Explore Vinci’s new Research page for technical papers, category frameworks, and validation resources on continuous physics reasoning.
Schedule a Demo
Back to all News

Podcasts + Vodcasts

Sarah Osentoski on Inference Demand and the Physical Constraints Shaping AI Infrastructure

4.9.2026 | By Vinci

In this TechTV panel discussion, Sarah Osentoski joins Bill Mew and other industry experts to discuss how the AI infrastructure landscape is evolving as inference demand grows. The conversation examines the technical and operational constraints shaping the next phase of AI systems, and why hardware design is becoming more complex as physical limits become harder to ignore.

Sarah’s contribution focuses on the engineering side of that transition: as systems scale, challenges such as overheating, warpage, and power density become more consequential. She explains why Vinci is building a foundation model for physics to help hardware teams evaluate those issues more effectively, and why secure, behind-firewall deployment matters as infrastructure requirements become more demanding.

Watch the Panel

What this Conversation Covers

  • How growing inference demand is changing AI infrastructure requirements
  • Why the next phase of AI systems will be shaped by technical and operational constraints, not just model performance
  • Why physical issues such as heat, warpage, and power density are becoming more important in hardware design
  • Why traditional workflows make these engineering problems difficult to evaluate quickly and consistently
  • How foundation models for physics can help hardware teams reason about physical behavior earlier and more continuously
  • Why secure, on-prem deployment matters in infrastructure and enterprise environments
  • How physical constraints influence performance, efficiency, and system design decisions

Why this Matters

As AI systems move from training-heavy buildouts toward broader inference deployment, infrastructure decisions are increasingly shaped by physical reality: power, cooling, packaging, reliability, and data movement.

This is where Sarah’s framing matters. Hardware teams are not just dealing with more demand. They are dealing with systems that are harder to design, harder to cool, and harder to validate using traditional workflows. If physical behavior remains accessible only through slow, specialist-driven analysis, it becomes a bottleneck on the broader system.

This conversation points to a broader shift: AI infrastructure is increasingly constrained by physics, and addressing that requires deterministic, solver-grounded systems that can support real engineering workflows.

Sarah Osentoski joins a TechTV panel discussion on the shift from training to inference and the physical constraints shaping AI infrastructure, including heat, warpage, power, and deployment requirements. The video runs 35:48.

Watch Now

A Different Kind of AI for Hardware Design

Vinci is a frontier lab building the foundation model for the physical world, with deterministic, solver-grounded systems already deployed in production engineering workflows on flagship programs.

Explore the Platform

Media inquires: vinci@bigvalley.co

Related News