Chip maker Blaize is drawing attention for a processor it calls the GSP, designed to run artificial intelligence workloads at the "edge" — meaning on or near the devices where data is actually generated, rather than in a distant data center.
According to Jon Peddie Research, which covers the graphics and processor industry, the GSP "streams AI graphs at the edge." In AI systems, a "graph" is essentially the map of operations a model performs as it processes data. A chip built to stream those graphs is meant to keep that work flowing efficiently through the processor instead of repeatedly shuttling data back and forth, which is where conventional designs often lose time and burn power.
The pitch behind edge AI is straightforward. Running models close to the source — a camera, a sensor, a vehicle, a factory machine — can cut the delay of sending data to the cloud and back, reduce bandwidth costs, and keep sensitive information local. Those benefits matter most in settings where a fast, reliable response is essential and a round trip to the cloud is not practical.
The available reporting from Jon Peddie Research is brief and focuses on the GSP's core design idea rather than detailed specifications, pricing, or availability. Independent performance figures and customer deployments are not described in the source provided here.
Why it matters: as AI spreads from data centers into everyday hardware, processors that can run models locally and efficiently are becoming a key battleground — and Blaize's GSP is one company's bet on how to win it.