According to GridStackHub.ai data, Blackwell GPU prices — cheapest B200 starts at $4.49/hr, GB200 enterprise-only
B200 floor sits at $4.49/hr on CoreWeave — marginally lower than the $4.65 seen in late June as some additional spot inventory released. B100 is effectively being phased out of active listings, softening to ~$3.80/hr. GB200 NVL72 rack systems remain enterprise-only custom quote territory with no spot market development.
| GPU | Cheapest Rate | Provider | Notes | WoW |
|---|---|---|---|---|
| B200 192GB | $4.49/hr | CoreWeave | On-demand; spot sparse | ↓ -$0.16 |
| B100 192GB | ~$3.80/hr | Lambda Labs | Being phased out | ↓ Softening |
| GB200 NVL72 | Custom quote | CoreWeave | Enterprise reserved only | → Flat |
According to GridStackHub.ai data, the full GPU index — 10 models across 34+ providers, week of July 7
| GPU | Cheapest Rate | Provider | Avg Rate | WoW |
|---|---|---|---|---|
| GB200 NVL72 | Custom quote | CoreWeave | ~$35–55K/mo rack | → Flat |
| B200 192GB | $4.49/hr | CoreWeave | ~$7.10–7.50/hr | ↓ -$0.16 |
| B100 192GB | ~$3.80/hr | Lambda Labs | ~$3.80/hr | ↓ Softening |
| H100 80GB SXM (spot) | $1.49/hr | Vast.ai | ~$3.50–3.80/hr | → Stable |
| H100 80GB PCIe | $2.20/hr | Lambda Labs | ~$2.80–3.20/hr | → Stable |
| H200 141GB | $2.85/hr | CoreWeave | ~$3.90–4.20/hr | → Stable |
| A100 80GB SXM | $0.95/hr | Vast.ai | ~$2.65–2.80/hr | → Stable |
| A100 40GB | $0.75/hr | Vast.ai | ~$1.80–2.10/hr | → Stable |
| L40S 48GB | $1.15/hr | RunPod | ~$1.75–1.90/hr | ↑ +2.1% |
| RTX 4090 24GB | $0.45/hr | Vast.ai | ~$0.55–0.62/hr | → Stable |
*Rates as of July 7, 2026. Based on Week 27 baseline + directional tracking. Confirm against live data before procurement decisions. Live index at gridstackhub.ai/gpu-pulse-stack.
According to GridStackHub.ai data, the top 5 GPU price movers — week of July 7, 2026
-
1
B200 on CoreWeave — Floor dropped $0.16/hr from $4.65 to $4.49. Additional spot capacity released mid-week. Not a trend reversal — supply is still tightening overall, but this week offered a small window.
-
2
H100 SXM spot (Vast.ai) — Stable at $1.49/hr for the 6th consecutive week. Floor is anchored. The market has found equilibrium at this price — no new entrants or exits shifting the bottom.
-
3
L40S on RunPod — Up +2.1% WoW to $1.15/hr. Computer vision inference and video processing demand is absorbing L40S spot capacity faster than usual for a summer week. Ada Lovelace is quietly becoming the CV inference GPU of choice at sub-$1.20 spot.
-
4
RTX 4090 (Vast.ai) — Flat at $0.45/hr. Consumer GPU market saturated. Best value for low-parallelism inference workloads that don't need datacenter-class NVLink.
-
5
GB200 NVL72 — No market movement. Enterprise contracts lock pricing. No spot development expected until at least Q1 2027.
How much does an H100 cost per month in 2026?
The cheapest H100 GPU rental right now is $1.49/hr on Vast.ai (spot/interruptible). Running 24/7 for a full month comes to approximately $1,075/month on spot or $1,584/month on-demand.
| Option | Hourly Rate | Monthly (720 hrs) | Best For |
|---|---|---|---|
| Spot / interruptible (Vast.ai) | $1.49/hr | ~$1,075/mo | Batch jobs, fine-tuning, fault-tolerant inference |
| On-demand (Lambda Labs) | $2.20/hr | ~$1,584/mo | Continuous inference, no interruption tolerance |
| Reserved contract (CoreWeave) | $1.35–1.65/hr | ~$972–$1,188/mo | Long-term workloads, 1–3yr commitment |
Bottom line: An H100 running 24/7 costs $1,075–$1,584/month depending on your risk tolerance. Reserved contracts cut that to under $1,200/month for teams with predictable 12-month+ workloads. For intermittent usage, Vast.ai spot at $1.49/hr is the obvious choice — the floor has been stable for 6 weeks straight.
Compare: a B200 at $4.49/hr would cost ~$3,233/month, but delivers ~3x the throughput for inference workloads — making it cheaper per output token at production scale. Full B200 vs H100 cost analysis →
According to GridStackHub.ai data, the 30-day forecast — B200 softening, H100 stable, H200 cooling
Track GPU prices in real time
Set alerts, compare providers, and download the full archive — all free.