NVIDIARelease: 2022-01-01
H100 SXM
Ideal for high-performance computing (HPC) applications and AI workloads.
ArchitectureNVIDIA Hopper™
Cheapest Rent$1.9/hr
Hardware Intelligence Dossier
memory gb
80
tgp watts
-
cuda cores
-
architecture
NVIDIA Hopper™
precision compute
3,958 teraFLOPS / 1,979 teraFLOPS / 67 teraFLOPS / 34 teraFLOPS / 3,958 TOPS / 989 teraFLOPS / 1,979 teraFLOPS
memory bandwidth gbps
3.35
Workload Suitability
The H100 SXM variant is best suited for AI training workloads, particularly for large language models and high-performance computing tasks.
Deep Learning Training
Scientific Simulation
High-Performance Computing
Asset Insights
"Preliminary specifications. May be subject to change."
Last Intel Refresh:February 2026
Verified Market Rates
Pricing Matrix
| Provider | Stock Status | Market Price | Network Link |
|---|---|---|---|
Hyperstack Verified Node2d ago | High Availability | $1.90per hour | Inspect Listing |
Hyperstack Verified Node1d ago | High Availability | $1.90per hour | Inspect Listing |
Oblivus Verified Node2d ago | High Availability | $2.94per hour | Inspect Listing |