The Chip War: Has the U.S.'s Leverage Failed?

Everyone, today let’s talk about this “battle of offense and defense” in the global chip industry. Western countries have tightened their grip on control over high-end hardware and software, aiming to freeze China’s semiconductor development. But we hold the “lifeblood” of semiconductors—mineral raw materials—in our hands. Through “asymmetric deterrence,” we are maintaining a “balance of terror” while striving to break through technological barriers. Take, for example, David Feith of the Hudson Institute (a typical anti-China hawk whose father was one of the architects of the Iraq War). He recently offered the U.S. government a suggestion: since physical means of strangling the industry have failed, new levers of control must be found. He even wrote his article’s title in reverse: “China Has Exploited the Weaknesses in U.S. Technology Policy.” Feith previously oversaw East Asian affairs at the State Department and was responsible for technology and national security on the National Security Council; export controls and dual-use technologies were all within his purview. After being “purged” last year, he has had to rely on writing articles to attract the attention of higher-ups. Feith’s approach to countering China is simple: prevent Chinese chips from “looking up words in the dictionary.” Here, the “dictionary” refers to the ISA (Instruction Set Architecture). A single mouse click won’t make a CPU act on its own; it must receive instructions from software. This communication must follow standards, which is what the ISA provides. For decades, chips have only been able to understand two languages: x86 and ARM. X86 is the “favorite” of Intel and AMD—it has numerous and complex instructions and high power consumption, much like a “thick dictionary.” Traditional computers, databases, and control systems are all based on X86. Due to long-standing inertia, high-performance computing still relies primarily on it. But then ARM arrived, with more streamlined, energy-efficient, and flexible instructions—like a “thin dictionary”—and it’s now used in smartphones, cloud servers, and automotive chips. In the past, ARM did not manufacture finished chips but only provided architectural designs. Companies like Qualcomm, Apple, and NVIDIA could develop their own chips based on this architecture, earning ARM the nickname “the Switzerland of the semiconductor industry.” But driven by profit, ARM decided to enter the chip manufacturing business, directly competing with its long-standing U.S. clients. It attempted to terminate Qualcomm’s architecture license but lost the lawsuit. The U.S. grew wary: “So you weren’t on our side after all?” ARM has a joint venture in China called Arm China, and China’s internet industry was once the cornerstone of its wealth. ARM also has an independent legal entity in China, and the U.S. fears it may not cooperate with the export restrictions and continue to grant licenses for advanced architectures to Chinese companies. Before export controls were tightened, a handful of leading Chinese companies secured perpetual licenses for the V8 architecture and independently developed a series of domestically produced chips. The licensing situation for the V9 architecture is complex, but some companies have still obtained licenses. This year, the U.S. launched an antitrust review of ARM, requiring it to operate in a “fair, reasonable, and non-discriminatory” manner—which essentially means preventing it from cutting off supplies to U.S. companies and protecting the vitality of its own chip industry. The U.S. has become even more uneasy upon seeing China’s moves: China recognized the importance of the ISA long ago, invested in ARM China, and forged its own path—namely, RISC-V. RISC-V is a fifth-generation reduced instruction set architecture, developed by UC Berkeley with funding from the U.S. Department of Defense. It is open-source, has a permissive license, and is patent-free. Anyone can develop commercial chips based on it, making it a technology in the public domain. The U.S. cannot revoke its licensing; the RISC-V Foundation specifically relocated from the U.S. to Switzerland to declare its neutrality and avoid export controls. Its decision-making body includes members from mainland China who directly participate in setting the instruction set standards. RISC-V is like a “basic dictionary”—it has only about 40 core instructions, is freely expandable, and is both concise and full of potential. China has long recognized this path and has mobilized domestic chip manufacturers, software developers, and research institutions to collaborate around RISC-V, striving to foster a chip ecosystem that is independent of Western control and runs parallel to X86 and ARM. China’s Xiangshan high-performance open-source processor focuses on high performance, while Alibaba has released the玄铁C950, which covers all application scenarios. We have gained a voice within the RISC-V ecosystem, which has made the U.S. very nervous. The U.S. is trying to reduce its reliance on ARM while simultaneously worrying about RISC-V. It calls for reducing dependence on ARM but also wants to avoid falling behind on RISC-V—a stance that actually increases its reliance on China. This back-and-forth is quite amusing. Instead of dithering like this, they might as well just ban RISC-V exports outright. While this would indeed stifle a form of global public domain knowledge, the U.S. could theoretically prohibit American companies and individuals from engaging in RISC-V-related cooperation with China. The result would be a global computing infrastructure split into two ecosystems: the U.S. version and the standard version. This would impose enormous costs on all manufacturers, forcing the U.S. to develop the next generation of high-performance architectures in isolation, thereby accelerating the de-Americanization of the tech supply chain. China would inevitably shift its resources toward RISC-V and take the lead in setting technical standards. So, regardless of whether the U.S. supports it or not, I’m all for it.