Current State of AI
summary
Claims that AI adoption drives exponential datacenter capex needing roughly 120 GW and that inference costs fall about an order of magnitude per year while model intelligence rises; cites supply bottlenecks with GPU lead times near 36 weeks and memory ASP-driven price jumps (DRAM/HBM shortages). Describes AMD stack: CPUs consolidating servers 8:1, GPUs with 400 GB HBM and 20 petaflops FP8 targets, DPUs giving 40X network acceleration, and silicon photonics cutting interconnect power 30–40%, improving tokens-per-watt and tokens-per-dollar.
AI-written summary · about
chapters (5)
- 0:00Introduction to AI talk
- 1:53AI industry growth since 2022
- 7:30Current AI infrastructure constraints
- 14:13AMD's value proposition in AI
- 20:53Summary and Q&A
related talks
29mUnlocking the Future of AI: Trends, Infrastructure, and Tokenomics
28mThe AI Tipping Point
20mOperating Frontier Intelligence at Scale
16mAI: Common Misperceptions and Uncommon Perceptions
06mThe AI Compute Crisis & Web3 Opportunity
06mAI Inference at the Edge: Beyond the Hype
24mManipulating Data in an Artificial Intelligence World
04m4 real-world AI deployments at enterprise scale