry that depends on heavy, debt-assisted investment in data centres and chips. That context does not prove a downturn, but it explains why investors are reading every revenue figure more carefully.

How to Read AI Revenue Claims

For business leaders, analysts and readers following AI companies, a few habits make headlines easier to interpret.

1. Ask which metric it is. Run rate, annual recurring revenue, bookings and recognised revenue are different things.
2. Check gross versus net. Partner-channel sales can swing totals by large amounts.
3. Look at the source. A number shared with investors in a specific context, such as a comparison exercise, may not be the company’s formal accounting.
4. Compare like with like. Cross-company comparisons require matching definitions.
5. Consider costs, not only sales. Revenue growth means little without understanding the compute spending needed to generate it.
6. Wait for audited figures. An IPO prospectus will eventually force standardised disclosure.

What This Means for Businesses Buying AI

Most organisations adopting AI tools are not investors, but the story still has lessons.

Vendor stability matters. Pricing, product roadmaps and support depend on suppliers having the funds to keep investing. A vendor whose revenue is lower than market chatter suggested is not necessarily at risk, but it is a reminder to review contracts and concentration risk.

Avoid single-vendor lock-in. Designing workflows so that models can be swapped reduces exposure to commercial or technical surprises.

Expect pricing changes. As providers face pressure to show profitable growth, expect adjustments to usage limits, tiers and enterprise terms.

Ask for transparency. Procurement teams can reasonably ask vendors about uptime commitments, data handling and what happens if services change.

What to Watch Next

• Whether OpenAI provides an official statement or breakdown of the revenue figures.
• How investors and the press adjust comparisons with Anthropic and others.
• Any update to the early-2027 IPO timeline and the disclosure it brings.
• Whether more AI-related listings are delayed amid market volatility and higher bond yields.
• Continued strength in chip and infrastructure supplier earnings.

Frequently Asked Questions

Did OpenAI’s revenue actually fall by $20 billion?

No. Reports indicate the lower number reflects a clarification of what the company told investors and how it is measured, not a decline in sales. The $70 billion figure arose from comparison efforts with Anthropic.

Why does Anthropic’s accounting produce a bigger figure?

Per the FT as relayed by TechCrunch, Anthropic counts sales made through its cloud partners, and OpenAI does not.

When will OpenAI go public?

CNBC reported in August that the IPO, previously expected in 2026, had been pushed to early 2027. No official date has been confirmed in the sources reviewed.

Is this a sign of an AI bubble?

One restated metric does not settle that debate. It does highlight how heavily the sector depends on forward-looking numbers and why careful definitions matter.

The Bottom Line

A $20 billion swing in a headline number is a useful reminder that AI revenue figures are still being defined in real time. OpenAI’s “approaching $50 billion” and the earlier “approaching $70 billion” are not necessarily contradictory; they reflect different ways of counting in a race where comparison itself is part of the competition.

For investors, the lesson is to demand like-for-like definitions. For businesses buying AI services, it is to plan for a market that is still maturing, where pricing, partnerships and financing conditions can change quickly. And for everyone watching, the IPO filings expected in 2027 should finally replace headline run rates with audited numbers.

Sources: TechCrunch (8 October 2026), citing the Financial Times and earlier reporting by Axios, CNBC and Inc.; BBC Technology headlines (9 October 2026).