20VC: Why Late-Stage AI Is the Only Game That Matters
How VCs Are Capturing All the AI Value Before IPOs
The 20VC team (Harry Stebbings, Jason Calacanis, and Rory) dissects the AI investment landscape: Lightspeed’s $9 billion mega-raise, Oracle’s 45% stock plunge, Disney’s billion-dollar OpenAI deal, and what it all means for venture capital. Their thesis: late-stage AI is the game that matters, and everyone else is losing.
On the multi-stage playbook: “Lightspeed put a billion in two rounds of Anthropic. They had Rubrik last year and Nano this year. They did the two things a multi-stage manager has to do: great early stage companies built over 10-12 years, and they stuffed a ton of money in the hot late stage deals. Tick tick, insert $9 billion.” The formula: seed wins that mature over a decade plus concentrated late-stage bets.
On the greatest gift to VCs: “All these leaders not IPOing is the greatest gift to venture in our lifetimes. The fact that you can flood these top 20 companies with venture capital, keep this for yourselves… the public markets aren’t getting this.” Tesla IPO’d at $1.7 billion. SpaceX will IPO at $1.7 trillion. All that value stayed in private hands.
On seed economics being destroyed: “What Lightspeed can pay at seed is completely irrelevant. They don’t care if it’s 30, 40, 50. It’s an entry ticket for them to see the A, the B, the C.” When you have $9 billion to deploy, seed pricing doesn’t matter—it’s marketing for later rounds.
On Oracle’s 45% plunge: “They signed all this revenue and everyone gets excited about the RPO. And now everyone’s like ‘oh my god, to meet this revenue obligation you’re going to have to incur a whole lot of expenses.’ They opted to enter a very capital-intensive data center business to service one or two customers—principally OpenAI.” The market is stress-testing AI infrastructure bets.
On the canary question: “Is this a micro market hiccup or a canary in the coal mine? The markets are always interrogating you for truth. Google, keep going, you got a plan. Microsoft, like to see more of a plan. OpenAI, we’re good for it. Coreweave and Oracle, it’s not clear why you’re doing it.” The market is filtering from AI hype to specific execution.
6 Insights From 20VC on AI Investment Strategy
- Late-stage AI is the game - Playing the growth super cycle bet is the winning play; if you’re not in the big game, you don’t matter
- Private markets captured all the value - Tesla IPO’d at $1.7B, SpaceX will IPO at $1.7T; 15 years of compounding stayed in VC hands
- Seed economics are marketing - Multi-stage firms don’t care what they pay at seed; it’s an entry ticket for later rounds where the real money deploys
- Oracle/Coreweave are the high-octane bets - No IP, no diversification; they’re the marginal providers that amplify whatever the AI trend does
- Growth slowdown is the existential risk - “Nothing is as terrifying as a high growth bet that slows down” - you go from being valued on growth to valued on cash flow
- Disney deal is an IP template - First content deals set the template; everyone else pays more later, and rates ratchet up in 3 years
What This Means for AI Investors and Builders
The 20VC conversation crystallizes the venture capital logic of AI: late-stage growth is the winning bet because leaders staying private means all the compounding value accrues to VCs, not public markets. But the Oracle plunge shows the market is now stress-testing who has real plans versus who’s just riding the wave. For organizations thinking about AI investment or partnerships, the filter is clear: the infrastructure plays without IP or diversification are the most volatile; the platform players with existing ecosystems are the safer bets.