Pages

13 November 2023

Size isn’t everything: China’s new chip is less earthshaking than you may have heard

 MATT BRAZIL and PETER W. SINGER

When Huawei released its new Mate 60 Pro phone earlier this year, it set off alarm bells in Western media, which fixated on the ultra-tiny 7-nanometer circuits in the phone’s Chinese-made chipset. Just a year earlier, the U.S. had imposed export controls meant to keep China from obtaining circuits of 16 nanometers or smaller. The announcement seemed to demonstrate not just China’s technical prowess but also the failure of a signature element of Western technology policy in its competition with China, that has become all the more timely with new Biden administration new chip controls and AI policy efforts, seeking to block American chipmakers from selling semiconductors to China that circumvent government restrictions.

Typical headlines were “New phone sparks worry China has found a way around U.S. tech limits” (Washington Post) and “Is Huawei's new phone proof China is gaining ground in the chip wars?” (ABC). Not so coincidentally, the news broke during the visit to Beijing of U.S. Commerce Secretary Gina Raimondo, whose Department’s Bureau of Industry and Security administers those export controls. A Chinese state media editorial crowed, “While the tech war is ongoing and the U.S. maintains an advantage in the high-tech sector, the momentum of the Chinese people's determination to catch up despite the pressure, and a strong sense of moral conviction, are something that the U.S. cannot match. This also determines that while the US' suppression might bring us some troubles in the short term, in the long run, these ‘wars’ are not ultimately in their favor.”

Yet, as in so many issues of technology and China, the story is more complex.

To be sure, the Kirin 9000s—the system on a chip at the heart of the new phone—is in several ways a triumph for Chinese technology. It was designed by HiSilicon, the circuit-design subsidiary of Huawei, one of just four companies in the world that can design a globally competitive system-on-a-chip that incorporates all the key functions of a smartphone. (The others are America’s Qualcomm, Taiwan’s MediaTek, and South Korea’s Samsung.)

The chip was fabricated by China’s Semiconductor Manufacturing International Corporation, or SMIC, a silicon foundry. (The Commerce Department says SMIC makes chips for the Chinese military, a charge the company denies.) SMIC produces the Kirin 9000s using Deep Ultraviolet lithography equipment acquired before the latest American sanctions. Huawei’s claim to have produced 7nm circuits was verified by the Canadian firm TechInsights and by Tilly Zhang of Gavekal Dragonomics.

Speaking on the ChinaTalk podcast, Dylan Patel of SemiAnalysis and Doug O’Laughlin of Fabricated Knowledge argued that SMIC’s successes are due to inadequate U.S. controls, China’s hitherto underestimated chip design and semiconductor fabrication skills, and the PRC’s technology diverters at the Ministry of State Security and elsewhere.

But this may not be the whole picture.

To begin, the DUV process “success rate”—that is, the percentage of chips coming off the manufacturing line that actually work as designed—may or may not be up to the industry norm. Neither SMIC nor Huawei are releasing much data about the new chip or its production process. As Gavekal’s Zhang has noted, SMIC may have pushed the DUV process as far as it will go, reaping only 40 percent or fewer usable chips. Huawei's competitors, such as Taiwan Semiconductor Manufacturing Co., have been producing 7nm chips for over five years; their success rate generally tops 90 percent.

Secondly, the CCP government has given crucial financial support to what is supposed to be a commercial product. Zhang wrote that Huawei has “demonstrated that they are willing to accept much higher costs than are normally considered worthwhile…it is only the combination of Huawei's own large financial resources and generous government subsidies that could allow it to sell phones using these chips at normal market prices.” Still, the government is pressing on with these subsidies, having just invested $2 billion in Changxin Xinqiao, a memory chip start-up, and now seeking to launch a $44 billion state fund to boost domestic chipmaking, which would be the third such fund.

Finally, China may be unable to produce advanced sub-10nm chips at scale, at least for now. For example, SMIC’s Shanghai fab has been turning out less-advanced 14nm chips since late 2019. But, as of a year ago, the actual volumes were so small that the company ceased to report the contribution of the product to its revenue, according to Anton Shilov, who reports on China for the authoritative online technology publication Tom’s Hardware.

In short, Huawei’s new mobile phone shows that SMIC has produced at least some 7 nm chips using repetitive DUV immersion. But it remains unclear whether they have reached, or even can reach any time soon, the high success rates needed for commercially viable production, at least comparable to its peers.

These questions about the chip and SMIC’s larger ability to surpass U.S. controls thus remain unanswered for now. But clues will emerge by February, when the gift-giving season of Chinese New Year will likely cause demand for Huawei’s new phone to grow “faster than bamboo shoots after a spring rain,” as the saying goes.

What is clear, however, is the information offensive that the news serviced. Huawei’s announcement of 7nm chips was intended to bolster a narrative: that China will advance in the tech sphere despite what Chinese commentators often call Washington’s “cold war mentality.” The timing was chosen to call into question American and allied export controls. But until SMIC shows it can produce the chips at international levels of quality and scale, the message lacks evidence to back it.

And the CCP’s trumpeting of the news reveals an unintended message, akin to Shakespeare’s famous line in Hamlet: "The lady doth protest too much, methinks.” Beijing’s continued complaining about America’s export controls (which actually date to the Tiananmen Square period of 1989, so are not a new “cold war mentality”), illustrates how these controls continue to have an adverse effect. Otherwise, they would not be such a central point of regime leader and media discourse. This is much in line with the history of export controls, going back to the actual Cold War. U.S. controls hindered the Soviet bloc’s acquisition of technology, increasing expenses and reducing quality and after-sales service. In short: just because a product can finally be made doesn’t mean the policy to limit it has been ineffective.

No comments:

Post a Comment