The artificial intelligence stocks saw a sell-off on Thursday after reports said OpenAI’s reported annualised revenue was $20 billion short of previously reported figures.
The report raised questions about whether the revenue generated by AI companies can justify the sector’s substantial investment in chips, data centres and computing infrastructure.
The figure was around $50 billion, less than the $70 billion previously reported, prompting investors to reassess growth expectations across the AI ecosystem.
The Nasdaq Composite fell 1.25% on Thursday, while Nvidia, Advanced Micro Devices and Oracle dropped 2.9%, 3.9% and 5.5%, respectively.
However, the discrepancy did not necessarily indicate a sudden deterioration in OpenAI’s business.
Different revenue calculations appear to explain at least part of the gap, while subsequent reports pointed to continued growth expectations.
The episode nevertheless highlighted the sensitivity of AI-linked stocks to developments involving major customers and the funding needed to sustain the industry’s expansion.
Nasdaq Composite recovered 0.51% on Friday.
Revenue discrepancy triggers concerns across AI stocks
The Financial Times reported on October 8 that OpenAI’s annualised revenue was approximately $50 billion at the end of September, below the $70 billion figure previously circulated.
The distinction centres on how revenue generated through cloud-computing partners is counted.
Some earlier estimates included gross revenue associated with those partners to make comparisons with rival Anthropic more straightforward.
OpenAI’s reported figure excluded the partners’ share of certain sales.
Consequently, the difference should not be interpreted as evidence that OpenAI had suddenly lost $20 billion in customer spending.
Instead, it raised questions about how investors compare the financial performance of AI companies and assess their underlying growth.
The market reaction was immediate. Nvidia, a key supplier of chips used in AI systems, declined alongside AMD and Oracle, which has significant exposure to AI infrastructure and cloud computing.
The sell-off demonstrated how concerns about a major AI customer can spread to companies supplying the hardware, networking and computing capacity required to operate AI services.
Some of those losses were partly recovered in trading on October 9. Oracle shares gained 5%, and Coherent was up 1.85%. However, Nvidia was trading down 0.31%.
Spending, cash flow and valuations under scrutiny
Beyond the revenue discrepancy, investors are assessing whether the rapid expansion of AI infrastructure can generate sufficient returns to justify its cost.
According to Barclays, Microsoft, Alphabet, Amazon and Meta spent a combined $165 billion on capital equipment in the second quarter of 2026, up from $88 billion a year earlier.
Their combined free cash flow after those purchases was approximately $6.7 billion.
Bain and Company recently said the AI industry needs to generate $6 trillion in annual revenue by 2031 to justify the infrastructure buildout.
The figures illustrate the scale of investment being directed towards data centres and related infrastructure, alongside the relatively small amount of cash remaining after capital expenditure.
Higher Treasury yields add another consideration by increasing borrowing costs.
Investors are also questioning whether AI companies can sustain their spending plans without placing greater pressure on their balance sheets or relying on continued access to equity and debt markets.
Lee Yang, an analyst at Alchemy Markets in an Investing.com report, identified AI customer revenue, cash remaining after spending, and continued growth in chip orders as important indicators to monitor.
Weakness across those measures would provide a stronger warning about the sector’s outlook than a single revenue discrepancy, particularly when the reported figures use different accounting approaches.
Firmus IPO setback adds to AI funding concerns
Concerns about the AI investment cycle also received attention after Nvidia-backed cloud-computing company Firmus Grid scrapped plans for a $5 billion initial public offering in Australia.
Investors weren’t ready to back the company at the proposed $30 billion valuation.
Firmus operates data centres and rents computing capacity, including Nvidia chips, to customers such as OpenAI and Meta.
The proposed valuation was nearly three times the $10.5 billion at which the company had reportedly been valued two months earlier. Firmus has two operational data centres and five more in development.
Independent analyst Richard Windsor, who publishes the Radio Free Mobile blog, said Firmus could still raise money privately, but argued that its proposed valuation was too high given the company’s early stage and associated risks.
The abandoned listing does not, by itself, establish that demand for AI infrastructure is weakening.
However, it adds to scrutiny of the valuations being assigned to companies seeking to benefit from the AI boom.
The stakes extend to Nvidia, whose business depends in part on sustained demand from AI developers and infrastructure providers.
The prospect of major public listings by OpenAI and Anthropic has also become part of the market’s focus as investors look for evidence that leading AI businesses can support their valuations.
OpenAI’s revenue report therefore exposed a broader sensitivity within the AI trade: investors are not only watching spending on chips and data centres, but also the revenue, cash flow and financing conditions needed to sustain that investment.
The revenue discrepancy has an explanation, but questions over profitability, capital requirements and valuations remain central to the sector’s outlook.
The post How OpenAI’s revenue discrepancy shook the AI stock trade appeared first on Invezz