Once a powerhouse in the semiconductor industry, Japan is trying to become a leader in chip manufacturing again. Chip companies such as TSMC and Micron have pledged hundreds of millions of dollars to build new chip manufacturing plants across the country, and have received strong support from Tokyo.
Now, Japan’s chip policy may have scored another victory, with Nvidia announcing that it will work with Japanese companies to build a new factory network in the country focused on producing graphics processors for artificial intelligence.
“Japan has all the technical expertise and industrial capabilities to create its own artificial intelligence,” Nvidia CEO Jensen Huang said during a meeting with Japanese Economy Minister Yasunari Nishimura on Tuesday, according to Bloomberg.
At an earlier meeting, Huang promised Japanese Prime Minister Fumio Kishida that the chipmaker would try to supply key artificial intelligence processors to Japan. The surge in interest in generative AI products has led to a tight supply of the processors used to power them.
Japan’s most advanced chip factories lag market leaders TSMC and Samsung by about a decade, according to an August report from the Center for Strategic and International Studies. Japanese producers who dominated dynamic random access memory (DRAM) in the 1980s have largely exited the industry, but the country remains competitive in some areas such as NAND memory and image sensors. Tokyo understands it may have only “one last chance” to gain a strong position in the global chip industry, the report said.
Japan announced its semiconductor strategy in June 2021, which includes a partnership with the United States and subsidies for domestic chip manufacturing. On November 24, the Japanese government passed an additional $13.5 billion in funding to support the country’s chip manufacturing industry as part of an additional $88 billion budget to help households cope with inflation and encourage companies to increase domestic investment.
“The semiconductor industry that Japan is now starting to develop and cultivate will be able to produce GPUs,” Huang said on Monday. “Countries like Japan are realizing that you need to own your own data, build your own AI factories, and produce your own AI.”
Tokyo Electronics, Japan’s top chipmaker, has pledged to invest about $290 million to build a new development facility in Koshi, a city on the southwestern island of Kyushu. The island will also be home to a new TSMC foundry, worth approximately $6.7 billion, which will open late next year. US chipmaker Micron Technology has also pledged to invest approximately US$3.4 billion in Japan over the next few years.
Nvidia is at the center of the artificial intelligence boom because its processors are key to training the large language models that underpin the technology. The supercomputer used to train OpenAI’s ChatGPT uses 10,000 Nvidia A100 GPU chips. Nvidia’s data center revenue, which is tied to customer interest in artificial intelligence, hit $14.51 billion in the latest quarter, up 279% from the same period last year. Data center revenue now accounts for about 80% of Nvidia’s quarterly revenue, a significant change from two years ago, when the company’s gaming business drove the company’s sales.