Snowflake’s Pavan Pothukuchi explains why enterprise AI agents still struggle to act on systems of record, and how governed data platforms, semantic layers, and scoped agent identity close the gap. For enterprise leaders, agent autonomy has become a data architecture decision.

On Connected, Diana Blass tours Applied Materials' $5 billion EPIC Center, where chipmakers, equipment developers, and researchers will work in parallel to move AI chip breakthroughs from lab to market faster. Earlier collaboration could cut years from development cycles as AI demand outpaces manufacturing capacity.

At OCP Global Summit 2026, MiTAC Computing's Raymond Huang explains why AI infrastructure has become a full-stack integration problem spanning compute density, cooling, firmware, and deployment speed. For data center operators, advanced cooling and open management are now as decisive to scaling AI as GPU selection.

Marvell’s Sandeep Bharathi, Xi Wang, and Russ Esmacher explain why connectivity has become AI’s next rate limiter, from the copper wall inside the rack to scale-across links between AI clusters. For hyperscalers, the connectivity bets made today must hold up across multiple bandwidth generations.

On the Six Five Virtual Webcast, Kestra's Kai Waehner explains how AI agents acting across fragmented automation are moving governance out of individual tools and into a unified orchestration layer. Enterprises that define accountability, access control, and sovereignty requirements in that layer first can scale agents without losing end-to-end control.

On Episode 323 of The Six Five Pod, Patrick Moorhead and Daniel Newman unpack the biggest stories from the week, including Microsoft and NVIDIA’s hybrid intelligence push for Windows, Marvell’s Investor Day, new models from Anthropic, Mistral, and Reflection, the AI labs’ accountability hearings in New York and Australia, Google’s model-neutral Gemini agent, and whether temporary construction jobs justify AI data centers. Plus, Bulls & Bears covers Thursday’s chip-led pullback, NVIDIA’s run past $240, and Samsung’s record Q3 guidance.

At AI Infra Summit 2026, Synopsys' Thomas Andersen joins Brendan Burke and Matt Kimball to explain how chip design is moving from AI copilots that suggest to agentic systems that execute tool orchestration, timing fixes, and verification from a written spec. For enterprises, that shift compresses spec-to-tapeout timelines and lowers the resource barrier to custom silicon.

The companies racing to build frontier AI just agreed to referee themselves. Over the same week, launches, investor days, and earnings put a price on that arrangement. Microsoft is charging for finished work, NVIDIA is selling agent containment as infrastructure, and the suppliers AI can't route around, from Synopsys to Micron, are collecting the premium. Patrick Moorhead and Daniel Newman separate signal from stagecraft, then split hard on whether the White House accord actually binds these key players to anything.

‍

HPE's Varma Kunaparaju talks with Patrick Moorhead and Daniel Newman about why virtualization licensing changes are forcing enterprises to confront a bigger decision than which hypervisor to choose next: what operating model sits underneath their private cloud and AI factory. He outlines how CxOs can sequence modernization in phases, extend governance across open and frontier models, and decide where automation should replace human approval in IT operations.