"The damage done by this outbreak of open technology warfare is likely to be much more serious and long-lasting than normally assumed in the media." EE Times interviews economist Dieter Ernst.
In deciphering trade relations between any countries, it always helps to step back and first put everything into historical context. To that end, EE Times turned to Dieter Ernst, a senior fellow at the Centre for International Governance Innovation (Waterloo, Canada) and the East-West Center (Honolulu). Ernst is an economist and a long-time observer of China, known for his research examining industrial and innovation policies in China, the U.S., and emerging economies, with a focus on standards and intellectual property rights.
In the following interview, Ernst predicts, “The damage done by this outbreak of open technology warfare is likely to be much more serious and long-lasting than normally assumed in the media.”
Ernst, who just came back from field research in China’s AI industry, also shed some light on the likely impact of the trade war on AI technology development.
EE Times: With Trump’s executive order in and the U.S.’s increasing hardline stance against China, where do you see U.S.-China trade relations today, compared to almost 20 years ago when China joined the World Trade Organization (WTO)?
Dieter Ernst: To understand how dramatically U.S.-China relations have changed, it is necessary to go back to U.S. President Richard Nixon’s 1972 visit to China. That visit ended 25 years of no communication or diplomatic ties between the two countries. Powerful geopolitical and economic reasons were behind this opening up of China. By splitting up the socialist camp, U.S. policy and defense elites expected to gain more leverage over relations with the Soviet Union. Securing China’s support in resolving the Vietnam War was a particularly important factor.
For U.S. business, the rapprochement to China promised access to the potentially huge China market and to its cheap labor. The U.S. government had banned trade with China until the early 1970s. However, with the full normalization of diplomatic and commercial relations in 1979, the United States became the second-largest importer to China and, in 1986, was China’s third-largest partner in overall trade.
China’s entry into WTO in 2001 was meant to accelerate the country’s integration into international trade, with a particular focus on its rapidly growing role as the global electronics factory. The rest is history. In 2018, China was the United States’ largest U.S. merchandise trading partner (total trade at $660 billion), third-largest export market ($120 billion), and largest source of imports ($540 billion). China is also the largest foreign holder of U.S. Treasury securities (at $1.1 trillion year-end 2018).
U.S. foreign direct investment (FDI) in China has grown rapidly, culminating in a U.S. FDI stock in China through 2017 at $256 billion, compared to China’s FDI stock in the U.S. of $140 billion. As a result, both economies are deeply intertwined into multi-layered global corporate networks of production and innovation.
EE Times: What has changed now?
Ernst: The rise of U.S. economic nationalism, as embodied in Trump’s “America First” doctrine, has drastically changed America’s China policy. In a fundamental break with established U.S. trade diplomacy, the new U.S. trade dogma is ostensibly predicated on the fundamental arguments of mercantilism that trade is essentially a zero-sum game, and hence, an optimal policy is to ensure that exports exceed imports in any bilateral trade relation. Add to this a fundamental belief that U.S. security requires unchallenged leadership in science and technology.
China’s efforts to forge ahead in advanced manufacturing and services are considered to be a major threat. China, thus, has become a strategic competitor — an “adversarial nation and bad actor” that should no longer be allowed “to steal our ideas, copy our technology, and cheat their way to leadership — in a field central to our national security.”
These words are now followed by action. On May 15, the battle against China culminated in Trump’s executive order to ban American telecommunications firms from installing Huawei or ZTE equipment. However, even more important is a rule announced on the same day by the Commerce Department. This rule has placed Huawei and 68 affiliates in more than two dozen countries on its so-called “Entity List,” which is basically a trade blacklist that bars anyone on it from buying parts and components from U.S. companies without the government’s approval first. Obtaining such approval is practically impossible. According to a Federal Register posting, “The U.S. government will review license applications under a policy presumption of denial.” This rule takes effect immediately, which indicates how serious the U.S. government is in its goal to weaken, if not destroy, Huawei.
‘Temporary general license’
A few days later, on May 21, the U.S. Commerce Department scaled back some of these restrictions, issuing a 90-day “temporary general license,” valid through Aug. 19. This temporary license is allowing sales and services provided to Huawei and affiliates to support existing Huawei handsets as well as existing networks and equipment. Maintenance arrangements made between Huawei and customers up to May 16 for existing networks and equipment fall under the moratorium. It is important to emphasize, however, that Huawei is still barred from buying American components to manufacture new products without license approvals that likely will be denied.
All-out technology war
In short, there is no indication that the U.S. strategy has changed. U.S.-China trade relations will continue to be shaped by an all-out technology war. The U.S. Government is drastically tightening the export control of a broad portfolio of information technologies, with a focus on China. Since August 2018, the Export Control Reform Act (ECRA) requires the Commerce Department to create lists of “emerging” and “foundational” technologies that are essential to U.S. national security. This legislation had bipartisan sponsorship and was passed by both houses of Congress. Many provisions of the ECRA are still under development and will be rolled out over the next year or more. It seems that the restricted technologies will include a broad set of generic technologies, such as artificial intelligence, biotechnology, self-driving vehicles, nanotechnology, robotics, and semiconductors. While any new regulations will require private sector review and input before going into place, the list of restricted technologies is likely to be quite extensive.
Add to this the U.S. Treasury’s Foreign Investment Risk Review Modernization Act of 2018 (FIRRMA) that drastically expands the mandate of the Committee on Foreign Investment in the U.S. (CFIUS). Of particular importance are restrictions on so-called “deemed exports,” which limit any information about a controlled technology to a foreign national. Once these restrictions are implemented, they will throw a wrench into the existing pattern of knowledge sharing that is the lifeblood of the global IT industry. This unprecedented arsenal of technology export restrictions is bound to constrain China’s options to use off-the-shelf chips from leading U.S. and other foreign vendors.
In short, mutual distrust and sometimes open hostility are now shaping U.S.-China relations. On the U.S. side, trade diplomacy seems to have given way to a “regime change” agenda. The current “China bashing” mood in Washington, D.C., reflects a broad consensus among U.S. defense, foreign policy, and economic policy-making elites that China’s rise in information technology poses a serious threat to America’s leadership in science and advanced technology. In this view, the U.S. needs to implement aggressive trade, investment, technology, and visa restrictions to contain China’s technological and geopolitical ambitions. This consensus embraces the political spectrum. Both Republicans and Democrats share it, and it will persist even if Trump does not win the 2020 election.
EE Times: What damage, if any, has the current administration done to the U.S.-China trade and, more broadly, to the global semiconductor market? In other words, who will be more likely to suffer in what ways?
Ernst: The damage done by this outbreak of open technology warfare is likely to be much more serious and long-lasting than normally assumed in the media. All participants in the global semiconductor industry will be affected, irrespective of their nationality.
There is no doubt that U.S. technology export restrictions will slow down China’s efforts to catch up in “emerging” and “foundational” technologies.
EE Times: I hear you were in China recently.
Ernst: Yes, I just came back from field research in China’s AI industry that covered key players — the BATs (Baidu, Alibaba, Tencent), Huawei, some AI unicorns, AI chip companies, and leading AI research institutes. Without exception, all interviewees were very concerned whether secure access to core components and support services might be disrupted through a progressive “decoupling” of existing U.S.-China IT value chains. An equally important concern is the growing incidence of visa restrictions imposed on Chinese students, on the engineers and managers of these companies, and on members of Chinese research institutes. Most of our interview partners were visibly hurt by the aggressive language used in the U.S. policy announcements — much damage has already been done to America’s once seemingly invincible “soft power” image. This deep sense of disappointment was even more palpable among students when I gave talks at China’s leading universities.
EE Times: Tell us more about this trade war’s impact on the field of AI.
Ernst: The greatest damage is likely to fall on some of the AI unicorns. They can hardly keep up with the rapid demand growth for AI in the China market. These companies are trying to recruit young engineering graduates from across China. They are fiercely competing for experienced top talent from overseas. VC funds are still available, albeit at a slightly slower pace than 2018. While money is not a constraint at present, much of senior management’s attention is focused on capturing China’s booming markets for AI applications. Hence, investment in AI research, both applied and basic, remains limited. There is a broad consensus in the interviewed companies that their own in-house R&D efforts might well be insufficient to cope with the intensifying technology war with the U.S.
China’s AI unicorns are well aware of the government’s renewed efforts to strengthen domestic innovations through an increasing array of support policies and incentives. There are strong reasons for these companies to participate in such schemes, if only to qualify for additional funds. But it is unclear how active and effective such participation might be, as long as these companies are overwhelmingly focused on competing in China’s rapidly growing mass markets for AI applications.
EE Times: Where do Huawei and the BATs stand in that AI context?
Ernst: The situation, however, seems to be different for Huawei and the BATs. Detailed findings of our research will be published later this year. Suffice it to say here that the larger Chinese companies have prepared their “Plan B” for quite some time. And some companies like Huawei have already turned their Plan B into a de facto Plan A that now shapes their corporate survival strategy. For instance, all the BATs are now developing their own AI chips, with Alibaba having placed orders for 7-nm AI chip solutions with TSMC and Global Unichip.
China vs. US: Who will Blink First?
EE Times: Give us your reading on Huawei. Where is it heading?
Ernst: Take Huawei, which is the prime target of Trump’s hardline technology restrictions. While much better prepared than ZTE, Huawei is still very reliant on foreign suppliers for CPUs, GPUs, FPGAs, and high-end memory, many of them from the U.S. According to Ren Zhengfei, Huawei’s founder and CEO, Huawei procures about $67 billion worth of components every year, with roughly $11 billion coming from U.S. suppliers. For leading-edge semiconductors, Huawei’s dependence on U.S. companies is especially high.
With regard to AI chips, Huawei has heavily invested in the development of leading-edge processors as part of an integrated AI stack around Huawei’s Mindspore framework. At the same time, Huawei also provides drivers that support leading AI frameworks such as Pytorch and TensorFlow as well as its proprietary MindSpore framework. According to The Linley Group, Huawei now offers a set of neural-network accelerators, including the Atlas 200 module for smart cameras and the Atlas 300 card for edge computing. These products use the Ascend 310, a chip the company designed itself. It employs an architecture called DaVinci that implements 3D tensor computations. The company is developing a more powerful chip, the Ascend 910, but products using this device have been delayed until later this year.
In its quest to become the world’s largest smartphone vendor, Huawei has asked its Taiwan supply chain partners (IC packaging, testing, chip probing, and optical components) to relocate their operations to China. However, Taiwan’s regulatory restrictions may make such relocation quite difficult. In addition, the U.S. government applies quite heavy pressure on Taiwan’s government to obstruct such moves.
A dramatically more serious challenge emerged on Sunday, May 19, when Google announced (apparently with much hesitation) that it would comply with the U.S. Commerce Department technology export ban, barring Huawei from critical updates of Google’s proprietary version of the Android operating system. If this ban remains in place, this would ring the death bell for Huawei’s efforts to eclipse Samsung as the world’s No. 1 smartphone maker. Huawei claims that it has developed its own alternative to Android. Yet, as the failure of Amazon’s Appstore to unseat Android demonstrates, Huawei would find it next to impossible to convince leading program developers to build applications for such a new Huawei OS. As one analyst puts it: “Deprived of Google’ software, Huawei would be selling featherless chickens. … There is no positive spin to this situation for Huawei.”
Following Google’s bombshell, most leading U.S. semiconductor companies have frozen the supply of critical components and software to Huawei, including (but possibly not restricted to) Intel, Qualcomm, Xilinx, Broadcom, Qorvo, Micron Technology, and Western Digital. It is too early to judge how seriously Huawei might be affected. Some observers argue that Huawei may actually be quite well-insulated from the Qualcomm impact, as it builds its own mobile processors and modems. And Huawei claims that it has been preparing for this blockade by stockpiling chips from U.S. suppliers to last it at least three months, which should be enough time to tell if the current measure is a scare tactic or a permanent imposition from the U.S. government. In addition, Huawei has reportedly promised not to cut its orders with Taiwan’s upstream and downstream supply chain partners through the second quarter of 2020. It has also inquired about the possibility of increasing their production capacity.
Of particular importance is the decision of German chipmaker Infineon to suspend shipments to Huawei, as this could signal that non-U.S. companies will follow the U.S. hardline stance on Huawei. STMicroelectronics, another European chipmaker, is said to discuss whether it will continue shipping to Huawei. In addition, Japan’s Toshiba Memory, the world’s second-biggest NAND flash memory provider, and Japan Display, the screens supplier, are investigating the implications for their businesses of the U.S. blacklisting of Huawei.
For Huawei, a fundamental concern is secure and guaranteed access to leading-edge process nodes at 7 nm from TSMC. So far, there are no signs that TSMC will give in to U.S. pressure and restrict Huawei’s access to its leading-edge foundry services. Huawei’s 16-nm, 12-nm, and 7-nm chips are all from TSMC. On May 20, TSMC announced that it will not stop the supply plan for Huawei. This means that TSMC will provide Huawei’s foundry Kirin 980 with security improvements in the second half of the year. The upcoming Kirin 985 processor will not be affected. In fact, Huawei is one of the three largest foundry service customers of TSMC. It obviously wouldn’t be easy for TSMC to lose an important “preferred customer” like Huawei.
As for Huawei’s recently announced Arm-based CPU Kunpeng 920, this is important, as it seems to indicate a fairly close relationship with Softbank’s Arm. Hi-Silicon has designed the CPU for Kunpeng 920, but that CPU appears to be slower than the Arm Cortex-A76 or Marvell ThunderX2. However, for the China server market, it seems that its performance is considered to be sufficient. More important for Huawei’s survival may be that Arm is working with Huawei and several others to build a server ecosystem around Arm under the Neoverse brand. Huawei can be an effective representative of this effort in China.
Huawei’s relationship with Arm/Softbank seems to be critical for Huawei’s preemptive defense against the Trump executive order. To strengthen its design and process technology capabilities, Huawei is investing £3 billion in a 400-person chip research and development factory close to Arm’s Cambridge headquarters.
EE Times: What's the damage done to the U.S. electronics indsutry?
Ernst: It is worth considering some of the hurdles that the U.S. technology export restrictions may face. This campaign may end up hurting the U.S. IT industry and its broader economy, as well as key partners of the U.S. Supply chains may be disrupted and jobs destroyed on a significant scale in China but also in the U.S., and customers would have less choice.
Trump’s executive order is already imposing collateral damage on the U.S. IT industry, especially in semiconductors, as well as on the ICT industries of Japan, Korea, and Taiwan.
Qualcomm among the hardest hit
Take U.S. semiconductor companies. Huawei acknowledges that it is still very reliant on foreign suppliers for CPUs, GPUs, FPGAs, and high-end memory. Qualcomm is among the hardest-hit U.S. chipmakers after the Huawei blacklisting. And Bloomberg reports that the toughening stance against Huawei hurt the stocks of some U.S. companies that supply it on May 16, one day after the technology blacklist was announced. Qualcomm fell 4% — the company received two-thirds of its sales from China in its most recent fiscal year. In addition, Broadcom Inc. was down 2.3% and Xilinx Inc. dropped 7.3%. Their losses helped drag down the benchmark Philadelphia Stock Exchange Semiconductor Index by 1.7%.
It will be interesting to see whether the U.S. government will be able to sustain an effective blockade of Huawei by U.S. companies like Intel, Qualcomm, Xilinx, and Nvidia. After all, these U.S. companies critically depend on China, the largest worldwide semiconductor market.
In addition, will Trump’s Huawei attack force non-U.S. companies like Arm-Softbank, the Dutch company ASML, Samsung, Toshiba Memory, Japan Display, HonHai/Sharp, and, most importantly, TSMC to follow America’s blacklisting of Huawei? This raises an important question: Will the extraterritorial power of the U.S. Commerce Department be strong enough to prevent non-U.S. companies across the semiconductor value chain from continuing highly profitable links with Chinese semiconductor companies?
Trade lawyers argue that underlying U.S. law is extraterritorial. In principle, foreign items being sold to Huawei that have more than 25% U.S. content might also require a U.S. export license. If so, license denials could effectively put Huawei out of business, at least until they locate or develop alternative sources of supply.
It is difficult to disagree with the Bloomberg Editorial Board’s conclusion that driving Huawei out of business would be “a serious mistake,” as it would risk “simply alienating U.S. allies, infuriating average Chinese, and raising the chances of confrontation, all to no obvious end.”
The Trade War’s Impact on the Supply Chain
Return of ‘red scare’?
In the end, the greatest damage of intensifying technology warfare will result from profound and enduring changes in both countries that might well result in long-lasting disruptions of international trade, investment, and global value chains. As hostility toward China is on the rise across America, there are signs of disturbing similarities with the “red scare” hysteria during the 1950s over the perceived threat posed by the Soviet Union. As China bashing is widely accepted, there is a tendency to restrict debates on international competition almost exclusively on the threat posed by China (i.e., Huawei) on America’s security. Trump’s hardening stance on China has led to a “re-nationalization” of the IT supply chain — the conviction that information and communications technology can be trusted only if it is produced in one’s own jurisdiction. The result would be a “politicization of the supply chain,” which is unlikely to be sustainable.
Such an almost exclusive focus of America’s innovation strategy on the perceived security threat from China (i.e., Huawei) distracts attention away from necessary efforts to upgrade America’s innovation system and to improve the security of complex, multi-layered supply chains. The first effort requires a substantial increase (not a reduction) in R&D investments, both by the government and by the private sector. It also requires a much more proactive unified innovation strategy that combines competition policy with best practice approaches to standardization and intellectual property protection. The second effort would seek to ensure trustworthiness and resilience against cyberattacks through a dedicated security strategy. Putting the blame on Huawei is no substitute for such a dedicated security strategy.
5G network security
Take the security of 5G networks. A European study shows that policies to improve 5G network security need to address four issues:
It is time to move beyond the current debate that assumes that only the deployment of Chinese network equipment bears risks for national security. We can no longer afford that our innovation and trade policies are dominated by such naïve and misinformed concepts.