"Four suppliers, real USDC, no accounts: the full results of our x402 GPU test"
On October 2, 2026, we completed the second live verification of the x402 pay-per-job service on Kilawatt Cloud. In Wave 1, every request went to a single supplier. This time we sent each request to a different one, so e
On October 2, 2026, we completed the second live verification of the x402 pay-per-job service on Kilawatt Cloud. In Wave 1, every request went to a single supplier. This time we sent each request to a different one, so every supplier could be evaluated on its own.
As before: no simulation. An autonomous agent requested GPU capacity, paid in USDC on Base, and received machines with no account, no API key, and no human intervention. We are publishing everything, including the request that failed.
The test
Four placements across three additional suppliers: Lambda, Hyperstack, and RunPod. Each request was paid for on Base, ran for ten minutes, and went to the production endpoint. Every payment was a real USDC transaction, $4.88 in total.
Results
| # | Supplier | Config | GPUs booked | GPUs ready | Time to ready | Paid | Outcome |
|---|---|---|---|---|---|---|---|
| 1 | Lambda | 8x A100 | 8 | 8 | 5m 14s | $3.40 | Delivered |
| 2 | Hyperstack | 2x A100 PCIe | 2 | 2 | 6m 15s | $0.58 | Delivered |
| 3 | RunPod | 8x A40 | 8 | 0 | No capacity | $0.60 | Not delivered, refunded |
| 4 | RunPod | 4x A40 (retry) | 4 | 4 | 13s | $0.30 | Delivered |
- Delivery rate: 3 of 4 placements delivered, which is 14 of 22 requested GPUs.
- Time to ready: delivered jobs reached a healthy state between 13 seconds and 6 minutes 15 seconds.
- Combined with Wave 1: all four suppliers (Vast.ai, Lambda, Hyperstack, and RunPod) have now completed a successful multi-GPU x402 job, with 30 GPUs delivered in total.
What did not go perfectly
The first RunPod request, for eight A40 GPUs, could not be fulfilled because the supplier had no matching capacity. The platform returned the $0.60 payment automatically, as account credit, without any manual action. We then sent a follow-up request for four A40 GPUs, matched to confirmed availability, and it was fulfilled.
The refund was issued as account credit, not as USDC sent back to the wallet.
What this tells you, and what it does not
At the other suppliers, where the requested configuration was out of stock, the nearest in-stock multi-GPU configuration was used. That is why Hyperstack shows 2x A100 PCIe and not a larger node. So this test shows each supplier can deliver multi-GPU capacity through the payment flow. It does not show that each supplier can deliver the exact configuration you want at any moment.
Limits of this test
- Small sample: four placements is not a benchmark. Treat the numbers as live data points.
- Different configurations: the jobs were not identical, so the times are not a like-for-like speed comparison between suppliers.
- Platform records: the figures come from Kilawatt Cloud's own records. The final RunPod job was running and healthy at the last status check, and we did not get independent confirmation from any supplier that each machine was terminated.
Read the full reports
Every live verification run we do is published in full, with the provider, GPU count, result, time and cost for each job, including the ones that failed. You can read all of them on our Super Intelligence page:
https://kilawattcloud.dev/super-intelligence
If you are building agents that need compute on demand, we would like to hear what you would want to see in a test like this.
Kilawatt Cloud: https://kilawattcloud.dev
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.