When the company that sells most AI chips looks at buying an AI lab, it reads like a move to compete with OpenAI.
Reuters, citing the Financial Times, reports Nvidia is in talks to invest further in Reflection AI. Reflection builds open-weight models - trained models that are published so anyone can download and run on their own hardware. In fact, it just announced a new open-weight model called Beam.
Nvidia hasn’t said why it’s interested. But let’s compare the following two futures:
In one, three companies (OpenAI, Anthropic, Google) supply nearly all the AI models, and everyone else pays to use them. Nvidia’s chips mostly go to those big three and their cloud partners.
In the other, open models get good enough that banks, hospitals, and software companies run their own. Now thousands of organizations need chips.
Nvidia doesn’t need to own the winning model. It only needs more organizations running models.
That only works if two things hold:
- Demand for computing has to grow faster than models get cheaper to run.
- And those organizations have to buy Nvidia’s chips rather than AMD’s or Google’s.
The second part is where Nvidia’s software comes in. An open model can run on anyone’s chips, but Nvidia also builds the tools that tune models to run well on its hardware. Working closely with a lab like Reflection could make that lab’s model faster or cheaper to run on Nvidia’s hardware.
AI infrastructure has layers: apps on top, then models, then the software that runs them, then chips.
Nvidia’s advantage has mostly been in the bottom two layers. But those layers depend on each other:
- How a model is built decides how much computing it needs
- Software decides how efficiently that computing gets done
- And how widely the apps get used decides how much computing is needed in total
By helping build open models, Nvidia can influence all three.
Loading comments…
Comments aren’t loading right now.
The post is unaffected; the comment service didn’t respond. Try again in a moment, or come back later.