VoltageGPU vs Hyperstack

Short answer for the people who typed "hyperstack soc 2" or "hyperstack hipaa": its security page, read on 17 September 2026, states that SOC 2 Type 2 was achieved via NexGen Cloud, the parent company, and lists ISO 27001 and HIPAA as in progress. It states end-to-end encryption in transit and at rest and the availability of data centre regions in the EU, Canada and the US. The page does not mention GDPR processor terms. That is an honest, mid-build compliance file for a fast-growing GPU cloud, and it will likely be complete by the time some readers see this page; the source is re-read monthly.

What a SOC 2 report attests is that controls exist and were audited. What no report attests is a property of the hardware: whether an operator with console access could inspect a running VM. A standard GPU VM on Hyperstack, like on most clouds, runs on a hypervisor the operator controls. For public-model training, evaluation and pseudonymised pipelines, that is fine. For the step where identifiable data is decrypted in GPU memory under professional secrecy, health regulation or GDPR Article 9, the reviewer may ask for the technical measure to be enforced rather than documented. That is where this comparison begins.

Both products are a virtual machine with a NVIDIA GPU and root access. Hyperstack's runs on a standard hypervisor, bills per minute, and comes in EU, Canadian and US regions from a UK operator. It is a clean, cheap way to get an H100 or H200 for an afternoon, and for most work that is the whole requirement.

VoltageGPU's VM runs inside an Intel TDX trust domain with memory encryption, the GPU passed through into the domain, and /dev/tdx_guest in your hands. You generate an Intel TDX quote over 64 bytes you choose and verify it offline against Intel; on the verified SKUs you generate a NVIDIA GPU attestation report over your own nonce and verify it with NVIDIA. The contracting entity is VOLTAGE EI, a French sole proprietorship (SIREN 943 808 824). The evidence is published with checksums and an open-source verifier, voltage-verify. One hour is billed upfront and the unused part is refunded per second.

On list price Hyperstack is cheaper: $3.20 per hour for an H100 SXM and $3.99 for an H200 (no TEE, read 17 September 2026) against $5.00/hr and $6.58/hr for confidential VMs on VoltageGPU. The gap is the confidential stack, and for a non-sensitive workload it is not worth paying.

Hyperstack wins on price, on B200 availability, on spot VMs and reserved clusters, on per-minute billing without a prepaid hour, and on region choice. It also has a SOC 2 Type 2 report through its parent, which VoltageGPU does not have at all: we hold no SOC 2 and do not sign a HIPAA Business Associate Agreement today. If your procurement checklist starts with a report, Hyperstack clears that line and we do not, and you should say so to your reviewer.

For a UK buyer, contracting with a London company under UK law may itself be the requirement, and VoltageGPU, a French entity, does not change that.

A certificate says the operator is audited. An attestation says the hardware refused the operator. A buyer under bar-association secrecy, a clinic decrypting identifiable records for inference, a fintech scoring cardholder data, needs the second and cannot get it from a region or a report. A research team fine-tuning a public model needs the first and should not pay for the second.

The practical design for a regulated team is often both: Hyperstack for bulk capacity on non-sensitive data, and a VoltageGPU confidential VM for the step where the sensitive data is in the clear. This page exists so the line is visible before the purchase.