A single chip the size of a dinner plate. If NVIDIA was the past decade of AI compute, Cerebras is a candidate for the next.
Cerebras inference is materially faster than GPU baselines on frontier models.
G42 partnership locks in sovereign AI compute demand from the Middle East.
Filed S-1 already — one of the closest AI silicon names to a real ticker.
Cerebras carves a durable 5-10% share of the inference market.
NVIDIA CUDA ecosystem lock-in remains the hardest moat in tech.