Oracle’s AI ambitions face scrutiny on earnings miss — An Oracle office in Silicon Valley, California. Image: ykanazawa1999

Oracle forecast sales and profit that missed analyst estimates on Wednesday, while saying that spending would rise by US$15-billion compared with earlier estimates – a sign that big capital outlays to chase AI cloud-computing customers is not turning into profit as fast as Wall Street had expected.

Shares of the Austin, Texas-based company slumped 10% in extended trading.

Oracle has leapt to renewed prominence with grand plans to build artificial intelligence cloud data centres, and its results are viewed as a sign of whether there is an AI bubble and how it will raise money to build that infrastructure.

Oracle said that adjusted profit for the current fiscal third quarter would be $1.64 to $1.68/share, below analyst estimates of $1.72/share, according to LSEG data. Oracle’s third-quarter revenue growth forecast of between 16% and 18% also missed analyst estimates of 19.4% growth to $16.9-billion, according to LSEG data, and Oracle’s entire forecast range of cloud sales growth also missed LSEG estimates of $8.9-billion.

At the same time, Oracle executives said that capital expenditures for fiscal 2026 are now expected to be $15-billion higher than the $35-billion figure the company estimated in September during its first-quarter earnings call.

“The ramp in capex and unclear debt needs are causing uncertainty among investors,” said Melissa Otto, head of research at S&P Global’s Visible Alpha.

Capital expenditure

For the just-ended fiscal second quarter, Oracle reported total revenue of $16.1-billion, compared with analysts’ average estimate of $16.2-billion, according to data compiled by LSEG. Adjusted operating income of $6.7-billion also missed Wall Street’s average target of $6.8-billion, according to LSEG data.

“Although Oracle’s shares are buoyed by its September surge, this revenue miss will likely exacerbate concerns among already cautious investors about its OpenAI deal and its aggressive AI spending,” eMarketer analyst Jacob Bourne said in a statement.

Oracle’s closely watched metric for future cloud contracts also missed Wall Street estimates.

Read: OpenAI bets $38-billion on AWS in cloud power grab

Oracle also reported $523-billion in future contracts, up 14.94% from the $455-billion it reported in September, when it revealed a slew of cloud-computing deals with ChatGPT creator OpenAI and others that sent its shares skyrocketing. But the $523-billion figure fell below analyst estimates of $526-billion, according to Visible Alpha data.

On a conference call with analysts, Oracle CEO Clay Magouyrk fielded questions on how Oracle would finance building the data centres needed for its cloud contracts.

“We have some other interesting models that we’ve been working on,” he said. “One of them is that customers can actually bring their own chips, and in those models, Oracle obviously doesn’t have to incur any capital expenditures upfront for that model.”

He added: “Similarly, we have different models that we’re working on with different vendors, where some vendors are actually very interested in a model where they rent their capacity rather than selling that capacity.”

Oracle posted fiscal second-quarter adjusted profit of $2.26/share, above analyst estimates of $1.64, according to LSEG data. However, Oracle said both adjusted and unadjusted profits were higher on a one-time $2.7-billion pretax gain on selling its stake in chip designer Ampere Computing.

Larry Ellison, Oracle chairman, said the firm chose to sell its shares in Ampere because it plans to have a policy of neutrality about which chips it uses in its data centres and that “we no longer think it is strategic for us to continue designing, manufacturing and using our own chips in our cloud data centres.”

Latest chips

Ellison said that Oracle would continue to buy Nvidia’s latest chips, but that “we need to be prepared and able to deploy whatever chips our customers want to buy”.

Oracle is building massive data centres for OpenAI, which Reuters has reported is working with Broadcom to develop its own custom AI chip.