Intel and SoftBank Develop New Power-Efficient DRAM Technology for AI Data Centers

Intel and SoftBank have announced a collaboration to develop a next-generation DRAM technology aimed at making AI data centers more power-efficient. The project, known as Saimemory, could mark a significant shift in memory design by offering an alternative to current high-bandwidth memory (HBM) solutions.

Project Saimemory: A New Direction for DRAM

The joint venture, Saimemory, is focused on creating a stacked DRAM product that consumes around half the power of today’s HBM chips. If successful, this kind of efficiency could drastically cut energy costs and reduce the carbon footprint of large-scale AI infrastructure.

Combining Global Expertise

The project brings together Intel’s advanced semiconductor technology and patents from Japanese academic institutions, including the University of Tokyo. This mix of commercial and academic expertise aims to fast-track innovation in the memory space.

• Prototype target: 2027
• Commercialisation goal: By the end of the decade

Strategic Importance for Japan

Japan was a dominant force in the memory chip market during the 1980s but has since fallen behind. Saimemory could help re-establish Japan as a key player, particularly within its domestic AI and data centre sectors.

With only a few companies—Samsung, SK hynix, and Micron—currently producing advanced HBM, there’s room for disruption. A successful rollout could give Japanese data centres a homegrown alternative and attract global interest.

Why It Matters

The demand for AI processing power continues to grow, and that means more memory is needed. But HBM comes with high power requirements, which makes it expensive to run at scale. A more efficient alternative could:

• Lower operating costs for data centres
• Reduce environmental impact
• Ease supply chain pressures around HBM

Market Implications

If Saimemory succeeds, it could challenge the market dominance of existing HBM leaders. That could lead to more competitive pricing, better innovation, and diversified supply chains—especially important as demand for AI-capable infrastructure skyrockets.

What’s Next?

With a prototype expected by 2027, it’ll be a few years before we see real-world applications. Still, the collaboration between Intel and SoftBank signals a serious investment in reshaping how memory fits into the AI data centre ecosystem.

For more details, you can read the original article on Tom’s Hardware.