Corvex Achieves Confidential Computing Milestone on NVIDIA HGX B200 Systems

Event summary

  • Corvex completed verified production deployment of confidential computing on NVIDIA HGX B200 systems on March 3, 2026.
  • The deployment validates encrypted GPU-to-GPU communication using Intel Trust Domain Extensions and NVIDIA Confidential Computing.
  • Near-native performance enables secure AI without compromising production scale.
  • Corvex is among the first companies to achieve this production-ready confidential computing milestone.

The big picture

As AI systems transition from experimentation to mission-critical production, the demand for runtime security assurances is growing. This deployment establishes confidential computing as a production-ready infrastructure primitive, particularly valuable for highly regulated industries and protecting valuable IP in AI workloads. The combination of hardware-enforced isolation and cryptographic attestation addresses the 'data in use' risk, enabling verifiable security and compliance at runtime.

What we're watching

Adoption Pace
How quickly enterprises in regulated industries will adopt this confidential computing capability for AI workloads.
Competitive Response
Whether competitors will accelerate their own confidential computing solutions for AI infrastructure.
Performance Trade-offs
The extent to which near-native performance claims will hold as confidential computing scales across diverse AI workloads.