Are talent-and-IP-licensing deals a better structure than full acquisitions for fast-moving AI targets?

Harry Stebbings · published Sep 28, 2026

Meta's acquisition of Scale to bring in Alex Wang was an extremely effective deal — essentially buying a second CEO for 0.5% of market cap — and Wang has crushed it

Harry was initially sceptical, but Wang now runs Muse and the price paid was trivial relative to Meta's market cap

63:06 Meta scale deal was effectively buying a ceo cheaply not just licensing tech

Tom Hulme · published May 8, 2024

The Inflection deal was misreported: Microsoft was not buying the Pi model but a cluster of 12,000 H100s because it is GPU constrained, which was an incredibly smart move by Nadella

The Pi model was already obsolete months ago and nobody is queuing to use it, so the compute cluster was the real asset

47:28 Inflection deal was really a compute cluster acquisition not a model acquisition

Miles Dieffenbach · published Aug 4, 2025

Big acquirers are doing talent-and-IP-licensing deals instead of full acquisitions because a twelve-month regulatory review risks the target being obsolete and breakup fees are escalating

The AI landscape changes so fast that hot companies cool off within months (Windsurf, Stability AI), and the Wiz deal's 10% breakup fee sets a precedent where a $30B deal could carry a $4.5B risk over twelve months

67:13 Licensing deals avoid regulatory and obsolescence risk of full acquisitions

Harry Stebbings · published Aug 4, 2025

The talent-and-license maneuver is the smartest way around acquisition risk, but only works when the people and technology are the asset — not when revenue and customers are

In Wiz's case the revenue and customers are the asset; without them the company isn't worth $31B

68:49 Licensing works only when people and tech not revenue are the asset