Nvidia's Vera CPU Spurs China's RISC-V Challenge in High-Performance Computing

Nvidia's newly released Vera CPU, designed for artificial intelligence data centers, is set to enter China's procurement market as early as August, priced above $20,000 per chip with complete 256-chip cabinets reaching approximately $10 million. This development underscores a critical infrastructure challenge for China: as AI demands skyrocket, domestic technology leaders face a strategic question about whether to remain dependent on ARM-based architectures or pursue an alternative path through RISC-V.

Vera represents Nvidia's first standalone CPU optimized for agentic AI, based on ARM technology and claimed to deliver 1.8 times faster performance than competing processors. Nvidia CEO Jensen Huang has acknowledged that advanced chip export controls have substantially reduced the company's China market share, making the Vera launch both a commercial maneuver and an attempt to reclaim lost ground. Despite this, Chinese industry observers note that Vera's ordering logistics, supply stability, and pricing negotiations remain controlled by external actors, raising fundamental questions about technological autonomy.

In response, Chinese technology leaders are increasingly focusing on RISC-V, an open-source instruction set architecture that originated over a decade ago and has already shipped over 10 billion units in embedded systems. What distinguishes RISC-V's current momentum is its application to high-performance computing domains—data centers, servers, and AI infrastructure—where it previously held no foothold. Industry consensus frames this as China's opportunity to escape the "impossible triangle" of CPU development: the traditional trade-off between prosperity, controllability, and autonomy. X86 offers prosperity but lacks controllability; ARM provides some control through licensing but sacrifices true autonomy; proprietary domestic architectures achieve autonomy and control but remain isolated in small ecosystems.

RISC-V's structural advantages lie in its open-source nature, modular design philosophy, and international standards compliance, theoretically enabling simultaneous achievement of all three objectives. Multiple motivations propel China's RISC-V momentum: an anticipated explosion of AI-driven computing demand expected to account for over 70 percent of global semiconductor spending by 2030; export restrictions tightening access to advanced foreign chips; the cost-reduction potential of open-source architectures eliminating licensing fees; and coordinated government support combining national-level industrial policy with local implementation incentives. Notably, six major domestic CPU manufacturers, who previously dismissed RISC-V advocates as "disorganized crowds," have now completed national testing and several are actively developing next-generation chips based on RISC-V architecture.

Technical progress validates commercial viability. Chinese teams have achieved SPEC CPU benchmark scores reaching 14.78 to 18 points at unit frequency, crossing the critical 15-point threshold considered essential for high-performance computing credibility. Multiple domestic vendors claim to have delivered RISC-V cores exceeding 15 points, with some achieving clock speeds above 3.4 GHz. However, technical benchmarks represent only an entry ticket; the real competition lies in developing complete computing subsystems integrating cores with memory, interconnects, and software ecosystems—requiring engineering complexity equivalent to the leap from single components to integrated platforms.