The startup, which emerged as an Intel spinoff in 2020, is introducing its Active Compute Fabric to solve a persistent bottleneck in AI development: the idle time GPUs spend waiting for data delivery. Unlike Nvidia’s tightly integrated ecosystem, which incentivizes users to stick to a proprietary stack, Cornelis offers an open architecture. This design allows engineers to pair the networking fabric with a diverse range of third-party accelerators and GPUs.
By enabling chips to process and transmit information simultaneously, the company hopes to reclaim wasted compute cycles. Shipping is already underway, and the firm is currently finalizing a next-generation iteration of the technology scheduled for release later this year. This push represents a broader industry shift toward specialized infrastructure designed to weaken Nvidia’s grip on the AI hardware market.
Comments (0)
No comments yet. Be the first!