06-reference

stratechery meta compute

Tue Jan 13 2026 19:00:00 GMT-0500 (Eastern Standard Time) ·reference ·source: Stratechery ·by Ben Thompson
metaai-infrastructureopenaiadvertisingreality-labs

“Meta Compute, The Meta-OpenAI Battle, The Reality Labs Sacrifice” — Ben Thompson

Why this is in the vault

Meta’s infrastructure bet and the Meta-vs-OpenAI competitive framing are important context for understanding who wins in AI and how advertising funds AI buildout.

The core argument

Three topics. First, Zuckerberg announced Meta Compute — a top-level initiative to build tens of gigawatts of data center capacity. Thompson dismisses the “Meta is becoming a cloud provider” narrative; Meta’s play is that the limiting factor in AI is infrastructure, and Meta’s ad-funded free cash flow can outspend OpenAI.

Second, Thompson reframes the AI competitive landscape: OpenAI’s real threat is to Meta (competing for user time, AI-generated content displacing social feeds), not Google. Meta’s offense is infrastructure dominance — outbuilding OpenAI’s compute while already having a proven monetization model. OpenAI is “starved for compute” and lacks a scalable ad model.

Third, Meta is cutting ~1,000 Reality Labs jobs (10%), pivoting from VR/Metaverse toward AI wearables and mobile. This isn’t an admission of failure but a strategic trade: trimming Reality Labs spending to give investors confidence in the Meta Compute buildout. Reality Labs’ poor ROI had eroded investor trust, making it harder to get a blank check for AI infrastructure.

Mapping against Ray Data Co

The Meta ads + AI compute flywheel is relevant context for RDCO’s automated Meta ads work (autoinv). Understanding Meta’s AI infrastructure priorities helps predict what ad tools and capabilities will emerge.