Semiconductor and cloud-infrastructure shares fell sharply after reporting that OpenAIs annualised revenue was lower than some investors had assumed. CNBC reported Nvidia down about 3 percent, Oracle nearly 6 percent and CoreWeave nearly 8 percent in the selloff described by multiple outlets.
Financial Times reporting, via Reuters and CNN Business summaries, put OpenAIs September annualised revenue near 50 billion dollars against a roughly 70 billion figure some had used. Part of the gap reflects how sales through cloud partners are counted when comparing with rivals such as Anthropic.
The market reaction was about comparability as much as demand: if headline growth rates are not measured alike, valuations built on them wobble. Analysts quoted in coverage cautioned against reading one month as a trend while still repricing near-term expectations.
Figures here are attributed company and market reporting, not audited results reviewed by Digital News Point. The article describes a valuation debate; it is not investment advice.
Reporting is based on statements and reporting available at publication time. Digital News Point verified the central facts against at least two reputable sources and attributes claims to their sources in the text. This story will be updated if confirmed new information materially changes the account, and corrections will follow the site corrections policy.
Why comparability moves prices
Technology valuations often rely on forward revenue multiples. When the underlying revenue definition changes, the multiple that investors are willing to pay can change even if underlying customer demand has not. A gross figure that includes the full amount billed to an end customer will look larger than a net figure that records only the amount retained after a cloud partner takes its share. Neither approach is unusual, but mixing the two in a single comparison can create a misleading gap.
Annualised run rates add a second layer of sensitivity. Multiplying one month of revenue by twelve gives a simple headline number, yet it assumes that month is representative. Seasonal buying, large enterprise contracts that land unevenly, and introductory pricing can all make a single month stronger or weaker than the trend that follows. Investors who built models around a higher run rate may therefore adjust positions quickly when a lower, differently defined figure enters public discussion.
What the selloff does and does not signal
A one-day decline in semiconductor and cloud-infrastructure shares signals repricing of near-term expectations, not a verdict on long-term demand for computing capacity. Data-centre builders, chip designers, and power and cooling suppliers operate on multi-year planning cycles. Orders, construction, and deployment continue on those timelines even when equity prices move sharply over a few sessions.
At the same time, valuation discipline matters because artificial-intelligence infrastructure requires substantial capital before revenue arrives. Companies must fund chips, buildings, networking, and electricity, and investors watch closely for evidence that spending converts into durable, paying workloads. When revenue measurement becomes clearer, markets can distinguish more readily between rapid growth with improving unit economics and rapid growth that depends on heavy discounting or partner subsidies.
What readers should watch next
Readers following this debate can look for three things in future disclosures. First, whether companies state clearly if figures are gross or net and what period they cover. Second, whether growth is described in audited financial statements or in unaudited operational updates. Third, whether commentary addresses costs alongside revenue, including the expense of training and serving models. Clearer answers on those points make comparisons with peers more meaningful and reduce the chance that a single headline number drives an outsized market reaction.
Digital News Point will continue to attribute figures to their original sources and to flag differences in definition where they affect comparability. That approach helps separate a genuine change in business performance from a change in how performance is counted.
