Version 3

H100 GPU spot prices remain roughly 40% below their May 2026 peak at approximately $2.42/hr [1], while one-year H100 contract prices on neoclouds have climbed from a trough of ~$1.70/hr to ~$2.65/hr over the same period [3]. SemiAnalysis argues this divergence reflects market maturation — buyers shifting from exploratory spot capacity to committed production contracts — not softening AI demand [5]. Broader market observers continue to read falling spot prices as a demand weakness signal [2]. A separate report with the title suggesting H100 prices are up 20% year-to-date cites Nebius Group as a potential beneficiary of tight AI cloud supply [8].

Spot price indexes are the most visible public data on compute health, and their decline has sustained a bearish demand narrative. If contract prices — which reflect committed production workloads — are moving in the opposite direction, observers relying on spot data alone may be drawing inverted conclusions about where sustained AI investment is headed.

H100 GPU spot prices have declined to approximately $2.42 per hour as of late June 2026, sitting roughly 40% below their May peak [1]. This sustained drop has generated concern among AI observers and investors that appetite for compute is cooling — a reading amplified by multiple public spot price indexes and by market commentators framing the decline as evidence of stress in the AI compute boom [2].

SemiAnalysis, which publishes a proprietary neocloud survey, posted a five-tweet thread on June 26 arguing that spot price indexes are the wrong instrument for measuring AI demand health [1][3]. The firm draws a structural distinction between spot and on-demand markets — which serve proof-of-concept runs, one-off evaluations, burst workloads, and overflow capacity — and term contract markets, where buyers commit to planned, recurring, revenue-bearing production inference or training workloads [4]. One-year H100 contract prices on neoclouds have climbed from a trough of roughly $1.70/hr in late 2025 to approximately $2.65/hr as of June 2026 [3]. SemiAnalysis attributes this to serious buyers locking in term capacity for production workloads, pushing contract prices higher even as the spot market remains loose [5].

The firm's conclusion inverts the prevailing read: falling spot prices alongside rising contract prices signal market maturation — buyers shifting from exploratory, flexible capacity toward committed production deployment — not a cooling of AI ambition [5][6]. A LinkedIn post by Cristian Sion makes the anti-bubble case directly, arguing that rising prices for older H100 hardware contradict bubble-deflation claims [7]. Separately, reporting that H100 GPU prices are up roughly 20% year-to-date [8] and ClusterMAX 2.0, SemiAnalysis's GPU cloud rating system [9], add further context to the contract-price-rising narrative.

A complicating data point is the SiliconData H100 Hyperscaler Index for April 2026, which showed pricing in flat mode [10] — a different signal from both the spot price decline and the neocloud contract price rise. Whether hyperscaler contract pricing has since diverged from the neocloud trajectory remains unresolved.