Expensive GPUs are sitting idle across the industry.

Utilization rates at many organizations don't crack 50%.

The hardware keeps getting faster, but the efficiency gains aren't automatic anymore.

Craig Tavares, President and COO of BUZZ HPC, joins leaders from Intel, Foundry, and Genesis Cloud to unpack the disconnect.

Craig brings over two decades of infrastructure experience and a full-stack view of AI compute economics, from driver-level optimization and multi-tenant workload scheduling to the buy-vs.-rent calculus that shapes every deployment decision.

This panel from the 2025 RAISE Summit in Paris explores what it will take to close the gap between raw compute capacity and real-world AI performance, and whether smarter software, better orchestration, and new pricing models can bend the cost curve before the next hardware generation arrives.