- 1 Top cover + fan wall
- 2 8x GPU tray (NVLink)
- 3 Chassis + power backplane
Server
Built to order
8x H200 HGX server
The next generation up from the H100 server — same 8-GPU NVLink design, close to double the memory per GPU.
SKU
SRV-H200X8-1128G
PriceiChecked against NVIDIA's own product listings and major retailers (B&H, Micro Center, NVIDIA's marketplace) as of July 2026 — real published prices, not estimates.
$370,000
Specifications
GPU memory
1.1 TB HBM3eiAn upgraded version of HBM3 with more capacity and bandwidth per GPU. Used in newer datacenter GPUs like the H200 and the Blackwell Ultra chips in the GB300.
Power draw
5.6 kW
Tensor / AI compute
~32 petaflops FP8 (8-GPU aggregate, sparse)
Rack units
8U typical (air-cooled 8-GPU HGX chassis; varies by system integrator)
Per-GPU memory141 GB HBM3e
Per-GPU bandwidth4.8 TB/s
GPU-to-GPU linkNVLink, 900 GB/s
Aggregate FP8 compute~32 petaflops
Which models this can run
Real memory math against the 3 models in the model advisor — parameters × 1 GB, plus 20% working room.
Qwen3.5-397B
476.4 GB required · this unit has 1.1 TB
Runs on a single unit.
DeepSeek-V3.2
805.2 GB required · this unit has 1.1 TB
Runs on a single unit.
Kimi K2.6
1,200.0 GB required · this unit has 1.1 TB
Needs 2x units clustered together (not a single unit).
Power and cooling
Electrical requirement: Draws 5,600 W — 3x what a single standard 120V/20A circuit can safely carry continuously (1,920 W). Requires 208V/240V power distribution and, at this scale, dedicated cooling — standard in datacenters, not offices or homes.
About 4.7x the continuous draw of an average home
(1200 W). Run flat-out for a month, that's roughly
4,032 kWh — about $524/month
at a $0.13/kWh commercial rate (US average, EIA — your
actual rate depends on region and utility contract).
Power cost, relative to the rest of the catalog
$524/mo
Cooling: Air-cooled chassis fans; liquid-cooled variants available from some integrators
What this is good for
- Same workloads as the H100 server, with more memory headroom per GPU
- Larger context windows or bigger batch sizes at the same GPU count
- Training runs that were memory-constrained on H100
What this is NOT good for
- Small teams or single-user workloads
- Cases where H100 pricing and availability already meet the need
Included components
Complete server: 8x H200 GPUs, dual host CPUs, redundant power supplies, and NVLink baseboard. Does not include networking switches, cabling, or rack enclosure.
Lead time and warranty
Delivery window estimateBuilt-to-order enterprise hardware — typically an 8–16 week lead time from order confirmation, depending on allocation.
WarrantyManufacturer warranty applies (enterprise service terms vary by reseller).