S. Korea, US, China, Taiwan embroiled in fierce competition to dominate AI semiconductors

Posted on : 2020-10-13 17:44 KST Modified on : 2020-10-13 17:44 KST
S. Korean government announces strategy for fostering domestic semiconductor industry
South Korean Prime Minister Chung Sye-kyun (third right) looks at an exhibit on AI semiconductor research before a meeting of government ministers related to the field of technology at the second Pangyo Techno Valley in Gyeonggi Province on Oct. 12. (Yonhap News)
South Korean Prime Minister Chung Sye-kyun (third right) looks at an exhibit on AI semiconductor research before a meeting of government ministers related to the field of technology at the second Pangyo Techno Valley in Gyeonggi Province on Oct. 12. (Yonhap News)

South Korea, the US, China, and Taiwan are in fierce competition for dominance in source technology for the AI-related semiconductor market. All four of these countries appear to be bringing their full technological capabilities to bear on the infant industry, which still lacks a dominant player. Not only traditional semiconductor manufacturers but also startups and non-semiconductor companies such as Google and Tesla have jumped into the fray.

Building the next DRAM by 2030

On Oct. 12, South Korean Prime Minister Chung Sye-kyun presided over a meeting of government ministers related to the field of technology at the system semiconductor design support center at the second Pangyo Techno Valley in Gyeonggi Province. At the meeting, the government unveiled its strategy for developing the AI semiconductor industry, with the goal of maintaining Korea’s status as a semiconductor powerhouse in the emerging field of AI semiconductors.

The government’s strategy consists of four main objectives: developing AI semiconductors into the next DRAM semiconductor by 2030; controlling a 20% share of the global market for AI semiconductors by 2030; developing 20 innovative AI semiconductor firms and training 3,000 highly talented workers; and seeking to create a technological “super gap” through an innovative processor-in-memory semiconductor based on Korea’s world-class capabilities in memory semiconductors.

South Korean technology still lags behind that of the advanced economies. Its AI semiconductor technology has been rated at 84% of the US’ level as of 2018. At 43.0% and 30.7%, respectively, the US and China vastly surpass South Korea’s 10.3% share of AI semiconductor patent applications between 2013 and 2016.

While system semiconductors are used throughout the process of data collection, storage/processing, transmission, and computation/control/learning/inference, AI semiconductors perform key AI calculations related to learning and inference. As semiconductors that perform large-scale computations with high performance and power efficiency, they process vast volumes of data simultaneously, achieving computational power efficiency roughly 1,000 times greater than previous semiconductors. They have been named as a “new weapon” ushering in a “Big Bang” for upstream and downstream businesses with convergence and expansion in different industries.

Forecasts for scale of AI semiconductor industry
Forecasts for scale of AI semiconductor industry
Google and Tesla joint development race

The global market for AI semiconductors is predicted to reach US$117.9 billion by 2030. They’re also expected to exceed 30% as a proportion of the system semiconductor market, which totals US$376.9 billion. It’s against this backdrop that different countries have been waging a heated technology battle.

Through the AI Next campaign, a next-generation AI research project by the Defense Department-affiliated Defense Advanced Research Projects Agency (DARPA), the US has been pursuing government-led R&D on areas such as AI and heterogeneous chip stacking and integration and neuromorphic chips. Global companies like Intel, Nvidia, Google, and Apple have also been investing in related M&A and R&D. Following its acquisition of Mobileye for US$15.3 billion, Intel acquired the Israel-based AI chip specialist company Habana Labs for US$2 billion. Nvidia recently acquired the UK semiconductor design company ARM for US$40 billion.

In China, major companies like Huawei and Alibaba have been boosting their AI semiconductor competitiveness under the slogan “strategizing next-generation AI developments.” The AI semiconductor market in China has been growing at an average annual rate of over 50%. Additionally, Taiwan, another major semiconductor power, has been pursuing government-led investment in AI processor chips and next-generation semiconductor design and process technology through its “AI semiconductor project” (2018-2021, US$132 million). A key example is Taiwanese foundry (manufacturer) TSMC, which has boosted its competitiveness by collaboration with US fabless (design) businesses.

Global companies are also concentrating investment to gain the advantage in key AI semiconductor markets such as cloud servers, smartphones, and automobiles. One particularly notable phenomenon has been the fast movement from non-semiconductor technology companies. Google has entered a server chip competition with Nvidia with the application of its independently developed Tensor Processing Unit (TPU) cloud processor to its data center, while Tesla has been pioneering automobile AI chip technology with the application of its own Full Self-Driving (FSD) chip to its new electric vehicle models. The startup Graphcore has reached the commercialization stage by combining its semiconductors with Microsoft and Facebook cloud platforms and drawing US$310 million in investment from Bosch and Dell. It also ventured into the South Korean AI server market last February.

By Cho Kye-wan, staff reporter

Please direct comments or questions to [english@hani.co.kr]

button that move to original korean article (클릭시 원문으로 이동하는 버튼)

Related stories

Most viewed articles