Built for AI verification research
Lucid Computing built the Verifiable Studio for researchers who need more than a login and a GPU. A multi-silicon cluster in the Bay Area, on bare metal or inside confidential computing, with fiber taps, side-channel probes, software-defined radios, and network analyzers on site, and engineers who will design and run the measurement with you.
The Verifiable Studio
The Verifiable Studio is a cluster for AI verification research.
The cluster is Slurm. Researchers log in over SSH, request GPUs, and run jobs on bare-metal NVIDIA and AMD accelerators. Home directories live on shared storage mounted on every node. From optical fiber taps to side-channel probes to network analyzers, we can add the instrumentation your experiment needs.
Researchers can arrive with a fully planned experiment or work with us through designing the idea.
The studio is in beta.
AI verification researchers
The studio is for anyone doing serious work on AI verification, anywhere in the world: academic labs, nonprofits, startups, industry teams, and government teams. If your work is about proving what compute is doing, such as attestation, hardware-enabled mechanisms, side channels, or verifiable reporting, this is built for you. Some researchers qualify for access at no charge through the Research Access Program.
The full cluster
Two H100 nodes are operational today through the beta. Scheduled additions bring the full cluster to nine nodes across NVIDIA and AMD silicon, with the official release on November 1, 2026. See the Verifiable Studio specifications for the latest.
DGX H100
8 × NVIDIA H100 SXM
H100 NVL
1 × NVIDIA H100 NVL
HGX H200
8 × NVIDIA H200 SXM
HGX B300
8 × NVIDIA B300 SXM
MI355X
8 × AMD Instinct MI355X
HGX A100
8 × NVIDIA A100 SXM
Talk it through with us
Every Tuesday we hold an open hour to talk through research ideas and deployment possibilities. Bring a half-formed experiment, a question about the hardware, or a plan you want to sanity check. No preparation needed.
If you are deciding whether the cluster fits your work, this is the fastest way to find out.