Pat Gelsinger: Memory, packaging and power—not chip speed—are AI's real bottlenecks
a16z Podcast · rss · 2026-10-09
a16z's Raghu Raghuram and Guido Appenzeller sit down with former Intel CEO Pat Gelsinger on AI buildout constraints. Key points:
- AI could transform chip design, but faster design alone won't fix the industry's biggest problems
- Bottlenecks: manufacturing, memory bandwidth, advanced packaging, and power
- The explosion of specialized AI chips may consolidate around fewer architectures
- New memory technologies and the copper-to-optical networking shift
- Energy capacity may become the major constraint on AI growth
- What virtualization looks like when infrastructure is built for agents rather than humans
More from Infra
- Why Starlink spooks regulators: a 10,000-satellite mesh can route around licensed gateways — kingslyj · 2026-10-09
- Why GEO satellites can't do this: small coverage, users locked to one satellite — kingslyj · 2026-10-09
- How is Starlink a national security threat? A thread explains the gateway problem — kingslyj · 2026-10-09
- Hyperscaler signs 210MW 15-year lease, locking in roughly $5.2B in contracted revenue — YvesMulkers · 2026-10-09
- Analyst Doubles Down: NVL72 AI Servers Are the Biggest Product Cycle in History — firstadopter · 2026-10-09
- New EXLR8 quant claims to beat EXL3 on every metric, GLM 5.3 cold-start TTFT at 30s — EAccelerate_42 · 2026-10-09