AI & Advanced Computing
Nvidia Acquires HuggingFace for $12.9B
Reported by EE Times on 7 September 2026
AI & Advanced Computing
Reported by EE Times on 7 September 2026
Buying the place where open models live is a software move with hardware consequences. CUDA already made the GPU hard to leave; owning the default distribution point for models makes it harder again, and every competing accelerator vendor has just been reminded that their silicon is worth exactly as much as the software path onto it.
The hiring read is that AI hardware companies are now in a software arms race most of them staff badly. Compiler, kernel and runtime engineers are the binding constraint at every accelerator startup and at every incumbent building an alternative, and they are paid like software engineers while being recruited by hardware companies benchmarking against hardware salaries. That gap is why those roles sit open for six months, and no amount of rewriting the advert closes it.
Read the original at EE Times .
This is one of the markets we recruit in every day. The AI hiring that matters in these markets is not the same as general software hiring. It sits close to the hardware: models that run on an edge device, vision on an imaging sensor, security in an embedded system.
Send the role, or send your CV, and it reaches Ramon rather than a queue. If it is not something we can do well, you will be told that straight away rather than a month from now.
Get in touch