Gao Jianlin, the former head of Tencent’s chip business and leader of its Peng Lai Lab, has left the internet giant to found a new startup focused on high‑performance RISC‑V‑based CPUs for AI servers and agentic AI applications, according to SemiWatch.
Gao brings a dual‑stack background in Linux systems and silicon development. He formed Tencent’s hardware team in 2013, moving from FPGAs to AI chips in 2018, and delivered Tencent’s first self‑developed chip in 2020. Since then, he oversaw architecture, front‑end design, RISC‑V R&D, verification, and backend teams – covering AI accelerators, general‑purpose CPUs, and RISC‑V lines.
Notable achievements under his leadership:
First FPGA‑based cloud server in China (2017) for AI image recognition.
Tencent’s first TPU‑style AI chip – taped out and deployed in cloud vision, replacing Nvidia GPUs in certain workloads.
Multiple AI chip generations (2022‑2025) now running in large‑scale datacenter deployment.
The Canghai 2 chip – featuring a self‑developed RISC‑V core with >15 CoreMark/MHz – the first domestic RISC‑V core at Arm Neoverse‑class performance, and the first RISC‑V chip deployed in a public cloud in China.
Unlike x86/Arm, RISC‑V’s open ISA allows custom extensions without costly licenses – historically used in MCUs/embedded, but now pushing into HPC, AI, automotive, and datacenters. BofA Securities forecasts the server CPU market to reach $210B by 2030 (36% CAGR) , with CPU:GPU ratio tightening from 1:4 in training to 1:1 as agentic AI drives multi‑step loop workloads – turning the CPU from a sidekick into the scheduler and orchestrator of AI inference.
Gao’s new venture taps this window: building differentiated, domain‑specific RISC‑V cores for AI compute, memory access, and networking – a departure from off‑the‑shelf Arm IP.
ICgoodFind Takeaway:
Top silicon talent moving into RISC‑V server CPU design is a strong signal. This venture could accelerate China’s homegrown cloud‑scale AI compute ecosystem.
