Levie now uses Salesforce 5x more than at any point before. The Box CEO @levie connected Salesforce's MCP server to Claude Code. Now he runs customer and market intelligence queries he would never have bothered pulling up by hand. The agent removes the friction. The underlying system gets queried more, not replaced. If you hold $CRM, the agentic era is an engagement tailwind, not a disruption risk - gated on whether the data platform handles the query load. Why incumbents gain from agents: Source: CXOTalk -

If you are sizing the SpaceX pre-IPO trade, @AravSrinivas gave the clearest 10-year bull case I have heard. On @twentyminutevc: SpaceX is N-of-one among the upcoming IPO class. Anthropic and OpenAI can claim to do what each other does. SpaceX builds physical planetary-scale connectivity. No one else is close. Starlink creates irreversible behavior change. Once you have it, you do not go back. Srinivas's framework: physical infrastructure moats outlast AI model moats by a decade or more. When models commoditize, Starlink's network lock-in does not. The comparison class for SpaceX is not Anthropic. It is global telco infrastructure - but built 20 years later and without legacy costs. Full breakdown: Source: 20VC with Harry Stebbings -

If you're paying $2T for SpaceX at IPO, this clause is in the S-1. Anthropic pays $1.25B/month to rent SpaceX's Colossus data center. That's $15B/year - the core of the xAI business case. The contract has a 90-day cancellation clause. Either side can walk. @rodriscoll on @20vcFund: one customer and one 90-day window stand between you and losing a third of the valuation argument. Any Anthropic statement on compute strategy - shifting toward AWS, Google, or in-house capacity - is a material event that moves in 90 days. The clause that wasn't discussed when the deal was announced: Source: 20VC with Harry Stebbings -

"Compute through 2027 is sold out." OpenAI's CFO said it on air, with no hedge. Where she's buying instead: @thefriley confirmed on @theallinpod with @chamath & others: no additional compute in 2026, 2027 "pretty limited as well." The Elon Musk anomaly - too much compute - is gone. Where she feels most short: 2030, 2031, 2032. The market prices infrastructure on a 2-year horizon. Friar buys on 6. Oracle, CoreWeave, and Nvidia are the proxies - but the real constraint is power and land. Source: All-In Podcast -

Microsoft canceled internal cloud code licenses. The token bill became untenable. Per Matt Turck (@mattturck) on The MAD Podcast. This is now the number one or two concern for enterprise AI leads, per Aaron Levie's (@levie) conversations with 200 CIOs this year. The mechanism: AI tools that launched on flat subscriptions - Cursor, internal Microsoft code tools - repriced to per-token billing. Enterprises built budgets for the old model. The new model hit them mid-deployment. This pressure compounds as models get more capable. It does not resolve. What the token cost shock means for enterprise SaaS positions: Source: The MAD Podcast with Matt Turck -

If SpaceX acquires Tesla, Elon ends up with 60-80% voting control of the combined entity. Dave Blunden walked through the math. SpaceX already has 10-for-1 super-voting shares, giving Elon ~85% of votes there. Tesla is 1-for-1, where Elon holds ~20%. If SpaceX acquires Tesla and absorbs its one-vote shares alongside SpaceX's existing super-voting structure, the voting math tilts hard in Elon's favor. Prediction markets (Kalishi): 50-50 odds within a year. @PeterDiamandis: 100%. Two blockers: Tesla's public shareholder vote (Elon doesn't control it), and SpaceX's IPO valuation needs to compare favorably to Tesla's market cap at deal time. If Tesla stock declines relative to SpaceX post-IPO, Tesla shareholders may welcome the deal. Combined entity: EVs, energy, solar, rockets, satellites, internet, humanoid robots, Mars. First $10T company thesis. Full voting structure breakdown and what changes if SpaceX IPOs first: Source: Moonshots with Peter H. Diamandis -