OpenAI Revenue Below Expectations Sends AI-Linked Stocks Lower

OpenAI told investors its annualized revenue was nearing $50 billion at the end of September 2026, according to a Financial Times report published on October 8, 2026. That is about $20 billion below the figure media outlets had been citing weeks earlier, reported as roughly $70 billion by the FT and about $68 billion elsewhere. The gap reached the market the same day, and AI-linked shares fell as investors reassessed how much growth is already priced into the sector's data-center spending. (CNN Business, AFP via Yahoo Finance)

Related reading: Anthropic and OpenAI Cut Model Prices in Their First Releases Since the Slowdown Debate · OpenAI's Astra Is Taking Anthropic's Enterprise Lead. Anthropic May Answer With a New Model.

The two numbers measure different things, which is where the confusion started. OpenAI's figure is what the company told its financial backers. The larger figures that circulated earlier came from investor information and, according to CNBC, included gross revenue from OpenAI's partners. (CNBC)

What OpenAI told investors

The FT reported that OpenAI shared the $50 billion annualized figure with investors as of the end of September 2026. CNBC confirmed the number and added two growth metrics the company also gave its backers: 77% total run-rate growth in the third quarter, and 107% run-rate growth in its enterprise business.

The figures reached the public through different channels. The FT based its report on documents shared with OpenAI's financial backers, and the earlier, larger figures had circulated in late September on the strength of investor information rather than a company disclosure. Each number is only as good as the source relaying it, and none has been reconciled against an audited statement.

Those percentages are strong by the standard of almost any software company. The complication is the base. OpenAI is private, and the growth figures are measured against numbers the company reports to investors rather than audited financial statements. No public filing lets an outsider check the arithmetic.

The higher figure counted gross revenue from partners. Gross revenue includes money that passes through to those partners, so it inflates the total when the goal is to compare one lab with another. CNBC reported that the partner-inclusive basis was used to line OpenAI up directly with Anthropic, its closest private rival. On a like-for-like basis, the distance between the two companies looks different from what the headline numbers suggested.

A lower figure is not automatically bad news for OpenAI. A company heading toward a public listing can prefer a number it can defend to a larger one it cannot, and the enterprise growth rate it shared alongside the $50 billion points to a business shifting toward higher-value customers. But the change in the headline number, from roughly $68 billion to $50 billion, is large enough to reshape the demand picture investors were working from.

Why the number matters beyond OpenAI

Both OpenAI and Anthropic are private, so investors have no audited revenue to anchor on. That makes the annualized run rate the number that moves markets, and it feeds straight into the case for the AI buildout. Data centers, chips and power contracts are justified by expected demand. When the demand signal from the industry's best-known company lands $20 billion below expectations, the math behind that spending gets harder to defend in the near term.

The mechanics matter here. A run rate annualizes a single recent month of sales, so it can move sharply on one strong or weak period, and it ignores costs entirely. OpenAI's spending on compute, chips and data centers is the other half of the equation, and it is not part of the figure investors were handed. A company can post rapid run-rate growth and still lose money at a rate that makes the growth hard to sustain.

The market reaction showed how much was riding on the estimate. On October 8, 2026, the Nasdaq Composite fell 1.25% to 27,193.34 and the S&P 500 fell 0.47% to 7,765.36, while the Dow Jones Industrial Average rose 0.1% to 51,231.64. The damage was concentrated in AI names: Oracle shares fell more than 5%, Nvidia about 3% and AMD almost 4%, and CoreWeave and other AI-linked stocks sank. (CNBC markets live blog, IndexBox)

The index-level moves understate what happened underneath. A modest decline in the S&P 500, paired with a much rougher day for the companies most exposed to AI capital spending, points to a rotation rather than a broad selloff. The Dow's small gain is a reminder that the selling centered on the companies whose valuations depend on AI revenue arriving on schedule, and not on tech as a whole.

Anthropic sets the comparison

Part of what makes the OpenAI figure awkward is where Anthropic stands. Anthropic is expected to launch its IPO later this fall, reportedly seeking a valuation around $2 trillion. In August 2026, Bloomberg and Reuters reported that Anthropic was on track for an annualized revenue run rate above $65 billion. Bloomberg noted that Anthropic's growth had helped lift global equities earlier in the same week that OpenAI's figure dragged them down.

If those reports hold, Anthropic's run rate sits above OpenAI's disclosed figure, which reverses the ordering many investors had assumed. That is the comparison CNBC's partner-revenue explanation speaks to: once the money that flows to partners is stripped out, the two labs may be closer than the earlier $68 billion figure implied, or Anthropic may genuinely be ahead. Neither company's numbers are audited, so the ranking rests on reported figures rather than verified accounts.

For Anthropic, the disclosure calculus is different because it is closer to a public listing. An IPO filing would put audited or at least reviewed figures in front of the market, which is more than either lab's private estimates can offer. If Anthropic's filing lands and shows revenue above OpenAI's $50 billion run rate, the gap will look less like a rounding difference and more like a lead heading into both companies' public-market debut.

OpenAI's other metrics are less ambiguous. The company had roughly 1 billion weekly active users as of late August 2026, according to CNBC, and its ChatGPT ads pilot crossed $1 billion in annualized run rate in under 200 days, a milestone it announced on August 31, 2026. Those are scale and monetization signals that support the growth story even as the headline revenue figure came in lower than expected.

Why the IPO timing raises the stakes

OpenAI is preparing to raise public money, and the Wall Street Journal has reported that the company missed internal revenue and user targets in its sprint toward a possible IPO. A company preparing to raise public money has more reason than usual to put its strongest number in front of investors, which makes the decision to disclose a $50 billion figure, rather than let the older, larger number stand, the more telling choice. It may reflect a preference for credibility with future public-market investors over a flattering private-round story.

Anthropic's expected IPO later this fall adds a competitive dimension. Two labs approaching public markets at the same time will be measured against each other, and the partner-revenue adjustment changes which one looks larger. The reporting suggests the comparison, more than either absolute number, is what unsettled AI-linked equities.

Caveats

Almost every number here is attributed rather than audited. Neither OpenAI nor Anthropic publishes financial statements, and the figures come from documents and information shared with investors and relayed by the FT, CNBC, Bloomberg, Reuters and the Wall Street Journal. The $50 billion figure is what OpenAI reportedly told investors; the $68 billion to $70 billion figures are what reporters had previously been given. The gross-versus-net distinction comes from a person familiar with the matter, as reported by CNBC, not from a public filing.

The difference between a run rate and revenue is not a technicality. Analysts and reporters use the run rate because it is the only window into a private company's demand, but it is a projection presented as a measurement. When two private labs are compared on run rates assembled from different bases, as OpenAI and Anthropic were, the comparison carries the assumptions of whoever compiled it.

Several open questions remain. It is not clear how much of the partner revenue is recurring, what OpenAI's costs look like against the $50 billion run rate, or whether the internal-target miss the Journal described reflects a temporary dip or a slower trend. None of those gaps can be closed from outside the company.

The annualized run rate is the number markets trade on, but it is an estimate the company gives its own investors, not a figure an auditor has signed off on.

What to watch

Three things will clarify the picture. First, whether OpenAI revises or confirms the $50 billion figure when it next speaks to investors, and how it describes the partner-revenue basis. Second, whether Anthropic's IPO filing, expected later this fall, discloses revenue on a comparable basis, which would let investors check the $65 billion report against a regulated document. Third, whether the data-center and chip commitments made on the strength of the higher estimates stay on schedule if the demand signal is weaker than the earlier figures implied.

OpenAI's weekly users, the ads pilot and enterprise growth give the company real momentum. But the market's reaction on October 8, 2026 showed how much rides on an estimate that only the company and its investors can see. One reported number, restated from $68 billion to $50 billion, was enough to pull AI-linked shares lower. Until audited financials exist, that uncertainty does not go away.

Cover image: Jernej Furman, Wikimedia Commons, CC BY 2.0.