Open-source chip design spurring on AI

Open-source chip design architecture RISC-V is gaining in popularity in China, experts said, as domestic tech companies such as Alibaba Group Holding Ltd work to sharpen their technological strengths in semiconductors amid the rapidly evolving era of artificial intelligence.

RISC-V technology manages memory systems shared between CPUs, and has been open and free for use since its debut in 2010. Developers can use it to design a chip tailored to their unique needs.

Chip designs have traditionally been expensive to license. Global chip designers, such as Intel and ARM, have kept their blueprints secret. This meant consumers had to buy manufactured chips directly, or pay more for customized designs.

Amid the increasingly tighter restrictions on US chip exports to China, the rising popularity of open-source RISC-V will help cut Chinese companies' future reliance on chip design architecture such as Intel's x86, and products developed by the United Kingdom's ARM Holdings, the experts said.

Ni Guangnan, an academician at the Chinese Academy of Engineering, said, "China's open-source ecosystem has flourished, becoming a significant driver of technological innovation."

Ni expects that by 2030, RISC-V chips will capture over 25 percent of the market in consumer PCs, autonomous driving, network communications, industrial control, smart devices and high-performance server applications, becoming a cornerstone of the global semiconductor industry.

"Open-source RISC-V is not just a technological innovation, but also a global transformation that will influence the future global computing landscape," Ni added.

This is especially true in the era of AI. As AI technology advances, traditional computing architecture often struggles to meet the needs of complex AI tasks, which require higher performance, lower power consumption and enhanced parallel computing capabilities, the experts said.

RISC-V, with its scalability, flexibility and customizable characteristics, is uniquely positioned to address these needs. By extending its instruction set, RISC-V can support AI-specific computing requirements and improve parallel computing capabilities, making it an ideal choice for AI architecture and offering a new pathway for future AI chip development, they added.

According to a report from market research company Omdia, RISC-V-based processor shipments are expected to grow at an annual rate of nearly 50 percent from 2024 to 2030, reaching 17 billion units globally by 2030. And RISC-V processors are projected to account for nearly a quarter of the global market share by then.

Guo Songliu, head of RISC-V industry ecosystems at the Institute of Software at the Chinese Academy of Sciences, said, "AI is evolving rapidly, and RISC-V, as the most flexible and open of the three mainstream instruction set architectures, is undoubtedly best suited for the pace of innovation in the AI era."

To grasp the opportunity, Damo Academy, a research institute of Alibaba, has been innovating in high-performance chips for AI. At a conference held in late February, Damo Academy's Xuantie processor family announced that its highest-performance processor for computer servers, the C930, will begin delivery in March.

According to data released by Damo, as of early 2024, the Xuantie series had over 300 licensed customers, more than 800 authorized licenses and cumulative shipments exceeding 4 billion units.

Alibaba is not alone. In China, hundreds of companies are already engaged in, or exploring, RISC-V development. Among the 23 premier members of RISC-V International — an international organization dedicated to promoting the use of the chip architecture — Chinese companies and institutions account for half, including Alibaba, Huawei, ZTE and Tencent. Notably, premier members hold board and technical committee seats, enabling them to directly influence the development of RISC-V standards and technical directions.

In 2022, 10 billion RISC-V chips were produced globally, of which half were made in China. As Chinese companies have scrambled to embrace RISC-V in recent years, the proportion has risen even higher, though no official updated figure is available yet, experts added.

Meng Jianyi, chief scientist at Damo, said, "For China's chip industry to break into the high-performance chip market and challenge the dominance of x86 and ARM by foreign chip manufacturers, it must leverage open-source RISC-V."

Since 2018, Alibaba has been investing in RISC-V architecture, making it one of the earliest Chinese teams to explore the technology. In 2019, the launch of the Xuantie C910 processor marked the beginning of RISC-V's transition from applications for the internet of things market to the high-performance chips market, sparking a wave of industry interest in RISC-V.

Over the past six years, Damo's Xuantie series has released 13 RISC-V processors across four series, covering high-performance, high-efficiency and low-power scenarios.

While advancing its own hardware and software technologies, Damo is also driving collaboration across the industry chain to build an ecosystem for RISC-V. Applications of RISC-V are rapidly emerging, such as AI-enabled personal computers and intelligent robots, as shown by Damo Academy's partners at the February conference.