Are sanctions a double-edged sword? What the data reveals about the US-China war for chip supremacy

After seemingly being blindsided by China’s strength in 5G technologies, the US government now fears losing its leadership position to its Asian rival when it comes to chip making. And given what is at stake, who can blame it? The semiconductor industry is worth about $400 billion in annual revenue and underpins the global tech economy, meaning that any company at the bleeding edge of these technologies has a massive advantage over its competitors.

As manufacturing has slowly migrated to Asia, concerns have been mounting in the United States that the country’s position at the top is slipping. In response, the government has pushed back against China by making it impossible for companies such as Huawei to use US technologies or acquire software and other tools that are used in semiconductor manufacturing. The question is whether the United States really has to worry about losing its title as the semiconductor superpower.

Some critics have been pushing the narrative that the United States has already lost its dominance in the industry. This message intensified after the news broke that Intel, the US poster child chipmaker, has considered outsourcing manufacturing – a key part of the market dominated by overseas companies. What is more, it comes against the backdrop of China making notable improvements to its semiconductor technologies, including in the lucrative market for AI chips.

High-level analysis of published patent applications, conducted by Sumair Riyaz and S Rajan Kumar, lead business development and lead technical analysis at Dolcera, shows that industry leaders and those from key jurisdictions still consider the United States to be a hub of innovation. However, if this situation continues, then the country could well have cause for alarm.

State of play: industry heavyweights

To get a better understanding of the semiconductor industry we must first identify the key players. Samsung is the world’s leading patent filer in this space based on sheer volume, with 45,938 published patent applications (see Figure 1). The South Korean giant is followed by Taiwan Semiconductor Manufacturing Company (TSMC), recording 40,195 assets, and Semiconductor Energy Laboratory, which has a portfolio of 33,075 patents. These figures provide immediate insight into the importance of the Asia-Pacific region, as only four of the top 12 patent-owning companies are based in the United States – and no US companies feature in the top three.

Figure 1. Top semiconductor patent filers

Despite the high visibility of Asian players, the United States is the most important jurisdiction for the world’s most prolific chipmakers by a landslide, with 351,035 applications filed there to date (see Figure 2). China is the second most popular destination with 136,311 patents, while Japan rounds out the top three, recording 103,826 published patent applications at the time of analysis. China’s position is significant here, given that no local businesses rank in the top filers list.

Figure 2. Geography breakdown of published patent applications – top filers (industry)

Drilling down into the data, Dolcera’s analysis also offers a more nuanced insight into how filing strategies at the world’s busiest semiconductor businesses have evolved. Figure 3 categorises the data into patents by priority year and priority country. It reveals a gradual decrease in the number of applications being lodged in the United States, Japan and South Korea, while a subsequent increase takes place in applications designating China after 2013.

Figure 3. Priority year versus priority country – top filers (industry)

Source: Dolcera PCS

It is certainly worth getting an idea of who the busiest chipmakers are and what is on their minds when it comes to the geographic scope of their portfolios, but there are other ways to identify the lead players in the industry aside from their level of patenting activity.

As such, Kumar and Riyaz conducted another search, this time focusing on top companies by revenue. They identified Intel, Samsung, SK Hynix, Micron Technology, Broadcom, Qualcomm, Texas Instruments, ST Microelectronics, Kioxia (formerly Toshiba Memory) and NXP as the top 10 players through this analysis (see Figure 4).

Figure 4. Top players by revenue – 2019

Intel is the top dog in terms of revenue, but Samsung leapfrogs its US opponent in terms of patent quality (see Figure 5). Dolcera has an algorithm to determine the value of individual patents, where a score of three is the benchmark for high quality. The South Korean multinational owns 8,615 published patent applications that receive a Dolcera rank of three and above. With 6,064 patents Qualcomm owns the second largest portfolio in terms of strength, followed by Intel with 5,349.

Figure 5. Top players by revenue – quality check (patents ranked three and above)

Figure 6. Geography breakdown of published patent applications – top players by revenue

Given the importance of China as a region for IP expansion among the most prolific semiconductor companies, it is important to consider how major players interact with the country. Indeed, Figure 7 shows that TSMC, Samsung, Intel and Qualcomm are all investing more heavily in protecting their assets in this market.

Figure 7. China applications – top players by revenue

Source: Dolcera PCS

State of play: geography focus

The first part of Dolcera’s analysis has established who the lead chipmakers are from a patent perspective and revealed that, while these companies are paying increasing attention to China, the United States continues to dominate the market. Now, it is time to consider the semiconductor landscape on a broader scale by examining the competitive patent filing strategies of companies based on their home country (see Figures 8 to 15).

Figure 8. US firms – filing strategy

Figure 9. Top US companies – portfolio breakdown by geography

Source: Dolcera PCS

Figure 10. South Korean firms – filing strategy

Figure 11. Top South Korean companies – portfolio breakdown by geography

Source: Dolcera PCS

Figure 12. Taiwan-based firms – filing strategy

Figure 13. Top Taiwanese companies – portfolio breakdown by geography

Source: Dolcera PCS

Figure 14. China-based firms – filing strategy

Figure 15. Top Chinese companies – portfolio breakdown by geography

Source: Dolcera PCS

Although Asia is home to a number of chip-making powerhouses, the United States is consistently the most important market for patent protection. That is, with the exception of China, where local businesses often have domestically focused portfolios. But even in this case, Chinese businesses file almost in near equal volume in the United States as they do at home.

What is more, China continuously features as one of the most important jurisdictions for patent protection. This reflects the country’s position as a significant single market, which offers a lucrative opportunity for the world’s key players. In 2018 China was the largest market in Asia-Pacific. It represented 56% of sales of semiconductors to electronic equipment makers in the region and accounted for 34% of these sales globally. The rapid growth in Chinese demand for semiconductors has lifted the industry worldwide, and because of the gap between consumption and production, 91% of this demand is satisfied by imports – creating a great opportunity on which to capitalise.

But if there is any fear that the United States has lost its lead status in the semiconductor industry, it should be reassuring for US investors to see how much more active US businesses are in terms of patent output. Chinese firms have lodged 7,416 patents in their home jurisdiction, whereas US businesses have filed 42,231. This is an important indicator that the United States is still a critical source of innovation in the sector.

Breakdown of major markets

United States

Patent protection at home is a top priority for US businesses (see Figure 8). China is the second most important market, followed by Taiwan, Europe, Japan and South Korea. While local companies do seek protection for their innovations abroad, these activity levels are in no way comparable to filing levels at home. The major players in the United States follow a similar strategy, but it is worth noting that their portfolios are highly diverse in terms of geography, with robust holdings in key markets (see Figure 9).

South Korea

The United States is the top market for firms based in South Korea. These businesses also file a significant number of patents at home, and a far smaller number in China (see Figure 10). Samsung has filed nearly twice as many patents in the United States as it has in South Korea. Out of the leading companies, only LG Corp is more active at home than abroad (see Figure 11).

Taiwan

As with their South Korean rivals, Taiwanese firms file most often in the United States, followed by China and then their home market (see Figure 12). While these companies have upped their output to China in recent years, the filing volumes cannot be compared to those in the United States. As a reflection of its business, TSMC’s portfolio is highly global, with the company having filed three times as many patents in the United States as in China, and with a US portfolio that is four times larger than its holdings in Taiwan.

China

Similar to firms from the other countries analysed, Chinese businesses have invested heavily in patent protection in the United States in order to tap into the lucrative semiconductor market. However, Chinese firms are the first example of companies prioritising local protection in Asia. Other markets around the world appear to be significantly less important to Chinese businesses, as seen by the relatively low volume of patents being filed elsewhere.

Semiconductor innovation and the patent decline

The chip-making market is built on high levels of R&D. The US chip industry invests about one-fifth of revenue in R&D – among the most of any sector of the economy, according to data from the Semiconductor Industry Association (SIA).

In fact, keeping the foot on the gas when it comes to innovation is so important that in the United States R&D intensity has increased regardless of annual sales cycles, with expenditures growing at a compound annual growth rate of approximately 6.6% from 1999 to 2019. In 2019 R&D investment in the US semiconductor industry reached $39.8 billion, according to the SIA. While the United States spends more on R&D as a percentage of sales than any other country’s semiconductor industry, all major chipmakers are splashing out cash on innovation to keep ahead of the competition.

Interestingly, though, Dolcera’s research shows that patent filing in the semiconductor industry has been on a gradual decline for the past decade (see Figure 16). There are a number of possible reasons for the drop in output, with one potential explanation simply that the scale of innovation is returning to a more normal pace after unprecedented growth over the past 50 years. Further, the slump in filings may not be a cause for concern given the high R&D levels and overall outlook for the semiconductor market, which is expected to grow anywhere between 6% and 7% over the next few years.

Figure 16. Patent filings relating to semiconductors – trends over time

That being said, there is a very real threat to R&D intensity in the industry as the trade war takes its toll on leading players, particularly in the United States.

According to a report by the Boston Consulting Group (BCG), entitled “How restricting trade with China could end US semiconductor leadership”, the United States’ status as the chip-making superpower is grounded in what it describes as a virtuous innovation cycle. Essentially, US leadership relies on revenue from global markets (including China) to achieve the scale that is necessary to fund investment in R&D, which leads to superior technologies, thereby reinforcing the United States’ position and profitability.

The current icy relationship between the United States and China puts this cycle at risk. The median year-on-year revenue growth of the top 25 US semiconductor companies plummeted from 10% in the four quarters immediately prior to the introduction of the first rounds of tariffs in July 2018 to approximately 1% in late 2018, reports BCG. Further, in each of the three quarters after the United States restricted sales to Huawei in May 2019, the top US semiconductor companies reported a median revenue decline of between 4% and 9%. The trade conflict is often cited as a significant factor in their performance.

BCG went a step further to describe the effect on the US semiconductor industry in two different scenarios. The first assumes the status quo, while the second considers an escalation that would halt bilateral technology trade. In both scenarios BCG found profound negative repercussions for the US semiconductor industry (summarised in Table 1). The most significant potential consequences are as follows:

• Over the next three to five years, US semiconductor companies could lose 8% of global share and 16% of their revenues if the United States maintains current restrictions.
• Over the same period, US companies could lose 18% of global share and 37% of their revenues if the United States completely bans semiconductor companies from selling to Chinese customers.

Table 1. US-China battle for leadership in semiconductors

The impact to the top line would mean that leading semiconductor companies would have to cut back on R&D spending, which would undoubtedly have a tangible impact on patent filing activity and could genuinely risk the United States’ position as the world’s semiconductor spearhead. BCG predicts that if the United States loses its virtuous innovation cycle, South Korea will overtake it in the short term, and China may arise as the new chipmaker superpower in the long term.

China’s threat: myth or reality?

Now comes the big question: is China – as it stands – really a threat to US dominance in the semiconductor industry?

It is clear that the major players, and the key semiconductor markets in general, view the United States as the most important country for innovation and patent protection. But that does not mean that China has not become a serious contender to the United States’ title of semiconductor king.

The Chinese government made chip making a priority in 2014 when the State Council set the goal of becoming a global leader in all segments of the semiconductor industry by 2030. The country further announced, under the Made in China 2025 industry plan, that it aims to produce 40% of the semiconductors it uses by 2020 and 70% by 2025. (For the record, BCG’s latest estimate puts China’s self-sufficiency to between 25% and 40% by 2025.) To help achieve its goals, the Chinese administration created a $29 billion state-backed fund to invest in the industry.

There is no doubt that China’s semiconductor technologies have improved, but the country has also shown itself to have comparative strength in an important area of the market. The chip industry is not just about smartphones, and the recent data boom has led to the innovation of chips for data centres and servers, and for inference and insights from data on the Cloud or edge. Critical to these technologies are AI and machine learning, fields in which China has worked hard to operate at a level that is on a par with that of the United States.

This, according to Dolcera’s analysis, is where the battle begins for technology dominance and what the Trump administration has been trying desperately to stall. The current government perceives China’s growth in AI to be a national security threat and the trade restrictions are designed to slow down the country’s high level of innovation in this space.

To be clear, an AI chip is a processor used to train AI algorithms and to infer. This forms the basis for products used for AI applications. The known players in the market are Nvidia (graphics processing units), Intel (central processing units), AMD, Xilinx (field programmable gate arrays) and other application-specific integrated circuit players. Other notable start-ups in this domain include Habana Labs, a subsidiary of Intel, Graphcore and Cerebras, which are based in Israel, the United Kingdom and the United States, respectively.

AI chip players have an upper hand in the AI algorithm training that forms the basis for further application development and the area is dominated by established players such as Nvidia, Intel and Cambricon. The latter is the only major player from China and it is worth noting that China’s start-ups are mainly focused on inference, which is less technologically intensive. This means that Chinese firms must rely on chips made by their US and European counterparts. With the trade ban in place, China has been forced to pin its hopes on Huawei’s HiSilicon, which briefly made an appearance as a top 10 semiconductor leader by sales earlier in 2020. Huawei has been in the data centre and cloud business for some time and offers AI chips and AI frameworks, which are popular in the region.

With all this in mind, it is worth considering who the key players are in the AI chip space from a patent perspective. Huawei is the world’s top corporate owner in AI in edge computing in data centres, followed by Nokia and Ericsson (see Figure 17).

Figure 17. AI in edge computing in data centres – key players

In terms of publication geography, the United States is the most sought-after market, closely followed by China (see Figure 18).

Figure 18. AI in edge computing in data centres– top geographies

The major players list for AI processors looks strikingly different from the one for AI in edge computing in data centres. US heavyweights IBM, Alphabet and Microsoft claim the top three spots, but a number of Chinese firms also feature, including Baidu, Tencent, Alibaba and Cambricon (see Figure 19).

Figure 19. AI processors – key players

Again, innovation is concentrated in the United States and China, with the latter pulling ahead slightly in terms of patent volume (see Figure 20).

Figure 20. AI processors – top geographies

The trade war impact on China’s industry

The restrictions implemented against China are certainly being felt. China imports more than $300 billion worth of integrated circuits every year, and most of it comes from the United States (see Figure 21). According to Dolcera’s analysts, the country’s reliance on external manufacturers is even greater.

Huawei, which has been the main target of the Trump administration, relied heavily on TSMC for its chip production, and being cut off from the foundry has taken a toll on the firm’s accelerated growth. The Asian telecoms giant has now turned to China’s most advanced chipmaker, Semiconductor Manufacturing International Corporation (SMIC), but the company’s technological capabilities significantly lag behind the world’s leading foundries, including TSMC. What SMIC needs is access to machines that use extreme ultraviolet lithography technology, but these are manufactured by ASML and the United States has ensured that the Chinese foundry is denied such tools. To make matters worse, the Trump administration is now considering blacklisting SMIC.

Trade restrictions have created significant hurdles to the ascendancy of China’s industry. At the 2020 World Semiconductor Conference, Wang Xuguang, chief executive officer of AINSTEC, a Suzhou-based company that develops 3D visual chips, said: “The entire chip industry is too fragile to defend itself. We are at least 20 years behind compared to Silicon Valley from scale and quality of talent to size of the ecosystem. If we can prosper (with the United States), that’s the best, but if the situation doesn’t allow this to happen, we need to think what we have on our hands.”

The fear for senior executives in the semiconductor market is that the US tech sanctions will extend to other companies and sectors – which could devastate the market. Gary Yang, a founding partner at Beijing-based venture capital firm Sky Saga Capital, told the South Morning China Post (SCMP): “The rise of Japanese and South Korean semiconductor industries and their high-end equipment manufacturing can largely be attributed to their ties with US industry chains, besides the investment from the state.” Yang continued to say: “It’s like having a customer with the entire know-how of an industry. They can help guide you at some crucial junctures. That is why investment alone can’t succeed and we need crucial directional guidance along the way.”

It may be difficult for China to reach self-sufficiency in chip making without any help from the United States. But while some experts and senior executives are hesitant, others are more optimistic about the future of the country’s semiconductor industry. SMIC founder and former chief executive Richard Chang Rugin told the SCMP that China can catch up to the United States in the next generation of chips. He argued that China was in pole position for the long game given its already strong presence in hi-tech applications such as wireless connectivity, AI and cloud computing.

Figure 21. China’s semiconductor imports

Riyaz and Kumar reiterate Rugin’s early optimism. “China is treading along the same path as the United States to become a dominant player in the market, and while the country’s semiconductor industry may be impacted in the short run, we cannot rule out the potential for it to bounce back,” says Riyaz. The Dolcera analysts point out that China has already started seeking out alternatives, it is ramping up local activities and is getting a second round of chip-focused funding that is backed by the state. The Chinese government has had its sights set on self-sufficiency since 2014 and the ongoing trade war has only made it double down on its long-term objective.

There is also a key loophole that China could exploit should the situation become dire. RISC-V is an open source instruction set architecture (ISA) that has been generating buzz in the industry because of the threat that it poses to UK-based ARM’s licensing business. RISC-V already has a number of bigwigs on board, including Nvidia, Western Digital, Qualcomm, Google, Alibaba, Huawei, ZTE and Huami (Xiaomi’s wearable devices partner). In addition, some leading companies are already utilising the RISC-V ecosystem. Take Alibaba, for example, which released its server chip Xuantie-910 based on the RISC-V open ISA last year with the aim of it being used in data centres.

“This creates opportunities for Chinese firms to tap into the nascent ecosystem that is evolving and help them meet their requirements for the semiconductor industry,” explains Kumar. “RISC-V does not currently fall under the trade-ban blanket. This means that leading start-up SiFive is exempt from restrictions and could license its technology to Chinese Firms. In the event that SiFive is banned, then China could rely on non-US firms like Syntacore, Nucleisys and Andes, among others.”