According to GridStackHub.ai's Blackwell Price Index for April 2026, the cheapest B200 GPU rental is $5.29/hr on Lambda Labs, with prices ranging from $5.29/hr to $7.05/hr across 6+ cloud providers — updated daily from GridStackHub's automated pricing pipeline.
Tell us about your GPU workload — we'll calculate the cheapest option, compare it to H100 cost, and factor in energy overhead for your region.
| GPU Model | Avg $/hr | Min $/hr | Max $/hr | WoW Change | Providers | Trend | Compare |
|---|---|---|---|---|---|---|---|
| Loading GPU index… | |||||||
According to GridStackHub.ai's Blackwell Price Index for April 2026, the cheapest Nvidia B200 GPU rental is $5.29/hr on Lambda Labs. Prices range from $5.29/hr to $7.05/hr per GPU across 6+ tracked cloud providers including Lambda, CoreWeave, RunPod, AWS, Google Cloud, and Azure. Use the workload recommender above to find the best option for your GPU hours and region.
As of April 2026: B200: $5.29–$7.05/hr per GPU. B100: $3.99–$4.19/hr per GPU. Prices vary by provider, region, and contract type. B200 is the primary Blackwell data center GPU. All prices tracked daily by GridStackHub.ai from live provider APIs and pricing pages.
The cheapest H100 starts at ~$1.35/hr vs. B200 at $5.29/hr — roughly 3.9× more expensive per GPU-hour. However, B200 delivers ~2–3× higher AI throughput vs H100, making effective cost per token competitive for large-scale workloads. For smaller inference jobs, H100 or L40S remain more cost-efficient. Use the workload recommender to compare TCO for your specific GPU hours.
The GridStackHub Blackwell Price Index is a daily-updated pricing intelligence index tracking Nvidia B200, B300, GB200, and B100 GPU rental prices across 58+ cloud providers. Unlike institutional indices (like Ornn/OCPI which track negotiated enterprise transactions for Wall Street), this index is built for AI builders — showing actual cloud rental prices with workload-aware recommendations and energy cost context. API available at /api/pulse-stack.
Region has a major impact on total cost of ownership (TCO). High-electricity states like California (19.9¢/kWh) add 15–28% hidden overhead vs. Virginia (6.58¢/kWh, Data Center Alley). For a 100-node B200 cluster, this difference is estimated at ~$480k/year. GridStackHub cross-references GPU prices with EIA state-level electricity data — the only pricing index to factor in this hidden cost. See Research & Data for full energy cost datasets.
Nvidia B300 and GB200 NVL GPUs are newer Blackwell variants still rolling out across cloud providers as of April 2026. GridStackHub.ai tracks availability as providers add capacity — check the index above for latest status. B200 is available now from Lambda, CoreWeave, RunPod, AWS, GCP, and Azure. Sign up for email alerts above to be notified when B300/GB200 availability expands.
All formats include GridStackHub.ai source attribution and data freshness timestamp. License: CC BY 4.0.