Hardware verification, side-channel analysis, TEE research, and network inspection require physical, bare-metal access to datacenter-grade GPUs that no cloud provider offers.
So the Verifiable Compute Foundation is providing it — built and operated by Lucid — and opening it to the community in November.
| Machine | Quantity |
|---|---|
| NVIDIA B300 | 1 |
| NVIDIA H200 | 5 |
| NVIDIA H100 | 1 |
| NVIDIA A100 | 1 |
| AMD MI355X | 2 |
From side-channel analysis of GPU workload behavior to probing TEE boundaries under real datacenter conditions, we can help you explore how systems leak and reveal information across every layer. This includes NIC-level experiments and controlled network manipulation on frontier-class hardware.
Need a different setup? We can reconfigure to meet your research needs.
Qualifying safety and verification research groups receive compute access at no cost on the Verifiable Compute Foundation's cluster, allocated by the Foundation through an open fair-share process and covered by research funders; Lucid builds and operates the hardware on the Foundation's behalf. If your needs exceed the free allocation, dedicated remote capacity is also available.
We host weekly virtual office hours every Tuesday from 10 to 11 AM Pacific Time to answer questions and discuss research and deployment possibilities.