By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Uncapped with Jack AltmanFormer CAA Talent Agent Turned Investor with $70B in AUM on AI and Venture Strategy | Ep. 29
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
Coatue’s Thomas Laffont on AI cycle, venture strategy, and founders
- Laffont frames the current AI moment as a cycle shift driven by escalating competitive intensity and a move from cash-flow-funded buildouts to leveraged, existential spending—highlighting Oracle and AI labs as key signals.
- He explains Coatue’s “wide aperture” approach: investing across public and private markets from inception through IPO, guided by themes (semis, data centers, power, models, and the evolving data layer) rather than rigid stage buckets.
- On enterprise software, he argues the “system of record” is depreciating as data unlocks from SaaS silos into platforms like Snowflake/Databricks, while value migrates to agents and intelligence built atop broader datasets.
- He also covers venture dynamics—zero-sum competition, conflicts and disclosure, decision-making by internal momentum—and closes with personal lessons on mentorship, integrity, and craft (from golf to meticulous ‘gift-wrapping’ as a young professional).
IDEAS WORTH REMEMBERING
5 ideasAI capex is shifting from optional to existential.
Laffont contrasts earlier AI spend (funded by massive FCF at Meta/Google/Microsoft) with a newer phase where even cash-flow-negative players (e.g., AI labs) invest heavily, implying higher stakes and the need for greater investor vigilance.
Competition in cloud is intensifying beyond the old oligopoly.
He notes the market feels structurally different with Oracle pushing aggressively into cloud/AI and GPU-specialists like CoreWeave emerging, expanding the set of serious infrastructure contenders.
Infrastructure will capture durable AI value: chips, data centers, and power.
Laffont’s highest-conviction layer is foundational: semiconductors (NVIDIA and Broadcom highlighted), plus the physical constraints of data center buildout and electricity generation as AI demand scales.
Near-term “power trades” favor deployable generation, not science projects.
He’s constructive on nuclear (including behind-the-meter deals) and natural-gas generation (e.g., GE Vernova turbines), while treating fusion as promising but not yet investable with high conviction.
Models matter, but applications are fuzzier—data becomes the bridge.
He expects foundational models to consolidate around a handful of leaders, while admitting app-layer outcomes are less clear; he positions the data layer as the critical enabler for agent-driven applications.
WORDS WORTH SAVING
5 quotesWhat was different about the Oracle announcement… is now you're seeing kind of some leverage come into… it’s actually not just the free cash flow-positive companies that are investing.
— Thomas Laffont
I do think the system of record is dead.
— Thomas Laffont
Every interaction within the enterprise, within three years, will be recorded… the default is gonna be record on.
— Thomas Laffont
The thing I dislike the most about venture is… the zero-sum nature of the business at times.
— Thomas Laffont
A lesson I learned at CAA is… star quality is real.
— Thomas Laffont
High quality AI-generated summary created from speaker-labeled transcript.
