The reported plan, covered by TechCrunch AI, underscores a critical tension in the current AI boom: the staggering capital expenditure on hardware may be outpacing the software's ability to generate standalone revenue. Meta's potential pivot to selling raw compute suggests its primary AI products—like Llama and Meta AI—may not yet be compelling enough revenue drivers to justify its $182.9 billion infrastructure commitment on their own. In our view, this represents a second-order bet that the real long-term value lies in owning the foundational compute layer, not necessarily in winning the model race.
However, it also exposes Meta to the same 'bubble' risks facing all hyperscalers: if end-user AI application revenue fails to materialize at scale, the demand for rented compute could collapse, leaving massive, depreciating assets.