Built by one engineer
VoltageGPU is a solo-founder company building confidential AI infrastructure from France.
Mission
AI is eating sensitive data. Law firms paste NDAs into ChatGPT. Accountants upload client financials. Clinics share patient records.
We build infrastructure where the enclave, not the operator, holds the keys. Hardware attestation you verify, not promises.
If you handle regulated data, you need compute that proves, cryptographically: it can't leak.
Products
Confidential Compute
Confidential GPU VMs (H200, RTX PRO 6000 Blackwell) on Intel TDX, with AES memory encryption and trust-domain GPU isolation.
Confidential AI Inference
TEE models with an OpenAI-compatible API. Models run inside Intel TDX enclaves at our inference provider; requests go through our API, and we do not claim we cannot see them.
Confidential AI Agents
9 domain-expert agents for legal, finance, compliance, medical, M&A, cyber, HR, tax and sovereign legal work, running inside hardware enclaves.
How it works
Your workload runs in an Intel TDX trust domain whose memory the CPU encrypts. On the Confidential VM you generate the attestation yourself and verify it against Intel and NVIDIA, not against us.
Company
Julien Aubry
Founder & CEO
Observation Regulated professionals need AI but can't trust cloud providers with client data.
Bet Hardware enclaves solve the trust problem, if the customer can verify the attestation instead of taking the operator's word.
VoltageGPU became A platform where AI runs inside Intel TDX enclaves, with attestation you verify yourself.
Vision Make confidential AI the default for every regulated industry.
Solaize, France
Founded
2025 · Solaize, France
Legal
VOLTAGE EI · SIREN 943 808 824
Infrastructure
Intel TDX · NVIDIA CC on the VM tier · French company, EU law
We serve
Law firms, accountants, clinics, fintech
Governance & team scaling
We believe procurement teams deserve a straight answer about who is behind the company they are about to trust. Here is the honest version.
Today
VoltageGPU is currently solo-founded and bootstrapped by Julien Aubry, supported by external counsel and the operations teams of our infrastructure providers, listed in /legal/subprocessors. No outside investors, no board of directors.
Q3 2026, Fractional DPO
Engagement of a fractional Data Protection Officer to formalize Art. 37 GDPR governance ahead of the first regulated-customer DPAs. Engagement letter will be linked on the Trust Center once signed.
Q4 2026, SOC 2 auditor
Auditor selection in progress. SOC 2 Type I attestation targeted Q4 2026, Type II 12 months later. We will publish the audit firm name on the Trust Center when the engagement is signed, and the bridge letter once the report is issued.
2027, Headcount
First full-time hires in Security Engineering and Customer Success, contingent on reaching the revenue thresholds that justify them. We would rather stay small and credible than grow a fake org chart for procurement appearances.
We will not invent VP titles, fabricate team photos, or stand up a fake advisory board. The single-point-of-failure risk that comes with a solo founder is real, we mitigate it through documented infrastructure partners, escrowed credentials with our counsel, and the public compliance roadmap on the Trust Center.
Values
Verify, don't trust
Hardware attestation replaces vendor promises. You generate the Intel TDX quote and the NVIDIA GPU report yourself and verify them offline.
Confidential by default
Every GPU workload runs inside an Intel TDX trust domain. For inference, our relay forwards your request over TLS to a model running in a TEE at our provider.
Built for regulated industries
Law firms, clinics, accountants, fintech, we serve teams that handle privileged data.
No data exhaust
No prompt or output retention on the inference API, the chat or the agents, no training, and no sharing beyond the sub-processors that run the service.
Start building with confidential AI
$3.80 welcome credit with your first top-up of $10 or more. Per-second billing, no subscription.