Fujitsu
Fujitsu carries real HPC pedigree (Fugaku, its own MONAKA Arm CPU, in-house liquid cooling), but its current Blackwell flagship is a Supermicro OEM product wrapped in Fujitsu services. The sovereign Made-in-Japan line is built at its own plant, still co-engineered with Supermicro. Watch the MONAKA platform for the return of real Fujitsu server design.
- Current flagship Blackwell server PRIMERGY GX2570 M8s (HGX B200) is a Supermicro OEM product: Fujitsu wraps SupportDesk maintenance and Infrastructure Manager tooling around a Supermicro design
- Earlier GPU generations (GX2570 M6 with HGX A100, GX2560 M7 with HGX H100) were Fujitsu-designed PRIMERGY servers with NVIDIA-Certified listings
- The Feb 2026 Made-in-Japan sovereign AI server line (HGX B300, RTX PRO 6000) is produced at Fujitsu's own Kasashima Plant but co-engineered with Supermicro
- Genuine in-house engineering on cooling (Cool-Central liquid cooling, Fugaku heritage) and silicon: the FUJITSU-MONAKA 2nm Arm CPU
- Co-developing Arm-based liquid-cooled servers around MONAKA with Supermicro for 2027, the platform to watch for real Fujitsu design depth
| Model | Platform | Form factor | Cooling |
|---|---|---|---|
| PRIMERGY GX2570 M8s Supermicro OEM product under the Fujitsu brand, available July 2025 with Fujitsu SupportDesk | HGX B200 | 10U air / 4U liquid | air or DLC |
| Sovereign AI server (name TBD) Made-in-Japan production from March 2026 at Kasashima Plant, built with Supermicro | HGX B300 | TBD | TBD |
| PRIMERGY GX2560 M7 4th Gen Xeon host, PCIe Gen5 | HGX H100 (4-GPU) | rack node | air |
| PRIMERGY GX2570 M6 NVIDIA-Certified, dual 3rd Gen Xeon | HGX A100 (8-GPU) | 4U | air |
- NVIDIA Partner Network member; long-standing NVIDIA collaborator in HPC, AI and VDI
- Multiple PRIMERGY models are NVIDIA-Certified Systems (GX2570 M6, GX2460 M1, RX2540 M5/M6)
- Oct 2025: expanded strategic collaboration integrating FUJITSU-MONAKA CPUs with NVIDIA GPUs via NVLink Fusion
- Not found in NVIDIA's public GB200 NVL72 or Vera Rubin rack-scale partner lists as of mid-2026
- Kasashima Plant, Japan: dedicated sovereign AI server line from March 2026, leveraging Fugaku manufacturing expertise
- Standard global GX2570 M8s SKU is Supermicro-built OEM product
- Buyer distinction: the global Blackwell SKU is a Supermicro rebadge with Fujitsu services; the sovereign line is Japan-built but jointly engineered with Supermicro
- Fujitsu's differentiated bet is MONAKA CPU plus NVIDIA GPU over NVLink Fusion, not the current PRIMERGY GX boxes
- Explicit Japan digital-sovereignty positioning tied to critical-infrastructure procurement
Is Fujitsu an OEM, ODM or integrator for NVIDIA AI servers?
Fujitsu sits in the "Global OEMs" tier of our index. Current flagship Blackwell server PRIMERGY GX2570 M8s (HGX B200) is a Supermicro OEM product: Fujitsu wraps SupportDesk maintenance and Infrastructure Manager tooling around a Supermicro design
Does Fujitsu design its NVIDIA servers in-house?
Fujitsu carries real HPC pedigree (Fugaku, its own MONAKA Arm CPU, in-house liquid cooling), but its current Blackwell flagship is a Supermicro OEM product wrapped in Fujitsu services. The sovereign Made-in-Japan line is built at its own plant, still co-engineered with Supermicro. Watch the MONAKA platform for the return of real Fujitsu server design.
Which NVIDIA platforms does Fujitsu ship?
Fujitsu ships systems on HGX B200, HGX B300, HGX H100 (4-GPU), HGX A100 (8-GPU). The systems table on this page lists the specific models.
Where does Fujitsu manufacture its AI servers?
Kasashima Plant, Japan: dedicated sovereign AI server line from March 2026, leveraging Fugaku manufacturing expertise. Standard global GX2570 M8s SKU is Supermicro-built OEM product