Megawatts, but few permanent jobs? An agent-assisted evidence review of the data-centre construction cliff

Purpose Data-centre capacity is expanding rapidly, yet job claims vary by an order of magnitude. This paper synthesises quantitative evidence on facility-attributable construction and operational employment, separates direct jobs from wider multiplier-supported employment, and benchmarks data centres against energy and other built-asset comparators. Design/methodology/approach A documented protocol governed an agent-assisted evidence synthesis: two search streams, deterministic identifier checks, record-level source review and six-layer coding. Roughly 4,545 candidate mentions across about 591 records were harmonised to 44 rows from 20 sources. All 44 rows have completed source-verification records; no independent double coding is claimed. Estimates were compared within, not pooled across, evidence families. Findings Evidence shows a sharp phase contrast: roughly 0.7–2.0 peak construction workers/MW (projected 1.3–3.6) versus 0.15–0.35 operational FTE/MW. Different time bases preclude a unit-consistent ratio. Gross input-output totals describe a different object from realised permanent or causal local employment. Research limitations/implications The synthesis is structured rather than PRISMA-systematic; per-MW ranges rest primarily on one bottom-up model, and causal evidence is overwhelmingly US-based. Off-site prefabrication and upstream manufacturing layers cannot yet be populated, so the framework is a measurement agenda rather than an immediately deployable tool. Practical implications Practitioners, policymakers and delivery teams should value temporary trades work explicitly, target binding electrical, cooling and MEP capacity and require employment claims to state phase, denominator, boundary and evidence type before incentive appraisal. Originality/value The paper contributes a measurement discipline and a six-layer employment boundary that, applied through a construction-management lens, separates the temporal labour profile from regional development announcements.