Rent H200 141GB
Compare live on-demand and spot rental prices across 97+ cloud providers. H100 die with HBM3e memory upgrade. Same compute, 76% more VRAM, 43% more bandwidth than H100.
Choosing the right billing model for H200 141GB
Provision and terminate at any time. Ideal for development, short experiments, and workloads with unpredictable duration.
Instances can be reclaimed when demand spikes. Best for fault-tolerant batch jobs, training with checkpointing, and preprocessing.
Lock in a rate for 1–3 months. Right for sustained production inference or long training runs where cost predictability matters.
Best use cases
- 70B+ model inference
- Memory-bound LLM serving
- Large context windows
H200 141GB — Specs & Benchmarks
Performance bars, compute tiers (FP32/FP16/BF16/FP8/INT8), memory specs, LLM model size guidance, and related GPU comparisons.
H200 141GB Rental Guide
The H200 141GB is justified when you need to run 70B parameter models at FP16 on a single card, or when memory bandwidth is the primary bottleneck for your inference workload. The 141GB HBM3e fits Llama 3.1 70B at FP16 (140GB) on a single card, eliminating the need for two-card tensor parallelism. For 70B inference, this typically reduces latency by 30–50% compared to a two-H100 setup, while also reducing infrastructure cost.
Spot instances for H200 are available from a limited set of providers including CoreWeave and Lambda Labs at 35–55% below on-demand rates. Availability is more constrained than H100 — check provider availability before planning a workload around H200 spot. On-demand H200 is appropriate for production 70B inference serving where single-card latency matters.
When comparing H200 vs H100 for your workload, the decision comes down to whether memory capacity or bandwidth is your bottleneck. If your model fits in 80GB and you are not bandwidth-limited, the H100 delivers identical compute at lower cost. If you are running 70B at FP16, using large context windows, or serving high-batch-size inference where bandwidth matters, the H200's 4,800 GB/s is worth the premium. Use the GPU cost calculator to model total cost across H100 and H200 for your specific throughput requirements.
Frequently Asked Questions
How much does it cost to rent a H200 141GB?
H200 141GB on-demand rental prices vary by provider and region. On-demand rates typically range based on availability and provider margins — use the comparison table above to see current live rates across all providers. Spot instances are generally 40–70% cheaper than on-demand but can be interrupted. Monthly cost estimates (hourly rate × 730 hours) are shown in the table for sustained workloads.
Which cloud provider has the cheapest H200 141GB?
The cheapest H200 141GB provider changes as providers update their pricing. The comparison table above shows live rates sorted by price, so the cheapest option is always at the top. Factors beyond headline price include region (latency to your users), availability (high/medium/low), and billing granularity (per-second vs per-hour minimums).
What can I run on a H200 141GB?
With 141GB of HBM3e, the H200 141GB can run LLM models up to approximately 70B parameters at FP16, 141B at INT8, or 282B at INT4/GGUF quantization. Common workloads include: 70B+ model inference, Memory-bound LLM serving, Large context windows. H100 die with HBM3e memory upgrade. Same compute, 76% more VRAM, 43% more bandwidth than H100.
Should I use on-demand or spot pricing for H200 141GB?
Spot instances save 40–70% vs on-demand but can be interrupted when the provider needs capacity back. Use spot for: batch inference jobs, training runs with checkpointing, preprocessing pipelines, and any workload that can tolerate interruption and restart. Use on-demand for: production inference serving, interactive workloads, and jobs that cannot be interrupted. Most providers bill per second, so short on-demand jobs are not penalized by hourly minimums.
How does the H200 141GB compare to the H100 for cloud rental?
The H100 80GB delivers 1,979 TFLOPS FP16 with 3,350 GB/s HBM3 bandwidth, compared to the H200 141GB's 1979 TFLOPS FP16 and 4800 GB/s bandwidth. The H100 is significantly more expensive — typically $2.50–$5.00/hr vs lower rates for the H200 141GB. For workloads that fit within 141GB and don't require FP8 precision, the H200 141GB often delivers better cost-per-token than the H100.
What is the memory bandwidth of the H200 141GB and why does it matter?
The H200 141GB has 4800 GB/s of memory bandwidth. For LLM inference, memory bandwidth is often more important than raw TFLOPS — each autoregressive token generation reads the full model weight matrix from VRAM, so bandwidth directly determines tokens-per-second throughput. Higher bandwidth means faster inference for the same model at the same batch size. For batch inference (processing many requests simultaneously), compute throughput becomes more important.
Can I use the H200 141GB for Stable Diffusion or image generation?
Yes — the H200 141GB is well-suited for Stable Diffusion and image generation workloads. Image generation is primarily FP32 and FP16 compute-bound, and the H200 141GB's 67 TFLOPS FP32 throughput determines images-per-second. The 141GB VRAM fits SDXL (requires ~6GB) and most ControlNet pipelines. For high-throughput image generation at scale, compare cost-per-image across providers using the GPU cost calculator.