Strategic Logic Behind China’s RISC-V Push

China’s move to issue a national policy document encouraging the use of open-source RISC-V chips, as reported by Reuters on March 4, 2025, is a clear signal of intent. For years, China’s semiconductor industry has been hamstrung by its reliance on Western-controlled instruction set architectures (ISAs) like x86 (dominated by Intel and AMD) and ARM (owned by SoftBank but subject to U.S. export controls). The U.S. technological blockade—exemplified by restrictions on Huawei, Phytium, and Nvidia’s GPU supply—has exposed the vulnerability of this dependency. RISC-V, being open-source and geopolitically neutral, offers China a way to bypass these chokeholds.

The open-source nature of RISC-V is a game-changer. Unlike x86, which is a closed ecosystem, or ARM, which requires hefty licensing fees and carries risks of supply chain disruptions, RISC-V allows Chinese firms to design chips without paying royalties or worrying about foreign patent barriers. This autonomy is critical in the context of AI, where customized hardware is increasingly vital for optimizing performance and energy efficiency. Alibaba’s Xuantie team achieving a 30% energy efficiency boost with RISC-V-based AI server chips is a concrete example of how this architecture can deliver tangible benefits.

RISC-V in the AI Era

The rise of AI has shifted the semiconductor landscape, and RISC-V’s flexibility makes it particularly well-suited for this new paradigm. Traditional ISAs like x86 were designed for general-purpose computing, resulting in bloated architectures that aren’t always efficient for AI workloads. ARM, while more adaptable, still operates within a proprietary framework. RISC-V’s modular design, however, allows developers to strip out unnecessary instructions and add custom extensions tailored to specific AI tasks—think edge computing, autonomous driving, or industrial robotics.

The DeepSeek model’s success on RISC-V, as highlighted by Li Chunqiang from Alibaba DAMO Academy, underscores this potential. By leveraging Mixture of Experts (MOE) technology to reduce computational demands, DeepSeek demonstrates that RISC-V chips don’t need to match Nvidia’s raw horsepower to be commercially viable. A RISC-V solution that’s 30% as powerful as an Nvidia GPU but costs significantly less could democratize AI hardware, especially for smaller companies or niche applications. This aligns with China’s fragmented but vast AI market, where cost and customization often trump brute force.

Complementary Ecosystems: RISC-V, LoongArch, and SW64

Your point about China balancing RISC-V with its own closed-source ISAs like LoongArch (Loongson) and SW64 (Sunway) is spot-on. This isn’t an either/or strategy but a layered one. RISC-V’s openness and low cost make it ideal for widespread commercial adoption and ecosystem-building, while LoongArch and SW64 cater to high-security, high-performance niches like national defense and supercomputing. For instance, Loongson powers strategic assets like Beidou satellites, and SW64 underpinned the Shenwei Taihu Light supercomputer. These closed systems ensure China retains control over critical infrastructure, while RISC-V helps expand the broader industrial base.

This dual-track approach mitigates risk. If RISC-V faces international pushback (e.g., U.S. efforts to limit China’s role in standard-setting), China can fall back on its proprietary ISAs. Meanwhile, RISC-V’s growth could indirectly bolster Loongson and SW64 by fostering a more robust domestic chip ecosystem—think shared talent, tools, and manufacturing capacity.

Challenges and Risks

Despite its promise, RISC-V isn’t without hurdles. The software ecosystem remains a weak link—ARM’s 20 million developers dwarf RISC-V’s community, and its tool chain is only 60% complete. However, China’s proactive response, like Huawei’s HarmonyOS (Hongmeng) fully adapting to RISC-V and Alibaba’s “Swordless Alliance” integrating EDA tools and compilers, shows a concerted effort to close this gap. This vertical integration could accelerate RISC-V’s maturity, especially within China’s borders.

Geopolitics is another wildcard. The U.S. has already signaled unease about RISC-V’s rise, with 2023 legislative efforts to restrict its export. If the U.S. successfully excludes China from RISC-V’s global standards bodies, it could slow China’s momentum. Yet, RISC-V’s open-source nature makes it hard to fully contain—China can fork the architecture and build its own ecosystem if needed, much like it has with other technologies.

Finally, in high-performance computing (HPC), RISC-V lags behind. Its energy efficiency at sub-7nm processes and single-chip compute density trail AMD’s MI300X by about two generations. For now, RISC-V shines in edge and mid-tier applications rather than HPC, but that could change with time and investment.

The Bigger Picture

China’s RISC-V strategy is about more than just chips—it’s about reshaping the global tech order. By championing an open-source architecture, China positions itself as a leader in the “de-Westernization” of tech infrastructure, appealing to other nations wary of U.S. dominance. If RISC-V captures 25% of key markets like PCs, autonomous driving, and servers, as Ni Guangnan predicts, it could challenge the x86-ARM duopoly and erode the leverage of Western giants.

This isn’t just a defensive play against U.S. sanctions; it’s an offensive move to seize discourse power in the AI era. Horizon Robotics’ X3M chip outperforming Nvidia’s Orin in power efficiency for autonomous driving shows how RISC-V can carve out niches where Western solutions falter. Over time, this “scenario-defined” approach could fragment Nvidia’s dominance, especially in emerging markets.

Final Thoughts

March 4, 2025, might be remembered as a turning point. China’s RISC-V policy isn’t just a reaction to U.S. pressure—it’s a bold bid to redefine the semiconductor landscape. By blending open-source innovation with national strategy, China is betting that RISC-V can be the backbone of its AI-driven future, while complementary ISAs like LoongArch and SW64 secure its sovereignty. The stakes are high, and the outcome will ripple across the global tech industry for decades.

What do you think—will RISC-V live up to this hype, or are the ecosystem and geopolitical challenges too steep to overcome?

Comments