Observed arrival · 2026-11-13
Nightshift’s time card for spare-GPU inference
Credibility concern recorded. The source reference remains available for verification and correction.
Nightshift is building AI inference on GPU capacity owners make available when it would otherwise sit idle.
- For
- Developers exploring open-model inference
- Worth noticing
- Its utilization graphic is labeled a simulation; the cited 52% figure is limited to one Microsoft training cluster.
Field notes
The sample EX-0417 card sorts 40 product descriptions and lists token counts, a 0.840-credit cap, a 0.547-credit charge, and a 0.293-credit release. The page says no job ran and nothing was charged. The pilot described is narrower than the spare-GPU premise: one warm, pinned model in one region. Its utilization graphic is labeled simulated, and the page limits its 52% figure to a two-month trace of one Microsoft training cluster.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue