Pages

27 December 2020

China’s Drive to Make Semiconductor Chips Is Failing


BY SALVATORE BABONES 

Shenzhen, a city of some 12 million people in southeastern China’s Guangdong province, is the consumer electronics capital of the world. Immediately abutting Hong Kong, it now towers over its restive regional rival in terms of population, skyscrapers, and by some counts even gross domestic product. It is also home to Huawei, the Chinese telecommunications company that dominates global 5G wireless infrastructure and sits at the center of the U.S.-China tech war.

Long a hub for mobile phone assembly, the city of Shenzhen is about to get into the business of making phones itself. In November, a consortium led by the Shenzhen municipal government struck an unusual deal to pay Huawei $15 billion and take over the company’s Honor budget smartphone brand. Huawei is fighting for its very survival since it was added to both the U.S. Commerce Department’s export licensing Entity List and the U.S. Defense Department’s foreign investment blacklist.

The strange spectacle of a city government funneling money into a global tech giant and ending up with a budget phone maker is emblematic of China’s problems in developing its own technologies. China has the ambition, and it can do things at scale. It can also raise the money, even (when necessary) from unlikely sources. But it lacks the broad ecosystem of commercial cooperation, intellectual property protection, and intelligent venture capital that makes deep technology collaboration possible. China’s command economy is a cookie-cutter economy, but high technology is a networking game.

The fact that the U.S. government could so easily hobble the world’s largest smartphone maker and 5G infrastructure supplier in less than one year is indicative of the fragility of China’s highly centralized high tech sector. China’s electronics industry relies on U.S., Taiwanese, South Korean, and Japanese suppliers for many key components, but the most strategic of strategic technologies is the microprocessor. And despite years of strategic investment, China has (so far) been unable to master the production of these highly specialized but utterly ubiquitous computer chips.
China’s electronics industry relies on U.S., Taiwanese, South Korean, and Japanese suppliers for many key components.

Silicon-based semiconductors are used in all sorts of computer chips, including memory chips, sensor chips, and a variety of other microprocessor chips. Most people are familiar with the general-purpose microprocessors known as central processing units (CPUs) that power smartphone and computer operating systems, but device performance increasingly depends on more specialized microprocessors such as graphics processing units (GPUs) and artificial intelligence (AI) accelerators. The most advanced CPUs now coming to market, such as the Apple A14 Bionic, the Qualcomm Snapdragon 888, and the Samsung Exynos 1080, include integrated GPUs and AI accelerators right on the chip—and have 5G wireless integration to boot.

The only serious Chinese rival to these advanced U.S. chips is the HiSilicon Kirin 9000, designed by Huawei’s own in-house “fabless” chip-design subsidiary. In the arcane lingo of semiconductor manufacturing, a fabless chipmaker is one that lacks its own fabrication facilities, known as “fabs” or “foundries.” Until this year, Huawei’s HiSilicon chips were actually made by Taiwan Semiconductor Manufacturing Company, but tightening U.S. sanctions put an end to that. Broader U.S. export controls on chip design software and foundry machine tools mean that Huawei now has little chance of developing an advanced fabrication capability of its own. As a result, the Kirin 9000 is effectively stillborn.

China’s most advanced chip foundry is Semiconductor Manufacturing International Corporation (SMIC), based in Shanghai. Like Huawei, SMIC is on both the U.S. Commerce and Defense departments’ watchlists, severely restricting its access to U.S. technology and finance. Without foreign help, SMIC is generations away from being able to produce a chip like the Kirin 9000. Like all of today’s flagship CPUs, the Kirin 9000 is designed for 5-nanometer silicon wafers. When it comes to semiconductors, thinner is better, and the best SMIC can currently manage is 14 nanometers. It has announced plans to produce 7-nanometer chips, but lacks the machine tools to make them.

If China has been unable to match its international competitors on microprocessors, it’s not for want of trying—or spending. China established a $22 billion National Integrated Circuit Industry Investment Fund in 2014 (known as the Big Fund) in a bid to reduce its reliance on imported chips, but to little avail. Today, only 16 percent of China’s semiconductors are made locally, and these tend to be the least sophisticated in every category. Last year, China announced a second Big Fund to invest a further $29 billion in semiconductor development. It remains to be seen whether or not China can make a success of it the second time around, this time in the face of aggressive U.S. sanctions.
China’s chipmakers and designers now seem to be short on cash.

Despite all the promised investment, China’s chipmakers and designers now seem to be short on cash. Huawei going cap in hand to the Shenzhen government is no surprise, given its high exposure to the U.S. and Indian markets. But other Chinese chipmakers with little connection to the United States are also facing financial difficulties. Wuhan’s Hongxin Semiconductor Manufacturing Company had promised to build China’s first 7-nanometer chip fab, but ran out of money in August. It has since been taken over by the municipal district government—effectively a bailout.

Also based in coronavirus-hit Wuhan, memory-chip maker Yangtze Memory Technologies announced plans in September to build a world-class $22 billion flash memory chip foundry. Two months later, its parent company, Tsinghua Unigroup, defaulted on a $198 million bond repayment. In addition to the now-stalled memory chip project, Tsinghua Unigroup owns a chip design firm and cloud computing platform, among other subsidiaries. A spinoff from the prestigious Tsinghua University (Chinese President Xi Jinping’s alma mater), it would once have been thought politically immune from failure.

Yet it is not the first prestigious university-linked spinoff to default on its debt. Peking University Founder Group, a diversified conglomerate, missed a bond payment last December, even before the coronavirus crisis hit. If China is allowing such well-connected firms as semiconductor foundries and university research groups to go bust, financial conditions in the country must be much more dire than its announcements of multibillion-dollar investment funds would suggest. So dire, in fact, that the governor of the People’s Bank of China, Yi Gang, felt compelled to publicly warn local governments last week not to expect bailouts for bad business decisions—such as taking over struggling companies or subsidiaries with uncertain futures.

China made chipmaking its top civilian technology priority of the last decade, but it has little to show for it. Even before many of its leading companies were hit with U.S. export and financial controls, China proved itself unable to establish a competitive presence in the market for relatively simple memory chips—never mind complex microprocessors. As the cutting edge of chip design goes even farther down the path of multiple integration, with CPUs, GPUs, AI accelerators, and wireless modems all printed on a single wafer a few molecules wide, China will find it ever more difficult to catch up.

No comments:

Post a Comment