Fujitsu and Nvidia’s NVLink Fusion partnership fuses MONAKA CPUs with Nvidia GPUs to power sector-specific AI agents today and supercharge FugakuNEXT by 2030.
Japan’s tech titans are doubling down on AI. Fujitsu and Nvidia just announced an expanded collab to build a full-stack AI infrastructure, fusing Fujitsu’s MONAKA CPUs with Nvidia GPUs over NVLink Fusion to deliver real-world AI agents today—and supercharge tomorrow’s supercomputer, FugakuNEXT, by around 2030.
Table of Contents
Getting Real with AI Agents
This ain’t your grandma’s server setup. Fujitsu’s throwing its next-gen MONAKA CPU tech into Nvidia’s GPU party via NVLink Fusion—think a super-fast expressway straight to the GPU city.
The goal? Ready-to-roll AI agents for:
- Healthcare: Agents that read scans, cross-check records, and spot patterns faster than any human doc.
- Manufacturing: Digital factory copilots handling planning, predictive maintenance, and motion planning in one shot.
- Robotics: Smarter perception-to-action pipelines in logistics and service bots, leveraging Japan’s robotics prowess.
Nvidia CEO Jensen Huang didn’t mince words: “Japan can lead the world in AI and robotics.”
He’s on a mission to build local, sovereign AI infrastructure at scale, and this partnership is step one of that playbook.
NVLink Fusion: CPU+GPU, No Strings Attached
Until now, NVLink mostly played nice with Nvidia’s own CPUs.
Opening it up to Fujitsu’s MONAKA chips means:
- Custom AI Factories: Deploy massive GPU clusters with your choice of CPU—no vendor lock-in.
- Slick Performance: Direct CPU-to-GPU links slash latency and boost throughput for continuous learning workloads.
In short, it’s like ripping out stoplights on the data highway.
DeepSeek Releases V3.2-Exp: Revolutionizing AI with Sparse Attention Technology
Discover how DeepSeek’s new model leverages innovative sparse attention mechanisms to enhance AI efficiency and performance.
Read MoreFugakuNEXT: Japan’s 2030 Showstopper
Fujitsu, RIKEN, and Nvidia are co-designing FugakuNEXT, slated for around 2030.
For the first time, Japan’s flagship will mix GPUs and CPUs in a hybrid AI-HPC platform, aiming for 5–10× the muscle of the current Fugaku and multi-exaflop FP64 peaks, all within a ~40 MW power envelope.
What’s Next?
- 2025–2027: Pilot AI-agent rollouts with Japanese enterprises, early software co-design, and MONAKA-X CPU reveals.
- 2028–2030: FugakuNEXT hardware deployment, full-scale AI-HPC fusion, and mass rollout of sector-specific agents.
This partnership isn’t just hype—it’s Japan’s bid to own the next decade of AI and robotics, from the hospital floor to the factory line to the supercomputer center.
