H200
141 GB SXM5$2.50/hr
- AWS$10.90
- Lambda$4.49
- CoreWeave$6.31
save up to 77%
01 — The rate card
Six chips.One honest meter.
Indicative on-demand rates, June 2026 — final price confirmed at launch. Per-second billing on everything. There is no fine print; this is the fine print.
02 — Versus the usual suspects
Same silicon. Public rate cards. You do the math.
H100
80 GB SXM5$1.99/hr
- AWS$7.57
- Lambda$2.99
- CoreWeave$4.76
save up to 74%
RTX PRO 6000
96 GB PCIe 5.0$1.80/hr
- RunPod$2.49
- Vast.ai$2.20
- CoreWeave$2.80
save up to 36%
A100
80 GB SXM4$1.40/hr
- AWS$3.77
- Lambda$1.79
- CoreWeave$2.21
save up to 63%
RTX 5090
32 GB PCIe 5.0$0.80/hr
- RunPod$0.94
- Vast.ai$1.02
- CoreWeave$1.20
save up to 33%
RTX 4090
24 GB PCIe 4.0$0.39/hr
- RunPod$0.69
- Vast.ai$0.55
- CoreWeave$0.80
save up to 51%
03 — The fine print
No fine print worth hiding.
- Per-second billing
- You pay for compute you use, metered by the second. Stop the box and the meter stops with it — mid-epoch, midnight, whenever.
- The price is the price
- One number per GPU. No egress drama, no “regional adjustment,” no surprise line items wearing acronyms.
- Credits, not contracts
- Top up credits and spend them on any GPU. No minimum commitments, no reserved-instance hostage math.
- Bulk & long-running
- Need 10–200 GPUs for 30 days to 6 months? Reserved clusters with prepay discounts that actually move the number.