Are Burn Multiples BS in an AI World? & Sam Altman Needs $1TRN of Energy

Jason Lemkin and Rory O’Driscoll argue burn multiples are broken for AI companies and warn founders at $15M ARR to take any term sheet they can get.

• AI-native companies under $100M ARR have -126% FCF margins vs -56% for non-AI, but lower burn multiples because growth is so fast.
• Lemkin calls it “terrible advice” to tell a $15M ARR founder growing 100% to wait for a better valuation — take the deal now.
• 70% of VC dollars are flowing into fewer than 20 deals; companies outside that cohort face a binary have/have-nots funding market.
• Kingmaker deterrence is real: raising large rounds to crowd out competitors (Harvey in legal, Abridge in healthcare) actively works in 2025.
• FiveTran (~$400M ARR) buying DBT (~$100M ARR) is framed as necessary portfolio math — 6,700 unicorns exist but only ~20 IPOs happen per year, implying a 30-year backlog.
• a16z holding stakes in both FiveTran and DBT makes the merger structurally easier; shared lead investor removes the core ownership-dilution conflict.
• PE’s SaaS buyout model is threatened on two fronts: AI accelerates product obsolescence, and AI agents replace human seats, compressing the recurring revenue base.
• Sam Altman’s $1T capex figure is called out as requiring $1,000 billion of revenue to cover — Rory calls it potentially absurd on current trajectories.

2025-10-02 · Watch on YouTube