We built the IPU from scratch for AI. Nvidia adapted a graphics chip. Architecture matters.

Being technically better doesn't guarantee winning. Ecosystems win, not specs.

$710 million to challenge the most valuable company in tech. We knew the odds.

AI chips are not one-size-fits-all. Different workloads need different architectures.

CUDA is Nvidia's real moat — not the chip. Software lock-in is the hardest barrier to break.

Bristol built the first computers. Building the next generation of AI chips here feels right.