Research Director, CPA
The decision last month by microchip manufacturer GlobalFoundries to build a chipmaking plant in China reinforced fears in the U.S. technology industry over China’s ambitious plans to become a major player in the world semiconductor industry.
While GlobalFoundries is a second-tier chipmaker and the technology destined for the China plant is not world-class, the decision is nevertheless an illustration of the growing clout China exerts across the entire technology industry as it grows inexorably towards becoming a self-sufficient technology leader and exporter. China’s strategy for reaching this goal includes conventional tactics like huge investment dollars but also unconventional, even illegal, tactics (under WTO rules) like forced technology transfer and theft of intellectual property.
GlobalFoundries is investing some $10 billion to build a fab (as chip fabrication facilities are known) in the central Chinese city of Chengdu. GF is getting support from the Chengdu municipal government, but has not revealed the level of Chinese government support. GlobalFoundries also operates two chip fabs in upstate New York (acquired from IBM) that provide chips for customers including the U.S. military.
China has made the semiconductor industry a strategic priority in its 13th Five Year Plan and its longer-term “Made in China 2025” plan. The Obama White House was sufficiently worried about China’s semiconductor plans that it published a detailed report on China’s semiconductor strategy in January. The report was surprisingly outspoken, coming from an Administration that was usually a flag-waver for free trade and international cooperation. The report charged that: “Chinese policies are distorting markets in ways that undermine innovation, subtract from U.S. market share, and put U.S. national security at risk.”
The scale of China’s semiconductor ambition is breathtaking. The January report, written by the President’s Council of Advisors on Science and Technology (PCAST), said China expects to spend $150 billion of government money over the next ten years to build up a world-class chip industry. According to press reports, some 20 new chip fabs are today under construction in China. Although the U.S. has 76 chip fabs, most are older technology. A new one is under construction in Utah and Intel recently committed to investing $7 billion to complete its state-of-the-art fab in Chandler, Arizona, but that’s about it for U.S. investment. Not a single silicon-based chip fab has been built in Silicon Valley this century. The PCAST committee includes former CEOs of Intel, Qualcomm, GlobalFoundries, and Applied Materials, as well as Google chairman Eric Schmidt, and several respected academic technologists.
Semiconductors are a strategic industry for a number of reasons. First, they are used in millions of devices, not just the obvious ones like smartphones and computers, but hundreds of consumer and industrial products. Today’s cars contain several dozen chips each, and the number of chips used in the world is only expected to increase. Secondly, chips are an industry where the U.S. leads the world, with companies like Intel, Qualcomm, and Broadcom each leading the world in its sector of the industry. U.S.-based chipmakers hold 50% of a global market worth $339 billion in revenue last year and they accounted for U.S. exports of an estimated $40 billion last year, making it one of America’s strongest export industries. Thirdly, the chip industry provides good jobs for 370,000 Americans, with average annual earnings of $70,660, according to BLS data. Fourthly, and most importantly, semiconductors are the enabling technology for much of the rest of the technology world. The popular catchphrase in the current cycle of venture capital is “software is eating the world,” but software is only eating the world because the speed and power of the each new generation of chip hardware makes more and better software possible. They key insight that made Google possible was (back in 1998) that the power of the then-latest generation of server chips made it possible to index the entire Internet quickly enough to turn Google-style search into a practical consumer service. Google, Facebook, Amazon, the smartphone, and thousands of other applications are only practical because of the power of today’s chips.
The key to successful software design in the constantly-changing tech industry is to design products that take advantage of the upcoming opportunities of semiconductors. It’s the Wayne Gretzky principle: when you sit down to design software, or a smartphone, or indeed an autonomous car, you don’t skate to where the puck is today, you skate to where it’s going. This is why the PCAST committee is so concerned about the Chinese industry. “Chinese industrial policies in this sector, as they are unfolding in practice pose real threats to semiconductor innovation and U.S. national security,” says the report. That’s an understatement. The concern is that not just semiconductors, but the entire U.S. software industry and development of the Internet itself could be hampered if China grabs a dominant position in semiconductors.
It would be one thing if China were winning share in a free market. But that’s not what’s happening. Today, China imports an estimated $100 billion worth of semiconductors a year. One of its objectives it to eliminate that dependence on suppliers that it considers unreliable—like any supplier based in the U.S. The PCAST report details four specific practices Chinese government bodies or companies are using to support their objective. Quoting directly from the report, they are:
- Forcing or encouraging domestic customers to buy only from Chinese semiconductor suppliers;
- Forcing transfer of technology in exchange for access to the Chinese market;
- Theft of intellectual property;
- Collusion. (referring to alleged collusion to lower the price of acquisition targets in the west, to aid Chinese efforts to buy western companies.)
The U.S.-China Commission (USCC), a bipartisan body established by Congress to monitor Chinese economic activities, said in its 2016 Annual Report that another Chinese tactic is trying to buy up western chip companies, to acquire the expertise necessary to succeed in an industry where the microscopic manufacturing techniques are very challenging. “Since 2014, China-headquartered firms have proposed or finalized more than 30 mergers and acquisition deals in the semiconductor industry, totaling nearly $20 billion,” said the USCC report. Some of those deals, like the acquisition of OmniVision and a stake in Marvell Technologies, have been successful. Others, like bids for Micron, Atmel, and Fairchild, failed, either because the U.S. government blocked them on national security grounds or a U.S.-based bidder emerged (probably with the tacit support of the U.S. national security community).
Lattice Semiconductor is a case in point. In 2012, the FBI indicted two Chinese men for trying to buy Lattice chips for illegal export to China. Lattice chips can withstand the high temperatures on spacecraft and export is controlled. Yet in 2016, China’s flagship semiconductor company, Tsinghua Unigroup, was permitted to take an 8.65% stake in Lattice. In another case, in 2014 IBM was permitted to license the technology for its Power8 processor to a little-known Chinese company, Suzhou Technology. Power8 has uses in data centers and supercomputers, and it has potential military applications. Power8’s complexity is clear from the Wikipedia description of its largest version as a chip that consists of 4.2 billion transistors, each 22 nanometers wide, fabricated with 15 metal layers, all on a one inch square substrate.
Game of Risk
The problem is scale—not the size of the chips, which are small, but the size of the production facilities, which are huge. Modern chip fabs cost between $5 billion and $10 billion each. Only large companies can finance that sum of money. With the help of aggressive government support, Taiwan and Korea took large shares in the merchant chip manufacture business, i.e. manufacturing chips for other chip design companies. Intel and Micron are two of the very few companies left that manufacture all or most of their own chips. Companies like Broadcom and Qualcomm rely on merchant fabs to produce their chips. Today’s chip industry is like a five-player game of Risk. There are two major players, the U.S. and China. The next three minor players, Taiwan, Korea, and Japan, will tilt towards wherever the prevailing market winds are blowing. If one player—China—is determined to be a world leader, and will favor doing business with companies that manufacture only or primarily on its own territory, what will companies in all five countries do?
Intel provided the answer when its vice president, Stacy Smith, told the New York Times last month: “China is the largest market for us on the planet…it makes sense to locate some production in China.” In addition to building a fab, Intel took a 20% stake in Tsinghua Unigroup. Intel won’t talk publicly about what was involved in such a deal, but doubtless the Chinese government is determined to get as much of Intel’s world-leading technology into China as possible. In this game of Risk, it seems inevitable that more and more production gravitates to China. What does that mean for the U.S.? Well today, Intel is the last man standing, the only company left willing and able to finance chip fabs in the U.S. on its own. The company is not growing as quickly as other major chip companies and on Wall Street its strategy is looking a bit questionable.
Put simply, this is a trade war in which China has a winning strategy. If the U.S. can be said to have a strategy at all, it looks like a losing strategy. Cynics argue the Chinese have been trying, and failing, for at least two decades to gain significant positions in the chip industry. But the Chinese are nothing if not persistent. Each year, the Chinese economy gets larger and its world influence grows. According to the Semiconductor Industry Association, last year, Chinese consumption of chips grew by 9.2%, while consumption in North and South America fell by 4.7%. Figure 1 shows that Chinese high-tech production is rapidly approaching the level of the U.S.—and will surpass us before long.
An instructive example is networking: the Chinese government decided two decades ago to free itself from dependence on America’s Cisco Systems for networking technology. China Brief says the policy is known within China as “de-Ciscoification.” Today, China’s networking giant, Huawei, is larger than Cisco. Now, the Chinese government wants Huawei and other electronics companies to be free of what China Brief calls “a dangerous over-reliance” on western semiconductors. China will make many mistakes but with a budget that starts at $150 billion, it can afford some huge mistakes and still come out successful. (For comparison, Intel’s annual capex budget is $12 billion.)
In the 1980s, the U.S. lost, and then regained, chip industry leadership from Japan. Since then, the U.S. has lost chip production leadership to Taiwan and Korea., Although American-headquartered companies today account for half the chip market, only 13% of the world’s chips are actually produced in the U.S. But the trend is for the industry to move away from the U.S. Research and development tends to follow manufacturing. Broadcom was an early leader in outsourcing the production of chips. Once based entirely in Irvine, California, Broadcom does R&D around the globe today. Figure 2, from a recent Broadcom presentation, shows that the company’s strong growth in R&D overseas, especially in Asia.
The chip industry is highly profitable. Last year, Intel showed an operating profit of $4.9 billion on revenue of $59 billion; for Qualcomm, the figures were $6.5 billion on $23.6 billion. Yet like every publicly traded company, chipmakers must focus on short-term growth. They cannot justify large capital projects in the U.S. when investors demand higher earnings—and when Asian nations incentivize, and in the case of China insist on, production on their territory. Intel estimates it costs 20% more to build a fab in the U.S. Is there room for a multi-polar chip industry? It would seem not. Today 55% of the world’s production is in Taiwan. As it happens, China’s long-term goals include not only being dominant in chips, but also annexing the island of Taiwan. If Taiwan loses leadership in chip manufacturing, its own citizens as well as the U.S. government may become less interested in preserving its independence.
In the view of many technology executives, forced technology transfer is the worst of China’s unfair practices, because it deprives a company of its “crown jewels,” the unique intellectual property that differentiate it in the marketplace. If I said to you I wanted to pay you $100,000 a year to have dinner with you and your wife at your home every Saturday night for the next three years, you might agree. But if I told you that at the end of three years, I would seize your home, your car, and your wife, and kick you out into the street, you would probably grab me by the scruff of the neck and throw me out of your house. Yet that’s what “forced technology transfer” is. It is a death sentence for companies that agree to it, yet most major American technology companies are doing it today, through gritted teeth.
When Intel CEO Brian Krzanich appeared recently with President Trump to celebrate the decision to spend $7 billion on Intel’s Arizona fab, political commentators looked at the event as a PR victory for the President. I saw the event instead through Krzanich’s eyes. Intel is the outstanding corporate citizen of Silicon Valley. It wants to do the right thing...for its employees, its customers, and America. But it needs the U.S. government to set the conditions that compel it to do the right thing. Privately, I believe Krzanich would thank the U.S. government for such decisions.
The PCAST report proposes that the U.S. launch large government-funded R&D projects in the mold of the Apollo space program of the 1960s. That’s a great idea. I’m sure the money can be found. But it doesn’t go far enough. We need:
- Tougher controls on technology that can be exported to China
- An end to forced technology transfer.
- Tougher controls on companies that can be acquired, based not only on defense considerations but on the importance of those companies and industries to the U.S. economy and taking into account the intentions of the acquiring company and its home country
- To explore ways to provide long-term finance to profitable companies in strategic industries
- A commitment to restore a significant portion of chip manufacturing to the U.S.
- To look at the entire electronics supply chain with a view to establishing the critical mass necessary to keep enough of the industry in the U.S. for it to continue to grow and invest of its own accord.
Figure 1. Chinese high-tech manufacturing output rapidly approaching US output. (Source: National Science Foundation Science and Engineering Indicators 2014)
Figure 2. US chipmaker Broadcom has as many design (R&D) sites in Asia as in North America. (Source: Broadcom)