The Compound and Friends Why Bubble Talk is Totally Wrong with Ankur Crawford
236 snips
Apr 17, 2026 
Dr. Ankur Crawford, an EVP and portfolio manager at Alger known for semiconductor and tech expertise. She explores accelerating AI timelines and practical demos. She explains compute and token economics, DRAM and chip bottlenecks, and why a few winners dominate tech demand. She discusses hyperscalers, Amazon’s role, consolidation across the semiconductor stack, and risks to software and screen-based platforms.
AI Snips
Chapters
Transcript
Episode notes
Challenge Vague Bubble Claims With Specific Fundamentals
- Test skeptics' bubble claims: ask them to justify which part is a bubble—CapEx, product, or valuation—then check the fundamentals.
- Crawford uses supply, consolidation, and CapEx visibility to counter generic 'bubble' claims.
Hyperscalers Verticalize To Lower Compute Costs And Capture AI Revenue
- Hyperscalers' CapEx targets aim to own the compute stack (chips, data centers, cooling) to lower costs and monetize AI.
- Amazon's Tranium chips and Anthropic partnership are examples of verticalizing to cut compute costs and drive AWS top-line.
Token Pricing Will Differentiate By Use Case And Value
- Not all tokens are equal: pricing and value will vary by use case (drug discovery vs. casual queries).
- Enterprise apps will pay more for high-value tokens while trivial consumer queries cost far less.