Chief executive Safra Catz told analysts the company signed four multibillion-dollar contracts with three different customers in the quarter ending 31 August. Despite missing analyst forecasts, Oracle still waved around a claimed $455 billion in performance obligations.

Catz hinted that more deep-pocketed punters would be added in the coming months.

“We’ve signed significant cloud contracts with the who’s-who of AI, including OpenAI, xAI, Meta, Nvidia, AMD and many others,” she said.

Futurum Group chief executive and analyst Daniel Newman said: “Oracle has made the right moves to clearly position itself as not only a meaningful player in cloud, but a meaningful player in the AI race.”

While the megadeals got headlines, it was co-founder and chairman Larry Ellison who drew attention to the company’s ambitions in AI inference, the bit where trained models actually generate stuff. Ellison said Oracle would chase both the trillion-dollar training market and the inference side of things.

Ellison said: “Millions of customers are using those AI models to run businesses and governments. AI inference will be used to run robotic factories, robotic cars, robotic greenhouses, biomolecular simulations for drug designs.”

Ellison claimed Oracle’s upper hand comes from the fact it already hoards so much of its customers’ private data inside its databases. The company is punting a new AI database to let businesses query that information directly, which of course fattens Oracle’s AI inference tab.

“Inference is the workload that will grow fastest and will create the biggest opportunity,” Newman said. With mountains of data and plenty of iron behind it, Oracle reckons it has an edge.

Ellison said Oracle is pouring cash into AI agents.

“The new [applications] that we’re building, they’re nothing other than a bunch of AI agents that we generate that are linked together with workflow,” he said.