VigFi · Field
AI & compute
GPU supply, model economics, HBM, and the next compute squeeze.
VigFi Consensus
A frontier lab ships a model under $1.00 / million output tokens before Q4 2026
Token economics are collapsing as open-weight competition and serving efficiency compound. The question is whether a frontier-class tier crosses the sub-$1 line this year — and which lab prints it first.
“Margin pressure plus open-weight competition makes a sub-$1 frontier tier the rational move.”
“Inference efficiency is loosening cost, but allocation politics and HBM still gate the floor.”
“From the compute-demand side we watch, the serving-cost curve keeps bending down.”
Each voice is weighted by its VigFi Score. Cross-field voices are included where their record bears on the question. Illustrative consensus over sample receipts; the ledger anchor is a deterministic demo, not a live chain.
Field leaderboard
AI & compute, by the record
Open calls on the record
Live, dated, and anchored.
A frontier lab ships a model under $1.00 / million output tokens before October 2026
HBM remains the binding constraint on advanced-GPU output through 2026
H100-class spot rental rates fall below $2.00/hr before Q4 2026
Companies on the record
They publish, dated and sourced — not ranked.
Companies put structured data, reports, and reviews on the record in their field. They publish but do not forecast, so they carry no VigFi Score.
Illustrative sample data while the product is in preview. The VigFi Score and Consensus are computed from the receipts; the ledger anchor is a deterministic demo, not a live chain.
The whole record, one number per voice.
See AI & compute ranked against every other field on the Reputation Index, or read the live field feed — every post dated, sourced, and anchored.