Policymakers increasingly worry that Chinese firms will use the RISC-V architecture to circumvent U.S. export controls or de-risk against potential future controls and are therefore looking for ways to regulate China’s access to the technology. But regulating RISC-V is legally tenuous and potentially counterproductive.
Since 2022, the U.S. government has sought ways to slow down China’s AI development by denying the country cutting-edge chips needed to train and deploy those systems. In October 2022, the Commerce Department’s Bureau of Industry and Security (BIS) imposed controls on the export to China of AI chips and the equipment used to make them, as well as the servicing of that equipment by U.S. persons. A year later, in October 2023, BIS further tightened certain dimensions of the controls to improve their collective effectiveness. These controls have been primarily focused on the hardware, and do not currently control chip architectures, except in narrow circumstances. Then, in November, members of the U.S. House of Representatives’ Select Committee on the Chinese Communist Party (CCP) wrote to the Secretary of Commerce, expressing concern that RISC-V could undermine U.S. export controls by facilitating the development of China’s chip design ecosystem. To understand the contours of this argument, we first need to understand instruction set architectures (ISAs) such as RISC-V.
What is RISC-V?
At the foundation of every processor’s design is a technical specification known as an instruction set architecture (ISA). The ISA describes how software will control the processor’s hardware. Specialized AI chips—for example those designed by AMD and NVIDIA—typically rely on an in-house ISA. For general computing (i.e. CPUs), however, rather than reinvent the wheel, chip design teams typically license an existing ISA from Intel or Arm (a British company). For example, for their respective CPU designs, AMD licenses from Intel and Apple licenses from Arm. Increasingly, rather than licensing an ISA from Intel or Arm, CPU design shops are considering a third ISA option: RISC-V.
RISC-V is an open-source ISA managed by a Swiss standards development body (RISC-V International). Chip design teams can access and implement the RISC-V standard free of charge, avoiding costly licensing fees they would otherwise pay to Intel or Arm. Companies around the world—including American and Chinese firms—are taking advantage of RISC-V to design chips of varying complexity from simple microcontrollers to complex systems-on-a-chip. RISC-V is a particularly attractive option for Chinese chip designers as they look for ways to reduce their reliance on foreign semiconductor technology. However, given that the recent export controls are focused on AI chips (which typically use in-house ISAs), it’s not immediately obvious how a general computing ISA like RISC-V could hinder U.S. export controls; in short, RISC-V doesn’t present a threat, yet.
Why Policymakers are Concerned about RISC-V
In the future, the U.S. may control general computing processors, and RISC-V could reduce the effectiveness of that control. Currently, only the most advanced, specialized AI chips for data centers are subject to U.S. controls, but BIS does not intend to ever loosen its control threshold. Eventually, even the performance of general purpose chips (CPUs) built into our phones and computers will likely exceed the threshold and require a license for export to China. In that future, the prevalence of an open source ISA like RISC-V could aid China’s design of controlled CPUs, limiting the effectiveness of U.S. export controls.
For this reason, the signatories of the Congressional letter view China’s involvement in the development and adoption of RISC-V as a threat to U.S. national security that warrants a response. Export controls targeting standards-setting activity for RISC-V are likely the only regulatory tools that have a chance to slow down China’s development and adoption of the technology.
Limited Regulatory Options to Control RISC-V
Controlling China’s access to RISC-V technology, however, is easier said than done. RISC-V International moved its headquarters from the United States to Switzerland in March 2020 in part to insulate itself from the creeping influence of geopolitics on the chip industry. This move severely limits the U.S. government’s regulatory options. Since RISC-V International operates outside the United States, BIS can’t restrict the publication of RISC-V standards or Chinese companies’ involvement in the process. BIS’s only clear option is to restrict the involvement of U.S. firms in RISC-V standards development (via controls on the activities of U.S. persons), but as this type of regulation often has unintended consequences, it’s critical to think through the potential costs and benefits.
Controlling U.S. involvement in RISC-V standards development would benefit U.S. economic security if it slows China’s development of RISC-V chip designs. However, given the importance of RISC-V to China’s semiconductor self-reliance efforts, even if U.S. firms were forced out of participating in standards development at RISC-V International, it’s likely that standards development efforts would continue, and it is unclear whether China’s RISC-V design efforts would be significantly hindered.
Meanwhile, controlling U.S. firms’ activities at RISC-V International risks ceding tech standards leadership to China. The U.S. government recently reaffirmed that leadership in standards development drives leadership in technology development. A similar situation played out in 2019 when export controls on Huawei forced U.S. firms out of critical telecom standards-setting bodies—temporarily abdicating technology leadership. In response, BIS issued and then expanded a standards-setting carve-out to the control. Restricting U.S. firms from participating in RISC-V standards would undermine that effort—likely hurting U.S. standards leadership in RISC-V and beyond.
In addition, locking U.S. chip design firms out of RISC-V standards development may lead firms to move overseas or to adopt less efficient technologies. Startups and hyperscalers alike see significant value in participating in RISC-V standards development. If the U.S. government prevents startups from participating, they may move overseas (or incorporate elsewhere to begin with) to avoid U.S. regulations. Furthermore, if larger chip design firms that would have bet on RISC-V in the long-term are restricted from influencing the standard’s development, they may return to a proprietary ISA, pushing these firms down a suboptimal technological trajectory and reducing their competitiveness with Chinese chip designers. In the long term, controlling U.S. involvement in RISC-V standards development could counterintuitively give Chinese chip designers an advantage over their U.S. counterparts.
A Way Forward
Given the limitations of controlling RISC-V development, the U.S. government could instead choose a different path: supporting, rather than controlling, the development and adoption of this new technology. If RISC-V truly does present an opportunity for Chinese design firms to leapfrog their U.S. competitors—which is not guaranteed, it stands to reason that encouraging U.S. design firms to adopt RISC-V architectures could mitigate that threat and facilitate innovation at home. Similarly, supporting U.S. firms’ participation in RISC-V standards-setting could help alleviate concerns that China will steer the technology’s development trajectory or introduce vulnerabilities.
Thanks to Chris Rohlf and Hanna Dohmen for their feedback on this article.