>>> Follow up of The Information intervie sent earlier...

AI INFRASTRUCTURE INVESTMENT

Vinod Khosla Interview ( {MSG /ID 68FE62A7000299D028410001 <GO>} )→ Portfolio Implications

Who is Vinod Khosla?
  • Co-founder Sun Microsystems
  • Founder Khosla Ventures
  • Early investor in OpenAI
  • Strong conviction on AI infrastructure economics

Key Insights
  • Current enterprise AI ROI is weak due to execution talent gaps, not demand
  • Compute demand is secular and driven by labor substitution at scale
  • Margins expand into 2030s via cheaper chips + better models (efficiency gains)
  • Nvidia maintains a strong moat; only credible disruption is photonics
  • Power grid capacity = primary bottleneck to AI scaling
  • Venture returns become highly concentrated (2-3 pct winners capture most value)

Trade Basket (Infrastructure > Apps)
  • NVDA (compute dominance)
  • MSFT (Copilot + GPT distribution)
  • AMZN (AWS infrastructure)
  • GOOGL (Search yield uplift)
  • AMD, AVGO (inference + networking exposure)
  • EQIX, DLR (data center scarcity → pricing power)
  • GEV, AES (flexible power capacity)

Catalysts (12–24 months)
  • New GPU cycles reduce $/inference → faster deployment
  • Enterprise AI moves from pilots to production scale
  • Data center leasing acceleration
  • First geothermal deployments for hyperscalers
  • Algorithmic efficiency breakthroughs → margins up

Risks and Mitigation
  • Adoption slower → tilt toward infrastructure over software
  • Power delays → add utilities + turbine exposure
  • Photonics arrival early → small hedge sleeve
  • High rates → partial Nasdaq hedge + energy rotation

Bottom Line
Demand is certain.
Power and execution determine timing.
Infrastructure is the most reliable winner of the compute decade.