Key Info
A developer at OpenCode argues that the idea of large hidden enterprise demand for open source models is misguided, pointing to the small number of top inference providers and the scarcity of engineers capable of running reliable inference.
Highlights
- Claims the top inference providers chasing enterprise business are all roughly the same size as his own company, and there are at most five of them
- Says very few people can run reliable inference, and those who can are highly paid and work at AI labs
- Argues enterprises seeking sovereignty won't get it by leasing GPUs from a third party
- Frames the hidden-demand argument as the wrong way to look at the market