Intel targets robotics with new chip series

Intel is shifting its focus toward robotics by planning to release a new series of System-on-Chip (SoC) solutions called "Razer Lake-AX" within the next two years. These chips are expected to be primarily used in robotics and humanoid systems, as confirmed by Intel executive Cho Min-sung during a conference in South Korea on October 16.

AI performance boost expected

The "Razer Lake-AX" chips are anticipated to deliver up to 700 TOPS of AI computing power, a significant jump from the best "Panther Lake" solutions, which combined CPU, GPU, and NPU performance to only 180 TOPS. This would nearly quadruple the AI performance of future Intel SoCs compared to current models.

New chips will arrive in stages

Intel plans to begin mass production of autonomous mobile robots with some customers in 2027, followed by humanoid systems in 2028. The "Nova Lake" solutions, expected in early 2027, will precede the "Razer Lake-AX" chips, setting the stage for the new series.

Details of the chips remain unclear

Earlier leaks suggested the "Razer Lake-AX" chips might include a graphics component with 16 or 32 Xe3P cores, along with more memory capacity. However, the exact specifications of these future SoCs are still largely unknown.

Comparison with competitors is complex

If Intel achieves the promised 700 TOPS of AI computing power, these chips would surpass the current capabilities of the Nvidia Jetson AGX Orin modules. However, they may still fall short of the upcoming Nvidia Jetson Thor systems, which are expected to deliver up to 2,070 TFLOPS of AI performance in sparse mode. Comparing these systems is difficult due to differences in performance metrics and use cases.

Potential impact on the PC market

The "Razer Lake-AX" chips could help Intel compete with the next generation of AMD Ryzen AI Max mobile processors. These processors are increasingly being used in mini-PCs designed to handle local large language model (LLM) tasks, suggesting a growing demand for high-performance computing in compact devices.