Meta's late-August analyst changes introduced a valuation angle that barely existed a year ago. Evercore raised its target from $820 to $860 and highlighted the possibility that Meta could eventually monetise spare AI compute capacity. Other firms remained less aggressive, with Truist trimming its target and several brokers keeping Hold recommendations.
The idea is unusual because Meta is not a public cloud provider. Its data-centre buildout has historically supported advertising, recommendation systems and internal products. If excess capacity can be sold externally, part of that infrastructure could become a revenue-generating asset in its own right.
The optionality is real, but it should not be valued like AWS yet
Selling surplus capacity is conceptually straightforward. Building a reliable cloud business around it is not. External customers require service agreements, software tooling, support, billing and predictable access to capacity.
That means the opportunity deserves option value before it deserves a full cloud multiple. Analysts assigning billions of dollars to the idea are making a judgement about strategic flexibility, not valuing an established segment.
Meta's core AI return still comes through advertising
The most important economic use of AI at Meta remains improving engagement, targeting, content ranking and ad performance. Those gains can raise revenue without creating a separate AI line item.
That is why the compute thesis should be treated as incremental. If internal AI spending already supports better advertising economics, selling spare capacity would be a bonus rather than the primary justification for the infrastructure buildout.
The analyst spread reflects uncertainty over capital intensity
Meta can grow earnings while spending aggressively, but the market still needs evidence that each new wave of infrastructure produces adequate returns. The more cautious targets effectively apply a discount to that uncertainty.
The bullish case is becoming broader, though. It now includes advertising productivity, consumer AI products and potential compute monetisation. That makes Meta harder to model, but it also gives analysts more than one route to justify upside.