Confidential Supercomputing: Essential Evolution or Elegant Illusion?

Sensitive workloads — spanning health and medical sciences, financial modelling, and AI training on proprietary data — are increasingly demanding confidentiality guarantees that traditionally out of scope for supercomputing Digital Research Infrastructure (DRI) . Confidential computing, long established in cloud and edge contexts through concepts such as hardware-enforced trusted execution environments, is now being proposed as a pathway to securing computation at supercomputing scale. But is this vision achievable, or does it fundamentally conflict with the architectural principles that make supercomputers fast?
This panel brings together experts from diverse background to debate whether confidential supercomputing is a necessary and achievable evolution of HPC infrastructure, or an aspirational posture that satisfies compliance optics without delivering meaningful security at scale. Panellists will examine the current state of practice and state of art and critically assess whether today's threat models justify the engineering trade-offs involved. Attendees will leave with an understanding of where confidential computing can genuinely deliver at AI and HPC scale — and where the gaps remain wide open.