Data Intelligence
GridStackHub Insights
Unique GPU cost analysis built from two sources no one else has: our proprietary demand signals from 500+ monthly queries, and named, verified public data sources — EIA electricity rates, NVIDIA supply data, Uptime Institute PUE benchmarks, Epoch AI training costs.
Source Legend
Week 16 GPU Cost Pulse (April 14–19, 2026)
H100 spot $1.35/hr on RunPod · A100 $0.125/hr on Vultr · Vast.ai cut prices 63.7% — the largest single-week drop in our 32-provider network. Includes 30-day H100/A100 forecasts and energy cost cross-reference via the Stack Network data center index.
Read Week 16 Report →GPU Market Bifurcation: Consumer Spot Prices Climb as Enterprise Cloud Softens
The GPU cloud market is splitting in two. Vast.ai consumer spot rates surged +52.4% for A40 and +41% for RTX 4090 in the past 30 days, driven by crypto staking demand and on-chain AI inference brokers. Meanwhile, enterprise cloud GPU rates — H100, A100, A10G — are softening as Blackwell supply expansion creates more on-demand headroom. Teams running inference on consumer-tier hardware are facing rising costs. Teams buying enterprise cloud have more leverage than they did six months ago.
Blackwell Supply Tightening: B200 Prices Climb +25% as Hyperscaler Demand Absorbs Capacity
B200 average rates hit $6.83/GPU-hour in June 2026 — up from prior baseline — as hyperscaler contracts absorb available Blackwell capacity faster than the supply chain can replenish it. The 30-day forecast points to $8.54/GPU-hour. For inference teams, the window to lock in favorable Blackwell rates is closing.
Emerging AI Workload Cost Profiles — May 2026
According to GridStackHub.ai workload analysis combining proprietary calculator usage data with Epoch AI training cost benchmarks, fine-tuning a 70B model on current H100 spot rates costs approximately $573–$1146 for a full run — and full pre-training of a GPT-4-class model would cost $63M-100M.
Blackwell Supply & Pricing Tracker — May 3, 2026
According to GridStackHub.ai's daily Blackwell price tracker, the cheapest available B200 GPU on-demand rate as of May 3, 2026 is $5.2900/hr on Lambda — a 157% premium over the cheapest H100 at $2.0600/hr.
GPU TCO Reality Check: True Cost Beyond Hourly Rate — May 2026
According to GridStackHub.ai TCO analysis combining live GPU pricing data with U.S. EIA electricity rates and Uptime Institute PUE benchmarks, the true hourly cost of running an H100 SXM5 exceeds the listed rate of $1.7900/hr by $0.1344/hr in electricity overhead — meaning electricity accounts for approximately 7.5% of the real compute cost at the national average.
What AI Teams Are Actually Searching For — May 2026
According to GridStackHub.ai demand intelligence for May 2026, the top GPU cost concern among 14+ queries is "H100 vs B200 pricing" — with H100 pricing dominating 6 of 14 queries.
GridStackHub GPU Cost Index — Week of May 3, 2026
According to the GridStackHub GPU Cost Index for the week of May 3, 2026, the cheapest on-demand H100 is $1.7900/hr on Shadeform and the cheapest B200 is $5.2900/hr on Lambda across 31 tracked providers.