In-Person Event

AI Infra Summit 2026

Meet WEKA in Booth 710 at the Santa Clara Convention Center
15–17 September 2026

Your GPUs got faster this year. Your cost per token didn’t budge. That gap is the misdiagnosis at the center of inference economics: the bottleneck isn’t compute—it’s memory and data movement, and agentic workloads are widening it with longer contexts, persistent state, and thousands of steps recomputed rather than remembered, each one paid for in full a second time. WEKA CTO Shimon Ben-David unpacks where the diagnosis went wrong and what the fix looks like: context and memory architectures built for inference, the three metrics that reveal the true cost per token, and the architectural decisions that drive them. You’ll leave this session with a diagnosis you can run on your own stack, and the numbers to prove it.

Shimon Ben David

Shimon Ben-David

Chief Technology Officer, WEKA

When: Tuesday, September 15, 3:50 pm
Where: Data & Models Track
Type: Solo Presentation, 20 minutes

Join Us in Santa Clara

Not registered? Use code “WEKAIOPLATINUM20” to take 20% off an all-access pass.