Rapidus CEO Atsuyoshi Koike will get a boost from Japan’s CHIPS Act and IBM to start the nation’s sole semiconductor foundry making the world’s most advanced silicon just two years behind industry heavyweight Taiwan Semiconductor Manufacturing Co. (TSMC), he said in an exclusive interview with EE Times in Brussels.

Koike and his team of more than 100 people are taking on the challenge of a lifetime.

Koike is a chip-industry veteran who recently worked for Western Digital and founded Trecenti, a Japanese foundry that didn’t take off, two decades ago. That company failed because it was too closely tethered to chipmaker Hitachi, Koike said.

He hopes his second chance, Rapidus, will realize his dream of creating a foundry that provides fast cycle time for customers.

“My idea is single-wafer processing,” Koike said. “It means the fab does not have any stock, so I can manufacture very precisely to control the cycle time.”

Koike aims to turn the standard practice of processing hundreds of wafers at a time on its head. By extracting data from a single wafer rather than hundreds, Rapidus plans to cut cycle time by eliminating production problems rapidly.

The company will also accelerate output using wafer bonding, a technique that’s just beginning to win industry adoption.

“We can manufacture differently by attaching one wafer to another to shorten cycle time,” Koike said. “That’s kind of a new idea.”

Rapidus will count on help from Japan’s CHIPS Act and an alliance with IBM to start production of 2-nm chips by 2027. That’s after TSMC launches its 2-nm process in 2025.

Rapidus is leading Japan’s global partnership for R&D and mass production of next-gen semiconductors, according to Satoshi Nohara, an official with Japan’s Ministry of Economy, Trade and Industry (METI). Nohara is responsible for implementing the nation’s CHIPS Act.

Japan and the U.S. have been working to reduce dependence on China in the semiconductor supply chain, and Rapidus is an outgrowth of that partnership.

“The step to global partnership for R&D and mass production of next-generation semiconductors is the most important part of our strategy,” Nohara said at a May ITF World event held by Interuniversity Microelectronics Centre (imec) in Brussels. “The Rapidus, imec and IBM partnership is its first project.”

Imec is set to help Rapidus develop building blocks necessary for the mass production of 2-nm chips that can be used for 5G communications, quantum computing, data centers, autonomous vehicles and digital smart cities.

$37 billion needed to begin

Rapidus will need to invest about $37 billion to start production, Koike said. Neither he nor METI’s Nohara disclosed how much the Japanese government will provide in subsidies.

TSMC’s new $8.6 billion chip fab in Japan will receive a METI subsidy accounting for 40% of the cost, according to a Foreign Policy article. Rapidus will get a subsidy from METI worth about $2.5 billion, according to a Reuters report.

Koike said the financial support from the Japanese government may not be as generous as other places like the U.S. and Europe that are implementing their own CHIPS Acts.

“Government support for the semiconductor industry is strong,” Koike said. “I think in Japan, not so good.”

Major Japanese companies like Sony, NTT Communications and a joint venture created by Toyota and Denso will invest in Rapidus, Koike said. He didn’t disclose the size of the investment.

Rapidus plans to do assembly and test, including heterogeneous integration, in its fab. That could also streamline cycle time.

“Japan has so many great equipment and materials companies for the backend,” Koike said. “They are willing to join us.”

IBM affirmed its partnership with Rapidus.

“We’re confident that our work with Rapidus will be successful and efficiently implemented,” Bethany McCarthy, an IBM spokesperson, told EE Times. “We recently welcomed the first of many Rapidus researchers who are joining us on-site at Albany NanoTech.”

IBM will help Rapidus achieve its 2-nm target, according to Koike.

“That is a key issue because Japan’s logic technology is 10 to 15 years behind,” he said.

Rapidus doesn’t plan to compete directly with TSMC in the foundry business, he added.

“We would like to (target) a specific fabrication market for this technology,” he said, noting that focus areas include high-performance computing and edge computing combined with AI.

Koike’s eyes brighten at the mention of automotive chips.

“Automotive is one of the chances,” he said. “Edge computing is potentially one of the good opportunities.”

Analyst weighs chances of success

The chances of success for Rapidus are good given the prudent strategy of the Japanese government, Paul Triolo, who advises global tech clients at Albright Stonebridge Group, said in an exclusive interview with EE Times.

“Japan has taken one of the more strategic and balanced approaches,” he said. “They’re not trying to jump and outcompete TSMC. They’re saying, ‘We want to have these companies to be players in the next generation of technology.’”

The main challenge for Rapidus will be to commercialize IBM’s 2-nm technology announced in May 2021, a world’s first by the company’s semiconductor research facility in Albany, New York.

“They have to work very closely with IBM on this very complicated technology that’s a little bit off the main roadmap,” Triolo said.

It can’t hurt that Kasunari Ishimaru is among the roughly 100 people on Koike’s team.

Ishimaru used to be the head of R&D at Kioxia, which Koike said will also be a Rapidus partner. Ishimaru also had experience working with IBM developing process technology from when he was with Toshiba.

“With Toshiba, I joined the IBM alliance to develop 32-nm-and-beyond technology with high-K metal gate,” Ishimaru told EE Times last month in Brussels.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved