Nvidia Is Spending $26B to Train Its Own AI Models

Nvidia is committing roughly $26 billion to build and train its own AI models. For a company selling the chips that run most of the world’s AI, it makes sense to control both the hardware and the software.

The investment will go into supercomputing infrastructure, data acquisition, and model training teams. Basically, they want to be able to run massive AI models end-to-end without relying on anyone else.

So far, Nvidia has been the backbone for labs and companies training AI. This move puts them directly in the model game, alongside OpenAI, Google, and other labs already spending heavily to train next-gen AI systems.

For the rest of us, this means Nvidia could start delivering AI services and models directly, not just selling the chips to power them. It’s a big bet, and one that could reshape who owns the AI tech stack from the silicon to the software.