AI Memory Products — Magic Quadrant

not gartner · Centripetal research · July 2026

X · COMPLETENESS OF VISIONHow much of the knowing-layer problem the architecture reaches: recall → provenance → procedure → governance → enforcement → self-repair → the meter beside the memory. High scores require reaching enforcement, not just retrieval.

Y · ABILITY TO EXECUTELive adoption and production readiness today: distribution, customers, shipping cadence, platform reach.

RECALL LAYER — vector/semantic ± temporal RUNTIME & PLATFORM-NATIVE STRUCTURED / SQL-NATIVE CENTRIPETAL OS — reference (governed knowing layer)

SCORES & RATIONALE — TABLE VIEW

PLAYER	CLASS	V / E	QUADRANT	WHY IT SITS THERE	SOURCE

Quadrant thresholds at 50/50: NICHE · VISIONARIES · CHALLENGERS · LEADERS. Survey population = the ten products in the 2026 "Top 10 AI Memory Products" landscape + Centripetal OS as the reference marker (not a categorical series — diamond + ring + label carry its identity beyond color). Ghost diamond = Centripetal with distribution solved. Categorical palette validated all-pairs on this surface (CVD ΔE ≥ 8, normal-vision ΔE ≥ 15, contrast ≥ 3:1). Recall is the commoditizing floor; the axis fight above it — refuse · meter · self-repair — is the open top of the category.