What’s trending in AI on 9 October 2026: The biggest AI story of the past 24 hours is not a model launch. It is a number. On 8 October the Financial Times reported that OpenAI has told investors its annualized revenue was approaching $50 billion at the end of September, roughly $20 billion below the $68–70 billion figure that had circulated only a week earlier. The Nasdaq fell 1.25%, its worst day since mid-August, and the companies that sell OpenAI chips, cloud capacity and data-centre space fell hardest. The twist is that OpenAI’s business did not suddenly shrink. The earlier figure came from investors who counted revenue the way Anthropic does, including sales through partners. Below: what was reported, why one company can have two very different revenue numbers, why the market reacted so sharply, what it means for businesses that buy AI, and a vendor stress-test scorecard you can use this month.
Key takeaways
- The number: OpenAI told investors annualized revenue was approaching $50 billion at the end of September, per the FT, versus the roughly $70 billion widely reported in late September.
- The cause is counting, not collapse. Anthropic includes sales made through cloud partners such as AWS and Google Cloud; OpenAI’s own figure does not. Investors who tried to compare the two like for like produced the higher number.
- Growth is still fast. OpenAI reportedly told investors its run rate grew more than 70% since July, with 77% growth in Q3 and 107% in its enterprise business.
- The market still flinched. CoreWeave fell 8%, Oracle about 5.5% and Nvidia 2.9% on 8 October, because so much AI infrastructure spending is tied to OpenAI’s ability to pay.
- For buyers, this is a vendor-risk story. If your AI tools depend on one lab, one cloud or one pricing promise, now is the time to check contracts, exit options and cost assumptions.
1. What the Financial Times reported
According to financial documents reviewed by the FT and summarised by TechCrunch, Yahoo Finance, Investing.com and others, OpenAI recently told investors that its annualized revenue was close to $50 billion as of the end of September. That landed awkwardly, because Axios had reported on 29 September that the figure was nearing $70 billion, and CNBC had put it at $68 billion. Both earlier numbers were based on information that had been shared with OpenAI investors.
The FT’s explanation is that the higher figures came from OpenAI’s own backers trying to build a direct comparison with Anthropic’s revenue. Anthropic, which sells Claude through Amazon Web Services and Google Cloud as well as directly, counts those partner sales in its annualized revenue. OpenAI does not count partner-channel sales in the same way. One source told SiliconANGLE that the $68 billion figure was a gross number including partner revenue, while the $50 billion figure is net.
| Figure | Amount | What it counts | Reported by |
|---|---|---|---|
| OpenAI run rate, July | ~$30 billion | OpenAI’s own net method | FT, via TradingKey |
| Investor-adjusted OpenAI estimate, August | ~$40 billion | Partner-inclusive, to compare with Anthropic | FT, via Investing.com and Yahoo Finance |
| OpenAI run rate widely reported, late September | $68–70 billion | The $40 billion base plus OpenAI’s reported 70%+ growth | Axios, CNBC |
| OpenAI run rate told to investors, end of September | Approaching $50 billion | OpenAI’s own net method | FT |
| Anthropic reported run rate | ~$65 billion | Includes sales through cloud partners | CNBC, Bloomberg (August) |
OpenAI did not respond to requests for comment from AFP or TechCrunch at the time of their reports. The company also cited strong momentum to investors: SiliconANGLE’s source said OpenAI pointed to 77% growth in its annual run rate in the third quarter and 107% growth in its enterprise run rate.
2. Why one company can have two revenue numbers
Annualized revenue, or run rate, takes a recent period, usually one month, and multiplies it out to a year. If a company billed $4 billion last month, its run rate is $48 billion. It is a useful snapshot for fast-growing private companies that do not publish audited quarterly results, but it is not a regulated accounting measure. The month you pick, whether you count gross or net revenue, and how you treat sales made through partners can all move the answer by tens of billions.
The partner question is the heart of this story. When a business buys Claude through AWS or Google Cloud, the cloud provider bills the customer. Anthropic counts the gross sale. OpenAI’s own run-rate figure, by contrast, leaves partner-channel sales out. Neither approach is wrong. They simply answer different questions: how much customers spend on our models versus how much money arrives in our accounts.
TradingKey’s walk-through of the FT figures shows how the two paths diverged. Start from the partner-inclusive $40 billion investor estimate, add OpenAI’s reported 70%-plus growth, and you get roughly $68–70 billion. Start from OpenAI’s own July figure of about $30 billion, apply the same growth, and you get about $50 billion. Same growth, different baseline.
This matters most for the comparison investors were chasing. At $70 billion OpenAI looked slightly ahead of Anthropic’s reported $65 billion. At $50 billion, on its own method, it looks behind, even though the two figures are not like for like. With both companies heading towards public listings, that headline gap shapes perceptions of who is winning enterprise AI.
3. Why AI stocks fell anyway
If the gap is mostly a counting difference, why did the Nasdaq lose 1.25% in a day? Because OpenAI’s revenue has become a proxy for the whole AI build-out. OpenAI is not only a model maker; it is one of the biggest buyers of chips, cloud capacity and data-centre space. TradingKey notes it has committed hundreds of billions of dollars to computing and is projected to need close to $280 billion in cumulative funding by 2030. Suppliers have raised money, signed leases and ordered hardware on the assumption that this demand keeps growing.
So the selling was concentrated in OpenAI’s supply chain. According to SiliconANGLE and TradingKey, CoreWeave fell 8%, Oracle about 5.5%, Intel 5.3%, Micron 4.8%, Broadcom 4.3%, AMD 3.9% and Nvidia 2.9%. The S&P 500 fell only about 0.5%. Constellation Research analyst Holger Mueller told SiliconANGLE that in investors’ calculations there is “no room for error”, because shortfalls ripple through an ecosystem built on very large commitments.
Not everyone saw a crisis. D.A. Davidson’s Gil Luria said the reports caused “unnecessary concern”, and Neostellar’s Evan Schlossman described the change as a clarification of reporting rather than weaker demand, according to TradingKey. Both readings can be true: demand is still growing fast, and the market had priced in a number that was never on OpenAI’s own books.
4. The bigger picture: AI’s money is getting more fragile
The revenue story landed in a week already full of signs that AI financing is stretching. The AI to ROI newsletter’s 9 October roundup, summarising WSJ and Bloomberg reporting, describes SpaceX seeking about $40 billion of debt to buy Nvidia chips, Broadcom arranging more than $50 billion of financing tied to OpenAI’s custom chip, and Oracle negotiating an arrangement in which another company would buy chips and lease them back. It also notes record credit-default-swap spreads for SpaceX and Oracle, a sign lenders are charging more to insure against default.
Meanwhile the stakes keep rising. OpenAI raised $122 billion in March, its latest valuation was reported at $852 billion, and AFP says the FT reported talks on a new private round that could value it at about $1.4 trillion. Its IPO, once expected in 2026, has been pushed to 2027. Anthropic is reportedly preparing to list as soon as next month. Every one of those numbers rests on revenue growth continuing, which is why a $20 billion swing in a headline figure moved markets. We looked at the same mismatch between spending and revenue in AI’s trillion-dollar gap, and at its effect on hardware prices in the AI memory shortage.
5. What it means for businesses that buy AI
Most companies will never buy a share in OpenAI or Anthropic, but almost all of them now depend on one or both. Your CRM’s assistant, your help-desk bot and your coding tools probably call one of these labs behind the scenes. That makes the labs’ finances part of your supply-chain risk, in the same way a key supplier’s credit rating is.
Three practical consequences follow. First, pricing is not guaranteed. The current price war, which we covered in the 90-minute AI price war, is partly funded by investors. If capital tightens, discounts can shrink and usage tiers can change at renewal. Second, vendor concentration is a real risk. A SaaS vendor that has built on a single model provider inherits that provider’s product decisions, outages and price moves. Third, vendor claims deserve the same scrutiny investors are now applying. If a $20 billion difference can hide in a definition, so can a vendor’s claims about usage, savings or seats.
None of this is a reason to stop using AI. It is a reason to buy it the way you would buy any critical service: with exit options, cost controls and contract terms that survive a bad quarter upstream. Our guide to AI cost overruns covers the budgeting side, and our open-weight versus closed model comparison covers when a self-hosted fallback makes sense.
6. Your AI vendor stress-test scorecard
Score each critical AI tool you use against the eight questions below. Count your reds. One or two reds on a tool that touches customers or revenue is worth a conversation with the vendor before your next renewal; four or more means you should have a written fallback plan.
| Question | Green | Amber | Red |
|---|---|---|---|
| 1. Which model providers does the tool depend on? | Named, and more than one is supported | Named, one provider | Vendor will not say |
| 2. Can you switch the underlying model? | Yes, in settings or by config | On request, with notice | No |
| 3. How long is your price locked? | 12 months or more, in writing | Until renewal only | Can change with 30 days’ notice or less |
| 4. What happens if the provider raises prices? | Cap or right to exit in contract | Vendor “intends” to absorb it | Passed straight through |
| 5. Can you export your data, prompts and configurations? | Documented, standard formats | Manual or partial | Not possible |
| 6. What is the plan if the AI feature is withdrawn? | Tested manual fallback | Fallback exists on paper | None |
| 7. Are usage and cost visible to you? | Real-time dashboard and alerts | Monthly invoice detail | Single line item |
| 8. Who holds your data if the vendor is sold or fails? | Deletion and transfer terms in contract | Covered in privacy policy only | Not addressed |
Questions 3, 4 and 8 belong in the contract, not in a sales call. The White House AI Accord left most vendor commitments voluntary, as we noted in our accord explainer, so written terms are where you get real protection. If you run AI agents built on a single provider, such as the always-on agents OpenAI showed at DevDay 2026, add question 6 to your agent runbook too.
7. What to watch next
- Anthropic’s public prospectus. An S-1 filing would give the first audited view of how a frontier lab counts revenue, and could settle the gross-versus-net debate.
- OpenAI’s next disclosure. Watch whether OpenAI starts publishing a partner-inclusive figure alongside its net run rate to make comparisons easier.
- Supplier guidance. Oracle, CoreWeave and chipmakers will face questions about OpenAI-linked backlog on their next earnings calls.
- Pricing at renewal. If financing tightens, the first sign for buyers will be smaller discounts and new usage tiers, not headlines.
Frequently asked questions
What did the Financial Times report about OpenAI’s revenue?
On 8 October 2026 the FT reported that OpenAI told investors its annualized revenue was approaching $50 billion at the end of September. That is roughly $20 billion below the $68–70 billion figure widely reported in late September, which had been based on information shared with investors.
Is OpenAI’s revenue falling?
No. The gap comes from a difference in counting rather than a drop in sales. OpenAI reportedly told investors its run rate grew more than 70% since July, including 77% in the third quarter, with 107% growth in its enterprise business.
Why do OpenAI and Anthropic report revenue differently?
Anthropic includes sales of Claude made through cloud partners such as AWS and Google Cloud in its annualized revenue, a gross approach. OpenAI’s own figure does not include partner-channel sales in the same way. Investors comparing the two like for like produced the higher OpenAI estimate.
What is annualized revenue?
Annualized revenue, or run rate, takes revenue from a recent period, usually a month, and multiplies it to a full year. It is a common snapshot for fast-growing private companies but is not a regulated accounting measure, so the period chosen and the method used can change the result a lot.
Why did AI stocks fall after the OpenAI report?
OpenAI is one of the biggest buyers of chips, cloud capacity and data-centre space, so its revenue is treated as a gauge for the whole AI supply chain. On 8 October CoreWeave fell 8%, Oracle about 5.5% and Nvidia 2.9%, while the Nasdaq fell 1.25%.
What should businesses do about AI vendor risk?
Find out which model providers your AI tools depend on, whether you can switch models, how long prices are locked, and whether you can export data and fall back to manual processes. Put price caps, exit rights and data terms in contracts. This is general guidance, not legal or financial advice.
Sources
- TechCrunch: OpenAI’s revenue is reportedly $20 billion less than previously projected (8 Oct 2026)
- Yahoo Finance: OpenAI’s annualized revenue $20 billion lower than prior investor estimates (8 Oct 2026)
- AFP via Yahoo Finance: OpenAI revenue gap report rattles AI stocks (8 Oct 2026)
- Investing.com: OpenAI annualized revenue at $50bn, far below reports – FT (8 Oct 2026)
- SiliconANGLE: AI stocks crumble on report that OpenAI’s annualized revenue is much lower than previously believed (8 Oct 2026)
- TradingKey: Why is OpenAI’s annualized revenue $20 billion below expectations? (9 Oct 2026)
- AI to ROI: News and analysis, 9 October 2026 (summary of WSJ and Bloomberg AI financing reports)