TensorDock vs Thunder Compute
Thunder Compute is cheaper on 2 of the 3 GPU configurations both providers list with on-demand pricing, updated 2026-10-02. TensorDock still wins on one, its best case being the H100 SXM · node count not reported at $2.25 vs $3.20. Platform facts come from each provider's documentation.
TensorDock
cheaper on 1 of 3 sharedThunder Compute
cheaper on 2 of 3 sharedPrices are per GPU-hour. Rows match exact GPU variants and published node minimums. When node count is not reported, confirm the required configuration before comparing whole-job costs. These are listed rates, not a guarantee of stock.
Price per GPU, side by side (2026-10-02)
| GPU (on-demand $/hr) | TensorDock | Thunder Compute | Price diff |
|---|---|---|---|
| RTX A4000 · node count not reported | $0.10 | not offered | – |
| Tesla V100 SXM2 · node count not reported | $0.17 | not offered | – |
| RTX 3090 · node count not reported | $0.20 | not offered | – |
| RTX A6000 · node count not reported | $0.45 | $0.35 | -$0.10 (-22.2%) Thunder Compute |
| RTX 4090 · node count not reported | $0.35 | not offered | – |
| RTX 6000 Ada · node count not reported | $0.75 | not offered | – |
| L40 · node count not reported | $0.95 | $0.79 | -$0.16 (-16.8%) Thunder Compute |
| A100 (unspecified) · node count not reported | not offered | $1.09 | – |
| A100 PCIe 80GB · node count not reported | $1.50 | not offered | – |
| A100 SXM 80GB · node count not reported | $1.80 | not offered | – |
| H100 SXM · node count not reported | $2.25 | $3.20 | -$0.95 (-29.7%) TensorDock |
Lowest recorded on-demand rate per configuration; highlighted = cheaper. Price diff is the absolute and percent saving on the cheaper side (named), relative to the higher price; – means one side has no matching configuration in the collected rates. Spot/marketplace tiers shown on the individual pages.
Tables rank by price, lowest first. Commercial relationships never affect ordering.How we collect and rank prices.
Platform facts
| TensorDock | Thunder Compute | |
|---|---|---|
| Billing increment | not recorded | not recorded |
| Egress | not recorded | not recorded |
| Storage /GB-mo | not recorded | not recorded |
| Spot offered | not recorded | not recorded |
| Reserved offered | not recorded | not recorded |
| Notebook support | not recorded | not recorded |
From each provider's documentation, updated 2026-07-07; sources on the fees matrix and enterprise matrix.
FAQ
Thunder Compute is cheaper on 2 of the 3 GPU configurations both providers list, updated 2026-10-02. TensorDock is cheaper on 1; 0 have equal rates. Compare the configuration you need.
The H100 SXM · node count not reported: $2.25 per GPU-hour on TensorDock vs $3.20 on Thunder Compute, a difference of $0.95 per hour, updated 2026-10-02.
TensorDock lists 7 GPU configurations with on-demand pricing that Thunder Compute does not: RTX A4000 · node count not reported, Tesla V100 SXM2 · node count not reported, RTX 3090 · node count not reported, RTX 4090 · node count not reported, RTX 6000 Ada · node count not reported, A100 PCIe 80GB · node count not reported and 1 more. Thunder Compute lists 1 that TensorDock does not: A100 (unspecified) · node count not reported.
Compare more rental options: the screener · price movers · methodology.