The AI Profit Fantasy

In years to come, we might look back at this period in tech as one in which the entire industry was in the grip of a mass delusion, namely the idea that artificial intelligence would prove to be good for business. Of course, it’s still early days. But so far, AI has proven lucrative primarily for the companies involved in making and selling AI chips and the servers they power—Nvidia, TSMC and Dell. For many others, it’s a money pit. We previously knew that was true of OpenAI, but a new example is Oracle, a software firm whose tiny cloud business is growing like gangbusters—at a cost to Oracle's fat profit margins. (The report sent Oracle stock tumbling as much as 5%, dragging down the overall market.)

As our story today revealed, Oracle earned a gross margin of just 14% on renting out servers powered by Nvidia chips in the three months ending with August. The company’s overall gross margin is closer to 70%. Seeing as Oracle has projected that AI cloud rental revenue will balloon through 2030, likely becoming its major source of revenue, its overall margins have to fall. And this issue isn’t just a problem for Oracle. Nvidia chips are expensive, and all the cloud firms that rent Nvidia-powered servers are suffering the impact. They might be generating AI cloud revenue, but profits in line with what they're used to? Fuhggedaboutit. (Catch our reporters talking about Oracle’s margins on The Information’s TITV today).