20VC: OpenAI Cuts Valuation to $1.25TRN | Public Market Meltdown: GOOG Loses $500BN in 12 Hours | DoorDash Buys Instacart for $17BN | Kalshi Raises $1BN and Cognition Eyes $80BN Valuation

What It Covers

Harry Stebbings, Rory O'Driscoll, and Jason Lemkin analyze OpenAI's valuation cut from $1.4T to $1.25T, Alphabet losing $500B in market cap, DoorDash acquiring Instacart for $17B, Kalshi raising $1B at $40B, and venture rounds for Cognition at $80B and Physical Intelligence at $20B.

Key Questions Answered

• •AI Valuation Reality Check: OpenAI's valuation cut from $1.4T to $1.25T with only $11B committed reflects rational price discovery, not AI collapse. When Anthropic is liquid at $1.5-2T, investors logically pay less for a slower-growing, still-unprofitable, illiquid OpenAI. Two securities competing for the same capital always produces this outcome. The narrative of AI cracking is misreading a straightforward comparative valuation exercise.
• •Anthropic Revenue Deceleration Signal: Anthropic's GAAP revenue grew from $5B in Q1 to $11B in Q2 to $17B in Q3, meaning quarter-on-quarter growth dropped from 100% to roughly 50%. The most actionable read is pricing pressure: OpenAI offers comparable products at 30% discounts, and Haiku 5.5 plus Jeb delivered 72% cost reductions on identical B2B workflows. Token deflation is now measurable in production environments, not just benchmarks.
• •Agentic Disruption Threat to Delivery Platforms: DoorDash disclosed only several hundred AI agent-placed orders against 3 million daily deliveries, but the structural risk is real. Agents can identify that a local restaurant delivers directly for $2 versus $14 through DoorDash, eliminating platform fees entirely. The third competitor to DoorDash is effectively nobody — direct restaurant delivery — and agents make that option frictionless and discoverable at scale.
• •Revenue Multiples Are Irrelevant in AI Markets: Physical Intelligence, with minimal revenue, is valued at $20B while Dropbox, generating billions in revenue, commands a fraction of that. The market prices future trajectory, not current output. Founders and investors should stop anchoring to revenue multiples when evaluating AI infrastructure and robotics companies. The only variable that matters is whether demand for the underlying capability will be materially larger in three to five years.
• •Venture Portfolio Construction Tradeoff — Velocity vs. Multiple: When choosing between Base Ten at $30B (inference), Cognition at $80B (coding agents), and Serval at $3B (ITSM), the multiple opportunity is clearest at Serval — ServiceNow trades at $110B, giving a concrete comparable 30x larger. Inference offers revenue velocity but faces commoditization from five competitors. Cognition offers enterprise coding momentum but requires a trillion-dollar outcome to generate a 10x return from current entry.
• •M&A Threshold for Enterprise Acquirers: Large acquirers including Salesforce consistently require $100M+ ARR before deploying their GTM resources against an acquisition. Below that threshold, deals underperform because the acquirer cannot make the target material within a $50B+ business. Founders targeting acquisition exits should treat $100M ARR as the functional floor for serious strategic interest, not just a vanity milestone. Salesforce's Cresta acquisition at 10x ARR versus Listen Labs at 100x illustrates the pricing gap between scale and early-stage targets.