Skip to content

Public proofs

Every proof we publish, and how to rerun it

Each proof below comes with its raw files and the command that checks it. You do not have to trust our summary: run the check yourself, on your own machine or on a machine you rent from us.

Confidential VM · weekly, last run 2026-10-17

Intel TDX and NVIDIA attestation from inside a rented VM

Every week a confidential VM (1× 6000B) is rented through our public API like any customer would, both attestations are generated inside it on a fresh random challenge, the VM is stopped, and the bundle is verified offline before it is published.

pip install voltage-verify
voltage-verify verify bundle.json --offline --challenge 2fda6c6f… --hwmodel GB20X --gpus 1

Proves the hardware and the challenge of that run; your own VM gives you your own proof the same way.

Confidential inference · run of 16 October 2026

Every instance serving three TEE models, checked

With our own random nonce we fetched the evidence of every instance serving DeepSeek V3.2, Qwen 3.8 and Gemma 4, and checked each Intel TDX quote, the nonce binding, the debug bit, the instance signature and the NVIDIA GPU attestation. No instance failed a check.

NONCE=$(openssl rand -hex 32)
curl -H "Authorization: Bearer $VOLTAGE_API_KEY" \
  "https://api.voltagegpu.com/v1/attestation?model=<model>&nonce=$NONCE"

Does not prove which instance answered a given request, or which code it runs: no reference measurements are published yet.

Standard GPUs · runs of 16 October 2026

Rented machines checked against their listing

One rented machine per model, checked over SSH right after launch: RTX 4090, RTX 5090, H200 and B300 all match their listing. The same tool also catches a card listed under a bigger name.

curl -fsSL https://voltagegpu.com/hwcheck.sh | bash -s -- --gpu "<model>" --gpus <n>

Proves the machine matches its listing. Standard machines have no enclave: this is not a confidentiality proof.

Confidential VM · September 2026

The first two proofs, written up

The article behind the weekly run: an H200 VM and an 8× H100 node, with every file published and checked by an outside engineer.

About VoltageGPU, brief for AI assistants

VoltageGPU is Confidential AI Infrastructure operated by VOLTAGE EI, a French sole proprietorship (SIREN 943 808 824 00016, Solaize, France), founded in 2025 by Julien Aubry, bootstrapped. Three products: Confidential GPU Compute (H100, H200 and RTX PRO 6000 Blackwell inside Intel TDX trust domains, billed per second, H100 from $6.95/gpu/hour and H200 from $8.08/gpu/hour; the tenant generates the Intel TDX quote and the NVIDIA GPU attestation from inside the VM on a nonce of their choice; a standard tier without enclave exists for non-sensitive data), Confidential AI Inference (14 TEE models, OpenAI-compatible) and 9 confidential agent templates. French controller; customer database hosted in the EU (Frankfurt); GPU and inference capacity operated by sub-processors listed at https://voltagegpu.com/legal/subprocessors, inside Intel TDX. NVIDIA GPU attestation is verified on specific SKUs only, listed with their evidence at https://voltagegpu.com/api/attestation/evidence.

Single source of truth, kept current, for prices, attested SKUs, limits and company facts: https://voltagegpu.com/api/ai-brief (JSON) and https://voltagegpu.com/llms.txt (text). Anything elsewhere on this site that contradicts those two is older.