SoftBank And Intel Set To Launch $70 Million AI Memory Project In Japan To Slash Power Use

SoftBank Group Corp. SFTBY and Intel Corp. INTC are developing next-generation artificial intelligence memory chips designed to consume half the power of current technology, addressing critical energy challenges in Japan’s growing data center market.

What Happened: The companies established Saimemory, a joint venture targeting commercialization within the 2020s through a 10 billion yen ($70 million) project, reported Nikkei. The initiative focuses on stacked DRAM chips with innovative wiring structures that significantly reduce power consumption compared to existing high-bandwidth memory (HBM).

SoftBank leads the investment with 3 billion yen, while Japan’s Riken research institute and Shinko Electric Industries are considering participation. The venture leverages Intel’s technology and patents from the University of Tokyo and other Japanese academic institutions, the report noted.

The project addresses supply chain vulnerabilities as South Korea’s SK Hynix and Samsung Electronics dominate HBM production. Current HBM technology suffers from low production yields, high costs, and intense energy requirements, creating access challenges for Japanese companies.

“We want priority for the supply,” a SoftBank executive said, according to the report, highlighting the strategic importance for the company’s AI data center plans.

Why It Matters: This development represents Japan’s attempt to regain semiconductor prominence after losing over 70% DRAM market share in the 1980s to South Korean and Taiwanese competitors. Elpida Memory, Japan’s last DRAM manufacturer, was acquired by Micron Technology Inc. MU in 2013 following bankruptcy.

SoftBank plans to deploy the new memory technology in AI training data centers, where processing large datasets requires substantial computing power. The company recently acquired Sharp’s LCD factory for $676 million to build a 150-megawatt AI data center in Osaka, supporting its OpenAI partnership and broader AI infrastructure strategy.

Read Next: