> What do they expect to get out of ... Nvidia did exactly this with CUDA, so go take a look at the ML world for an example of how it works. It seems to be goi

> What do they expect to get out of ... Nvidia did exactly this with CUDA, so go take a look at the ML world for an example of how it works. It seems to be going quite well for them. A common enough refrain is "I don't really want to buy Nvidia, but my toolchain requires CUDA". Pretty much every FPGA and SoC vendor does exactly this as far as I understand things. It's why you can't trivially run an open OS on most mobile hardware. Apparently such schemes don't meaningfully affect the purchasing decisions of a sufficiently large fraction of people to disincentivize the behavior.