Compute Cheap H100/H200 GPU Cloud Pricing at $2.04/hr and $3.00/hr

Takeaway

Compute Cheap advertises on-demand pricing of $2.038 / GPU‑hour for NVIDIA H100 and $2.995 / GPU‑hour for NVIDIA H200, making it the cheapest publicly listed GPU cloud for these high‑end cards as of September 2 2026.

Pricing Overview

  • H100 on‑demand: $2.038 per GPU‑hour (updated Sep 2 2026).
  • H200 on‑demand: $2.995 per GPU‑hour.
  • Spot prices: H100 $1.187 / hr, H200 $1.773 / hr.
  • Launch size options: 1×, 2×, 4×, 8× GPUs.
  • Memory specs: H100 80 GB HBM (SXM), H200 141 GB HBM3e.

The site provides a live chart comparing these rates to the 31‑day market benchmark, refreshed daily.

Why the Prices Are Low

  1. Marketplace competition – Capacity owners bid for workloads, preventing fixed‑tier price floors.
  2. Utilizing idle capacity – Underused GPUs are offered at market‑driven rates rather than remaining idle.
  3. Reservation model – Users pre‑pay for reserved capacity at the on‑demand rate; there are no multi‑year contracts.
  4. Minimal service stack – The platform focuses solely on raw GPU access, avoiding the overhead of managed services.

These factors collectively reduce overhead and allow Compute Cheap to undercut traditional cloud providers.

Community Reactions

"I like thundercompute for my personal stuff, especially a A6000 for $0.35/h … $1.09/GPU/hr for a H100!!" – calgoo

"Seems fraudulent. Who’s the operator/owner? If it sounds too good to be true, it probably is." – otterley

"vast.ai has H100 at $1.93/h, H200 at $3.8/h but no $2500 minimum. This feels a bit comparatively dodgy." – yread

"Not the cheapest. Many providers have spot H100s for $8‑12, and H200s for slightly less than the above prices." – zaptrem

"This company would be more trustworthy if it listed an HQ, about page, history founder info, or anything..." – khelavastr

The comments highlight two themes:

  • Skepticism about legitimacy – Users question the operator’s identity and the feasibility of such low rates.
  • Price comparison – Some point out that other providers (e.g., vast.ai) already offer competitive H100 rates, though often with different minimum commitments.

Potential Compliance Concerns

One comment asked, "how do services like this stay compliant with export controls on these devices? I assume a credit card is all I need to access these things?" – Export regulations for high‑end GPUs (especially those with advanced memory) typically require verification of end‑user and destination. Compute Cheap’s public documentation does not detail its compliance workflow, which may be a risk factor for enterprises.

How to Evaluate Compute Cheap

  1. Verify operator identity – Look for corporate registration, physical address, or leadership information.
  2. Check compliance mechanisms – Ensure the provider conducts end‑user checks consistent with U.S. export control laws.
  3. Compare total cost of ownership – Include any hidden fees (e.g., data egress, storage) that may affect the advertised hourly rate.
  4. Test with a small workload – Validate performance, reliability, and support before committing large‑scale training jobs.

Conclusion

Compute Cheap’s advertised on‑demand rates of $2.038 / hr for H100 and $2.995 / hr for H200 are among the lowest publicly listed prices for these GPUs, achieved through a marketplace model that monetizes idle capacity and minimizes service layers. However, prospective users should scrutinize the provider’s legitimacy, compliance practices, and hidden costs before relying on these rates for production‑grade AI workloads.

Sources

Related