SoftBank and Intel Launch Strategic Partnership on Advanced Memory

SoftBank-backed Saimemory has entered a major technology partnership with Intel, aiming to bring a new class of high-performance, energy-efficient memory to market as artificial intelligence workloads continue to surge worldwide.

The collaboration centers on developing next-generation memory designed specifically for AI systems and high-performance computing environments. Known internally as the Z-Angle Memory program (ZAM), the initiative seeks to overcome the growing limitations of traditional Dynamic Random-Access Memory (DRAM), which is increasingly struggling to keep pace with the data-intensive requirements of modern AI models.

According to SoftBank, working prototypes are expected by the fiscal year ending March 31, 2028, with full commercialization targeted for fiscal 2029. The announcement was met positively by investors, with SoftBank shares rising over 3% and Intel gaining around 5% in after-hours trading, signaling market confidence in the long-term potential of the project.

Saimemory itself was established in December 2024 as a dedicated SoftBank unit focused on advanced memory technologies, underscoring the group’s broader strategy to position itself at the center of the global AI hardware ecosystem.

What Makes Z-Angle Memory Different

At the core of the partnership is Intel’s work on new memory architectures and assembly techniques that aim to significantly improve both performance and power efficiency.

Intel engineers have been developing these technologies through earlier initiatives, including research conducted under the U.S. Department of Energy’s Advanced Memory Technology program. That effort focused on creating foundational breakthroughs in memory design, particularly for servers and data centers where power consumption and latency are critical bottlenecks.

Under ZAM, Intel brings its expertise in DRAM optimization, while Saimemory provides commercialization pathways and strategic backing from SoftBank. The companies say the new architecture can deliver:

• Higher bandwidth and faster data access for AI workloads
• Lower power consumption per operation
• Reduced overall system costs at scale
• Improved manufacturability for future mass deployment

Intel leadership has emphasized that conventional memory designs are no longer sufficient for large-scale AI, noting that their new approach combines architectural innovation with advanced packaging to unlock better performance while simultaneously lowering energy usage.

The goal is to create memory solutions that can support everything from cloud-based AI training clusters to edge computing and next-generation supercomputers over the coming decade.

Responding to Exploding AI Memory Demand

The timing of the partnership reflects mounting pressure across the global memory supply chain.

Demand for AI-related hardware has exploded over the past two years, driven by rapid adoption of generative AI, large language models, autonomous systems, and enterprise automation. These applications require massive volumes of high-speed memory, often operating continuously in power-hungry data centers.

As a result, memory supply has struggled to keep up, leading to tight inventories and rising prices across several segments of the DRAM and high-bandwidth memory market.

Industry analysts estimate that AI servers can require several times more memory than traditional enterprise systems, while hyperscale data centers are now among the world’s largest consumers of electricity. Against this backdrop, energy efficiency has become just as important as raw performance.

The ZAM program directly addresses this challenge by prioritizing lower power consumption alongside higher throughput, aiming to reduce the operational costs and environmental footprint of AI infrastructure.

A Broader Ecosystem Effort

The Intel–Saimemory collaboration is part of a wider push to rethink how memory is designed and deployed for AI-era computing.

In addition to Intel and SoftBank, Japan’s Fujitsu is also reported to be involved in the broader project, contributing systems integration and computing expertise. This multi-company approach highlights how complex next-generation memory development has become, requiring coordination across chip design, packaging, manufacturing, and system architecture.

For SoftBank, the partnership aligns closely with its long-term vision of building an AI-centric technology portfolio spanning semiconductors, computing platforms, and data infrastructure. For Intel, it represents another step in strengthening its position in advanced memory and AI hardware after several years of aggressive investment in foundry services and next-gen architectures.

Looking Ahead: From Prototype to Production

With prototypes expected in 2028 and commercialization planned for 2029, the Z-Angle Memory program is a medium-term bet on where AI hardware is headed.

Over the next few years, development will focus on validating performance gains, refining manufacturing processes, and integrating the new memory into real-world computing platforms. If successful, the technology could become a foundational component of future AI data centers, supercomputers, and enterprise systems.

As artificial intelligence continues to reshape industries from healthcare and finance to manufacturing and transportation, memory innovation is emerging as one of the most critical enablers of progress. The SoftBank–Intel partnership signals a clear message to the market: the race to power AI isn’t just about processors anymore. It’s increasingly about smarter, faster, and greener memory.

‍