INFRASTRUCTURE◐ DEVELOPINGM · Material▽ Strained
35–50% of planned AI data centers slip—compute timelines crack
Sep 3, 2026SOURCE: smithschool.ox.ac.uk
SO WHAT
In the thousand-day window, schedule risk delays real compute, pushing the phase transition into rationing-by-timing rather than choice. Watch network power and capital capture during the stall.
A new Oxford Smith/Marex white paper estimates 35–50% of planned AI data centers are behind schedule, citing a delivery gap. The old system’s build-out cadence is colliding with the inversion-driven demand curve.
This is Negative Resistance’s reframed reading of a reported signal. The headline and analysis above are our interpretation through the thousand-day-window lens. The original reporting lives at the source linked above.