AIBRIEF
Tokyo backed 2 nanometer chip foundry says AI data center demand is running ahead of plan; volume production still slated for late 2027 into early 2028.
Japan's state-backed chipmaker Rapidus, racing to mass-produce 2-nanometer logic chips, is sticking to its plan to start volume production in the second half of Japanese fiscal 2027 (late 2027 into early 2028), and said AI-chip and data-center demand is running ahead of expectations, CEO Atsuyoshi Koike told Reuters.
Tokyo is preparing to put another roughly ¥150 billion (about US$944 million) behind the bet. The Ministry of Economy, Trade and Industry will seek the top-up in the fiscal 2027 budget, Nikkei reported, as the government works to recapitalize the firm and crowd in lending from Japan's top banks.
Rapidus has already cleared the technical gate. The company confirmed a 2nm gate-all-around transistor prototyping milestone at its Chitose foundry pilot line, the prerequisite for the planned H2 FY2027 ramp. On June 5, the company closed a ¥150 billion round from the Information-Technology Promotion Agency, bringing cumulative funding to roughly ¥425 billion.
The customer pipeline is still the open test. Tenstorrent was named as the first 2nm client in February 2024, and TrendForce reported in October that IBM is now at the front of the US queue. Whether those order books translate into commercial volume at yield by early 2028 is the load-bearing question.