Databricks at $100BN, CoreWeave’s $11B Debt Bet & Nubank’s $2.5B Profit Shocker - Ep.19

Rory O’Driscoll and Jason Lemkin on 20VC debate whether Databricks at $100B is cheap, CoreWeave’s $11B debt bet, and whether AI infra spend can actually reach a trillion dollars.

• Databricks crossed ~$4B ARR growing 50% vs. Snowflake at ~$4B growing 26%, making Databricks look undervalued at 25x revenue by comparison.
• a16z’s Databricks position, held since ~2013-2014, could return $30B+ if the company IPOs at $200B as implied by current valuation trajectory.
• Amazon, Alphabet, Microsoft, and Meta combined are already spending $365B on infra annually, straining balance sheets — Facebook is now borrowing money.
• Rory argues Sam Altman’s ‘trillions in AI infra’ is a metaphor for ‘a lot,’ not a real capex plan; no company has ever spent $1T in capex.
• Jason’s team runs 10 AI agents replacing 5 humans, spending $500K/year on AI tools — predicts B2B AI consolidation coming faster than the SaaS wave.
• Chimath’s SPAC return correlates with the 2021 bubble peak; Rory calls SPACs a structurally bad vehicle where promoter incentives misalign with investor returns.
• Nubank posted $2.5B profit, cited as strongest fintech globally; CoreWeave’s $11B debt load is a concentrated bet that AI infra demand holds.
• Key risk framing: AI adoption slowdown would trigger rapid repricing across the entire venture portfolio — platform risks are being systematically ignored in 2025.

2025-08-21 · Watch on YouTube