Intel has spent years selling AI silicon to laptop buyers. Its next flagship is aimed at machines that walk.
At a robotics gathering the company sketched a roadmap that puts Nova Lake in early 2027 and Razor Lake-AX in 2028 with a target of roughly 700 TOPS. Against the Panther Lake development board now used for robotics work, rated up to 180 TOPS across CPU, GPU and neural engine, that is close to a fourfold jump on paper.
The venue carried the message. Razor Lake-AX is not positioned as a Core Ultra successor for PCs but as infrastructure for physical AI, systems that fuse cameras, lidar, force sensing and proprioception with inference close to the hardware. That workload is latency-sensitive in a way chatbot inference never was, since a control loop cannot wait on a round trip to a data center.
The number deserves care. Intel has not said whether 700 TOPS is dense or sparse, a distinction that can swing vendor comparisons by a wide margin, and it described a roadmap goal rather than shipping silicon.
The rest of the industry is converging on the same bet. Renesas has stood up a physical AI team and Micron is developing memory for humanoid platforms. Whether the market arrives on schedule depends on locomotion, manipulation and safety certification, none of which a chip fixes.
