Per-query compute costs now ration the consumer AI features that run on them.
Because frontier inference is leased at substantial marginal cost per query, consumer platforms must meter usage and compete on useful intelligence per watt rather than raw capability.
The same conclusion keeps arriving from across the workspace's research — 1 topics independently instantiate this theme. Filter the evidence by where it came from:
Per-token metering has become costly enough that owned hardware is being pitched to corporate buyers as the escape — the meter now shapes purchasing decisions.
The hyperscaler metric of success is now useful intelligence per unit of capacity and watt — per-query economics reaching the platform layer itself.
Microsoft's Useful Yield metric institutionalizes exactly the shift from raw capability to useful intelligence per watt that metered inference economics demands.
Apple rations its marquee consumer AI features because leased Blackwell inference carries substantial marginal cost per query — per-query compute, not capability, now sets the usage policy.