The labs
behind the models.

The GPU monopolist that powers virtually every frontier model.

LPU silicon for ultra-low-latency LLM inference.

Reconfigurable dataflow silicon for enterprise AI.

Wafer-scale silicon — entire model on a single chip.

Ascend silicon + Pangu LLM — China's domestic AI stack.

IPU silicon — UK's homegrown AI accelerator.

AI training accelerators — acquired by Intel in 2019.

Jim Keller's RISC-V AI silicon company.