In the past fifty years, chip design and EDA software in the United States, semiconductor materials in Japan, memory in South Korea, photolithography equipment in Europe, and wafer foundry in TSMC-Union have become the regional strengths of global semiconductorization, with supply chain clustering effects. However, the Sino-U.S. technological competition, the new crown epidemic, and the Russia-Ukraine conflict have broken this semiconductor globalization.
Korea is in a hurry!
In June 2023, South Korean President Yoon Seok-yul held a unique state meeting. “Chip semiconductors are the lifeline of Korea and the core of our industry. Korea’s semiconductors account for 20 percent of the overall export size and are the foundation of our economy.”
Korea’s Ministry of Science and ICT announced on June 27, 2023, that the Korean government will invest 1.02 trillion won (approximately $5.2 billion) in the research and development of cutting-edge artificial intelligence (AI) semiconductor technologies and expand joint research with AI pioneer countries on PIM (memory processing) semiconductors, next-generation neural network processors (NPUs), and system software. In addition, Korea plans to launch a large-scale data center chip project in 2024.
Subsequently, the South Korean government launched a program to support chip development, hoping to attract a total of 622 trillion won ($471 billion) in private sector investment from 2024 to 2047 through a series of financial incentives to build the world’s largest “supercluster” of chip factories and R&D facilities.
Japan is in a hurry!
In June 2023, Japan’s Ministry of Economy, Trade and Industry (METI) released a revised Semiconductor and Digital Industry Strategy, which plans to triple the sales of Japan’s domestically produced semiconductor industry by 2030 to 15 trillion yen (approximately $108 billion).
Japan has divided the revival of the semiconductor industry up to 2030 into three phases:
(1) Accelerate infrastructure development for semiconductor production.
(2) Cooperating in the development of next-generation semiconductor technologies.
(3) Develop disruptive semiconductor technologies based on existing technologies.
The Japanese government’s strategy does hit the core of the decline of Japan’s semiconductor industry, but also from between the lines kind of feels that the previous development of digital chips left behind “remorse”.
How high is the bar for the chip industry?
Objectively speaking, the threshold of the chip industry is mainly the commercial threshold, followed by the technical threshold. Chip products have a natural scale reproducibility, the first entrant is king, the winner takes all, as long as the price and profit control in place, will not give the latter any opportunity.
From a business perspective, the semiconductor chip industry has passed its peak, is a monopolistic industry that is going downhill, the rise of Silicon Valley in the United States in the past few decades, the birth of the golden age of the semiconductor industry, the industry’s initial stage through a large amount of capital investment to complete the basic technology, ecology, IP and supply chain construction, with the United States Intel, Nvidia, Qualcomm, TI and other chip companies as the leader, firmly Intel, Nvidia, Qualcomm, TI and other chip companies in the United States as the leader, firmly monopolized the entire industry, with absolute control and the right to speak, but also led the high-speed development of the IT field in the past few decades, creating the miracle of the times and great social wealth.
Moore’s law has failed, the development of chips has also decelerated, and even the development of chips in some industries has exceeded the application requirements. 28nm is already sufficient for many industries, 5nm and 2nm technologies only have application value in a small number of miniaturization and power consumption of very sensitive industries, and the investment threshold in 5nm and 2nm technologies is getting higher and higher, the supply chain is getting more and more fragile, and there is a technology overload already.
The core problem in high-end chips such as 5nm and 2nm is the input-output ratio, and the demand is very certain.
For relatively low-end chips (such as 28nm), the core issue is the cost and ecology, The chip business depends on the scale and ecology to support, if there is no scale chip company simply can not survive, maintain the scale of the moat is ecological, ecological construction needs to be corresponding to the timing of the entry and the cycle, there is no suitable entry opportunities or no build ecological allies are unlikely to succeed, both are indispensable.
Chip production and manufacturing process links, high precision requirements, especially suitable for “concentrated efforts to do big things”, a device can raise a, and can only raise a large international company.
Under the general trend of AI big model, GPU chips with big computing power have become the hottest now, and the dual pattern of GPU chips in China and the United States has been initially formed. With the outbreak of the AI big model, the global technology giants have launched fierce competition in the field of AI over the past year. Tesla CEO Elon Musk compared the AI arms race to a high-stakes poker game, saying that companies involved in the competition need to invest billions of dollars in AI hardware every year to remain competitive.
In terms of the expected future market size of AI chips, according to the latest forecast, by 2024 the AI chip market size will grow 25.6% from the previous year to reach $67.1 billion, and it is expected that by 2027, the AI chip market size is expected to be more than twice the size of 2023, reaching $119.4 billion.
Musk said Tesla will spend more than $500 million on Nvidia AI chips in 2024 alone, but warned that Tesla would need “billions of dollars” worth of hardware to catch up with its biggest competitors.
“While $500 million is a lot of money, it only equates to about 10,000 NVIDIA H100 chips.” Musk said in a post on X. “Tesla will spend more than that on NVIDIA hardware this year. To be competitive in AI, you have to invest at least billions of dollars every year.”
Despite NVIDIA’s dominance in the AI chip space, AI chips developed by its competitors are also growing in popularity.2023 In December, tech companies such as Meta, OpenAI, and Microsoft said they would be using AMD’s newly developed AI chip, the Instinct MI300X- a chip that directly benchmarks Nvidia’s H100.
AMD reported fourth-quarter fiscal 2023 results, with PC chips, data center chips, and chips for the gaming business all coming in below expectations, while AI chip (MI300) sales exceeded expectations by $400 million, and are expected to reach $3.5 billion in sales in 2024, up from $2 billion in previous forecasts.
Qualcomm and Arm’s “patent war” broke out, Qualcomm to RISC-V layout more actively. Manju Varma, Qualcomm’s director of product management, said that Qualcomm has already built many products based on RISC-V, and by the end of 2022, the shipment of Qualcomm’s chips using the RISC-V architecture will have exceeded 650 million units. RISC-V microcontrollers are currently used in Qualcomm’s SoCs for PCs, mobile devices, wearables, connected cars and AR/VR headsets. It is predicted that by 2027, there will be 25 billion RISC-V-based AI SoCs on the market, with revenues expected to reach $291 billion in the same year.
RISC-V is a chip architecture developed at this point, born in the era of the chip supply chain game, the previous monopoly faced the U.S. government’s various restrictions, had to give up some market opportunities, the potential customers also vaguely feel that the past chip supply chain there is a huge risk of uncertainty, but also more or less to support the development of RISC-V, or at least hope to have a new supplier such as RISC-V appeared, it should be said that RISC-V was born at the right time. RISC-V is born at the right time.
RISC-V is oriented to open source and free of charge, in the field of software open source and free of charge has been verified as a successful ecological construction mode, there are mature sets of methods, and through the open-source of the software to cultivate a large number of open source operation personnel, quite several companies in dealing with open source operation has also been mature experience and team. RISC-V is not subverting the chip technology but is subverting the traditional chip supply chain model. RISC-V is not subverting the chip technology but is subverting the traditional chip supply chain model. The era of excess chip technology gives time and space for the operation model to change, and the RISC-V architecture chips will be gradually applied in various industries, just like the rise of ARM in the mobile field in those years.
More than a decade has passed since the birth of RISC-V, which started in a lab at the University of California, Berkeley. The emerging instruction set architecture, which is often compared to X86 and Arm, is realizing its potential in more applications such as the Internet of Things and the cloud, and has attracted some companies to set up shop: Imagination, an IP vendor, launched its first commercial RISC-V cores in the summer of 2022 for use in system-on-chip (SoC) designs; and Intel announced its official membership in the RISC-V International Association in February 2022, as well as its investment of €400 million in Spain’s supercomputing industry. In February 2022, Intel announced that it had officially joined the RISC-V International Association and invested 400 million euros in cooperation with the Spanish Supercomputing Center, with plans to develop RISC-V-based supercomputing CPUs within 10 years; Google will launch a new open-source operating system for RISC-V chips; and in 2023, the world’s first mass-produced RISC-V single-board computer with an integrated 3D GPU, Fang-Starlight 2, will be born. In 2023, the world’s first mass-produced RISC-V single-board computer with integrated 3D GPUs will be born.
The difficulty in industrializing RISC-V is that there is no clear market entry point, i.e., a ‘killer’ application scenario. Just as in the first few decades of the emergence of smartphones, the development of ARM is also relatively slow.
The outbreak of global competition in science and technology has made Chinese chip enterprises lose the “nutrients” of foreign advanced technology and equipment, but they have an independent self
The main development of a broad “soil”. In the face of a huge domestic demand market, China’s chip boom will certainly continue in the future.
