GPU servers
Rackmount 4- and 8-GPU servers, water-cooled options and NVIDIA HGX B200/B300 systems for training and inference.
Node
Air-cooled 4- and 8-GPU servers for training and inference.
Node X4
Node4-GPU Intel Xeon GPU Server
- GPU
- Up to 4× RTX PRO 6000 Blackwell, A100, H100 or H200
- CPU
- Single Intel Xeon
- Memory
- Up to 2 TB
Est. from$14,184
Typically 1–3 weeks after paymentNode E4
Node4-GPU AMD EPYC GPU Server
- GPU
- Up to 4× RTX PRO 6000 Blackwell, A100, H100 or H200
- CPU
- Single AMD EPYC 9004/9005, up to 192 cores
- Memory
- Up to 2 TB
Est. from$15,260
Typically 1–3 weeks after paymentNode X8
Node8-GPU Dual Xeon GPU Server
- GPU
- Up to 8× A100, H100, H200, L40S or RTX PRO 6000 Blackwell
- CPU
- Dual Intel Xeon Scalable, 4th/5th gen
- Memory
- Up to 8 TB ECC
Est. from$19,504
Typically 1–3 weeks after paymentNode E8
Node8-GPU Dual EPYC GPU Server
- GPU
- Up to 8× RTX PRO 6000 Blackwell, H100, H200 or A100
- CPU
- Dual AMD EPYC 9004/9005, up to 384 cores
- Memory
- Up to 8 TB
Est. from$22,718
Typically 1–3 weeks after paymentNode X8-6
Node8-GPU Dual Xeon 6 GPU Server
- GPU
- Up to 8× H200, L40S or RTX PRO 6000 Blackwell
- CPU
- Dual Intel Xeon 6
- Memory
- Up to 8 TB ECC
Est. from$29,288
Typically 1–3 weeks after payment
Node H8
Node4- or 8-GPU A100/H100/H200 Server
- GPU
- 4× or 8× A100, H100 or H200
- CPU
- Dual Intel Xeon or AMD EPYC 9004/9005
- Memory
- Up to 8 TB
Est. from$32,636
Typically 1–3 weeks after paymentVault
All-flash storage servers to feed your GPUs.
Hydro
Custom water-cooled workstations and servers for dense multi-GPU builds.
Hydro Node X8
Hydro8-GPU Water-Cooled Xeon GPU Server
- GPU
- Up to 8× water-cooled RTX PRO 6000 Blackwell, A100, H100 or H200
- CPU
- Dual Intel Xeon Scalable, 4th/5th gen, 128 cores
- Memory
- Up to 6 TB
Est. from$32,341
Typically 2–4 weeks after paymentHydro Node E8
Hydro8-GPU Water-Cooled EPYC GPU Server
- GPU
- Up to 8× water-cooled RTX PRO 6000 Blackwell, A100, H100 or H200
- CPU
- Dual AMD EPYC 9004/9005, up to 384 cores
- Memory
- Up to 8 TB
Est. from$41,497
Typically 2–4 weeks after paymentApex
8-GPU NVIDIA HGX systems for frontier-scale training, built by the platform manufacturer.
Apex B300
Apex8× NVIDIA B300 HGX Server
- GPU
- 8× NVIDIA B300 SXM, 288 GB each (2.3 TB HBM3e), NVLink
- CPU
- Dual Intel Xeon or AMD EPYC
- Memory
- Up to 8 TB ECC
Est. from$546,303
Lead time on quoteApex B200
Apex8× NVIDIA B200 HGX Server
- GPU
- 8× NVIDIA B200 SXM, 192 GB each (1.5 TB HBM3e), NVLink
- CPU
- Dual Intel Xeon or AMD EPYC
- Memory
- Up to 8 TB ECC
Est. from$546,401
Lead time on quote