Should Nvidia be worried as OpenAI leans on Google's chips?
Hosts Jake Conley, Julie Hyman, and Pras Subramanian discuss reports that OpenAI (OPAI.PVT) is working with Google (GOOG) to make AI chips. They consider whether Nvidia (NVDA) should be worried as companies look for alternatives to relying on one chip maker.
OpenAI is also looking at chips, though not necessarily memory chips. I think this is a big move for OpenAI. The Information reports that the company plans to work with Google to make AI chips for its own use. This is a move away from Nvidia, as many companies are trying to find alternatives to its chips. OpenAI is also designing its own chips, called Jalapeño.
Jalapeño.
Jalapeño, right. Adding Google's chips gives OpenAI another way to build its own computing power. This is a big move. But should Nvidia be worried?
Maybe. I think OpenAI has already been renting some of Google's chips. Now it may buy them instead. We saw something similar with Anthropic as part of the S12 deal: it had been renting computing power and is now going to buy more. That difference is interesting and says something about where we are in this cycle. The short answer is yes, Nvidia should be somewhat worried. But that's why Nvidia has invested in all these other things, right?
Right. That's why Nvidia is investing in other AI companies.
Exactly. Those investments help Nvidia spread its risk in case this happens.
For Google, building more of its TPUs is a big win. It means getting a bigger deal and building a stronger relationship with a major customer. OpenAI is one of two major customers Google could target. But as you said about Nvidia, this is what Jensen Huang has been saying on earnings calls. Nvidia is about more than chips. The company has a lot of money invested across the business. It doesn't just make chips. It also provides the racks, the hardware, and everything else customers need.
It also provides software. And Nvidia is becoming a networking company and a software company. It is now also a security company with its OpenShell product. So yes, it is trying to cover every part of the market.
That's Huang's point: Nvidia wants to be the foundation that the whole industry depends on. The idea is that a rival like Google won't be able to replace Nvidia just by winning a few customers with its chips.
What's interesting is that SpaceX relies almost entirely on Nvidia. It is strongly committed to Nvidia and doesn't seem to be moving to other suppliers. I mention that because Oracle also announced at Oracle World that it will offer Google's TPU chips through its cloud computing services. Oracle also announced a deal with OpenAI, allowing OpenAI tools such as Codex and ChatGPT to run through Oracle's cloud services. Many companies are working together to offer these solutions, possibly giving customers one place to get everything they need.