Staff Reporter |
Now that mass production is largely sorted out, China’s RISC-V is aiming higher.
At the 5th Dishui Lake China RISC-V Industry Forum held in Shanghai’s Lingang area on August 13th, Wei Min Dai—Chairman, CEO, and President of VeriSilicon—recalled that the industry used to face the nagging question, “Can you even get the chip out?” Today, that’s a non-issue. The real focus has shifted to how fast these chips get deployed and how robust the ecosystem becomes.
RISC-V is an open instruction set architecture that sits right alongside Arm and x86. Born in a Berkeley lab back in 2010, it quickly caught the industry’s eye and made its name primarily in MCUs and IoT. But now, with chip volumes ramping up, RISC-V is boldly pushing into far more complex arenas—servers, AI, and automotive.
Over the past year alone, heavyweights like Qualcomm and NVIDIA have placed their bets on RISC-V through acquisitions and partnerships. Dai believes RISC-V’s momentum is simply “unstoppable,” and it’s only a matter of time before it forms a three-way standoff with x86 and Arm.
The Dishui Lake Forum, now in its fifth year, has showcased over 40 domestic RISC-V chip designs, with a mass-production rate above 90%. Dai told us that the forum now receives so many chip submissions that organizers have to carefully vet them, sometimes even doing on-site checks to confirm production status. The fact that companies are “fighting to get in” shows just how many Chinese players are shipping real, scalable products.
The market size is exploding too. Citing SHD Group data, Dai noted that global RISC-V SoC shipments hit roughly 6.9 billion units in 2025, and are projected to surge to about 36 billion by 2031—a jaw-dropping 31.7% compound annual growth rate. China’s market alone is expected to balloon from around $24 billion in 2025 to nearly $86.4 billion by 2031. Rong Yi, Secretary-General of the Shanghai Integrated Circuit Industry Association, also cited Wedbush Securities’ assessment at the forum: RISC-V’s market penetration has reached 25%, meaning one in every four chips globally now uses the architecture.
RISC-V’s tentacles are reaching far beyond its traditional turf. Hu Zhenbo, founder of Nuclei System Technology, sees the embedded space as already “a hundred flowers blooming.” Mobile application processors have a solid foundation, the PC sector is still relatively quiet, but the server market is showing serious muscle.
Here’s the thing: RISC-V used to be the go-to for cost-sensitive, low-performance applications that valued customization. Now, it’s stepping up to much steeper challenges.
Among the 10 chips showcased this year, server-oriented progress is drawing major attention. Blue Chip Computing’s RISC-V+AI converged server CPU, the LX500, has already found its way into mass-produced servers. In July, the company teamed up with Lenovo to roll out RISC-V servers, and initial deployments are quietly happening in data centers run by telecom operators and state-owned enterprises. Meanwhile, SpacemiT’s server chip V100 has completed tape-out, and LinkRay Intelligence has released its P100 aimed squarely at servers.
Dai breaks RISC-V’s market strategy into three fronts: “up top” is the high-end data center market, “down below” is the fragmented IoT space, and “off to the side” are edge scenarios like industrial and autonomous driving. His takeaway? “The high-end is advancing faster than we ever expected.”
AI is pouring gasoline on this fire. As AI computing demand skyrockets, more devices—robots, cars, cameras—need to handle inference, perception, and control locally. This gives RISC-V a golden opportunity to sidestep traditional competition and overtake on the curves.
Take SpacemiT’s K3, which is already powering robot applications. Or TPS’s chip that packs perception, control, and navigation into a single piece of silicon. And ESWIN’s automotive-grade electronic rearview mirror chip, the EAI8800, has slashed end-to-end latency to under 20 milliseconds.
As for which scenarios will hit critical mass first, the industry doesn’t have a unanimous answer. Li Huaqing, VP of LinkRay Intelligence, believes smart security—where algorithms are already pretty mature—might ramp up sooner, while robotics is the longer-term play. Wang Zhiwei, a partner at华山资本, is more cautious about an AI glasses boom in the short term, suggesting robots will likely infiltrate from industrial settings first. Dai, for his part, warns that trying to be everywhere at once could backfire—the key is picking your battles and deliberately steering toward the paths of least resistance.
But as RISC-V climbs toward the high end, the challenges it faces are evolving too.
Dai is quick to emphasize that RISC-V is “open,” not “open source.” China’s RISC-V industry has already formed a patent alliance, growing from 9 members in 2023 to 24 today, with a pool of 161 patents. His reasoning? An open instruction set is like handing everyone a free dictionary—you can write whatever you want, but you can still commit plagiarism. Consolidating patents is about building a defensive shield so member companies aren’t left exposed when intellectual property disputes rear their ugly heads.
But patents are just the tip of the iceberg. Software is the real mountain to climb. Dai points to ARM’s own history—when it moved upmarket, it faced the same ecosystem pressure. When RISC-V enters the server and AI markets, the ability for developers to quickly compile, debug, and deploy on these chips becomes absolutely critical. A mature architecture ecosystem requires operating systems, development tools, software applications, and a vibrant developer community all pulling in the same direction.
Shanghai, home to VeriSilicon, is currently the city throwing the most weight behind RISC-V chip development. In 2018, the China RISC-V Industry Alliance was founded here. In 2024, the Shanghai Open Processor Industry Innovation Center was unveiled in Lingang. And with five consecutive editions of the Dishui Lake Forum, chip designers, upstream and downstream players, and application developers keep converging in this city to hash things out.
In Dai’s view, the ecosystem problem will eventually solve itself. There’s no fundamental bottleneck blocking China’s RISC-V development—”it’s just a matter of time.”
