Therapeutic stagnation in Alzheimer’s disease (AD) highlights the role of non-genetic mechanisms like metabolic dysregulation. We propose a paradigm: exercise reprograms phosphoglycerate dehydrogenase (PHGDH) from a pathogenic nuclear transcription factor to a protective cytoplasmic metabolic enzyme through brain-periphery crosstalk. Under AD-associated stress, PHGDH translocate to the nucleus, where its helix-turn-helix (HHTH) domain activates pro-inflammatory genes (e.g., IKKα, HMGB1), suppresses autophagy, and accelerates Aβ deposition. Centrally, exercise-induced myokine irisin suppresses PHGDH nuclear translocation via AMPK/PGC-1α signaling, preserving its metabolic role. Peripherally, exercise inhibits PHGDH in hepatic Kupffer cells, reducing systemic IL-1β release and neuroinflammation. Notably, the HHTH-targeting inhibitor NCT-503 reduces Aβ plaques by 44–62% in APP-KI mice, validating this strategy. This framework positions exercise as a precision intervention, enabling novel diagnostics (e.g., PET probes for PHGDH compartmentalization) and synergistic “exercise-pharmacotherapy.”