Data Center NVIDIA

NVIDIA H100 PCIe

Hopper Architecture · 80GB HBM2e · PCIe

VRAM
80GB
FP16
102.0
TDP
350W
Hardware Price
$28k
MSRP: $25k
Cloud from
$2.39/hr
3 providers
Cheapest at RunPod →

Quick Insights

Performance/Dollar
3.64 TFLOPS/$k
FP16 performance per $1000
VRAM/Dollar
2.9 GB/$k
VRAM per $1000
vs Data Center Average
-43% perf
FP16 TFLOPS comparison
Cloud Availability
3 providers
from $2.39/hr

Specifications

VRAM 80GB HBM2e
Memory Bandwidth 2.0 TB/s
FP16 TFLOPS 102.0
Tensor TFLOPS 1.5k
FP32 TFLOPS 51.0
TDP 350W
Form Factor PCIe
Architecture Hopper
NVLink No
Release Date 2022-09

H100 PCIe Price and Rental Notes

H100 PCIe searches include current price, used H100 PCIe price in 2026, and cloud cost. This page should preserve the primary H100 intent while giving buyers safer used-price and rental context.

Used H100 PCIe price in 2026

Used H100 PCIe quotes can look attractive, but buyers should verify memory health, warranty transfer, cooling configuration, and whether the card is locked to a specific server platform.

H100 PCIe cloud price baseline

Cloud H100 PCIe pricing is useful as a benchmark even for hardware buyers. If your expected use is below the break-even estimate, renting can avoid depreciation, power, cooling, and maintenance costs.

Best fit

H100 PCIe is strongest for AI training, fine-tuning, and high-throughput inference where mature CUDA support and 80GB HBM memory matter more than the lowest hourly cost.

Related: H100 rental cost calculator Compare H100 cloud prices MI300X high-VRAM alternative H100 PCIe vs A100 80GB

Buy vs Rent Analysis

Buy Hardware
$28k
  • One-time cost, unlimited usage
  • Full control over hardware
  • Electricity & cooling costs extra
  • Depreciation over 2-3 years
Best if using >11715 hours total
Rent Cloud GPU
$2.39/hr
  • Pay only for what you use
  • No upfront investment
  • Scale up/down instantly
  • No maintenance required
Best for <11715 hours or variable usage
Breakeven Point
11,715
hours of usage

At $2.39/hr cloud pricing, buying the hardware pays off after 11,715 hours (~488 days or 16.3 months of 24/7 usage).

Usage Monthly Cloud Cost Months to Breakeven
100 hrs/month $239.00 118 months
200 hrs/month $478.00 59 months
500 hrs/month $1.2k 24 months

Cloud GPU Pricing

Rent NVIDIA H100 PCIe from 3 cloud providers. Prices shown per GPU per hour.

Provider Type Instance GPUs On-Demand Per GPU Spot Availability
RunPod gpu-cloud NVIDIA H100 PCIe 1x $2.39/hr $2.39/hr Cheapest $1.25/hr (-48%) -
Lambda Labs gpu-cloud lambda-h100-pcie 1x $2.49/hr $2.49/hr - -
CoreWeave gpu-cloud coreweave-h100-pcie 1x $4.25/hr $4.25/hr - -
Best Spot Deal: RunPod offers spot pricing at $1.25/hr (48% off on-demand).

H100 PCIe vs Alternatives

Compare NVIDIA H100 PCIe with similar GPUs from other brands.

GPU VRAM FP16 TFLOPS Bandwidth Hardware Price Cloud Price
H100 PCIe Current 80GB 102.0 2.0 TB/s $28k - -
AMD Instinct MI210 AMD 64GB (-20%) 181.0 (+77%) 1.6 TB/s - - Compare
MI300 AMD 128GB (+60%) 490.3 (+381%) 5.3 TB/s $15k (-46%) - Compare
AMD Instinct MI300A AMD 128GB (+60%) 980.0 (+861%) 5.3 TB/s - - Compare
AMD Instinct MI250X AMD 128GB (+60%) 383.0 (+275%) 3.3 TB/s - - Compare
AMD Instinct MI250 AMD 128GB (+60%) 362.0 (+255%) 3.3 TB/s - - Compare

Best Use Cases

No specific use case recommendations for NVIDIA H100 PCIe yet.

Browse All Use Cases →

Compare H100 PCIe

Other NVIDIA GPUs
Alternatives
Add to Comparison Tool

Frequently Asked Questions about H100 PCIe

Check warranty status, seller history, memory error logs where available, server compatibility, cooling requirements, and whether the price advantage beats cloud rental for your expected usage.

H100 PCIe is often easier to deploy in standard servers, while SXM platforms can deliver stronger multi-GPU bandwidth. Compare full system cost, not only the GPU card price.

The NVIDIA H100 PCIe has a market price of approximately $28k (MSRP: $25k). Cloud rental starts at $2.39/hr. Prices may vary based on retailer, region, and availability.

Yes, the NVIDIA H100 PCIe with 80GB VRAM is suitable for many AI/ML workloads. For large language models, you may need multiple GPUs or consider higher-VRAM options like A100 or H100.

The breakeven point is approximately 11,715 hours of usage. Buy if you'll use it more than this; rent for shorter projects or variable workloads. Cloud rental from RunPod starts at $2.39/hr.

With 80GB VRAM and 102.0 FP16 TFLOPS, the NVIDIA H100 PCIe can run: Large language models (7B-13B), Stable Diffusion XL, video AI, and professional 3D rendering.

The NVIDIA H100 PCIe offers 80GB VRAM and 102.0 FP16 performance at $28k. Compare with similar GPUs using our comparison tool above. Key factors: VRAM for model size, TFLOPS for speed, and price for budget.