From Industry to Services? China’s Big Data Pilot Zones and County-Level Structural Transformation, with Implications for Energy Intensity

China’s National Big Data Comprehensive Pilot Zones (BDPZs) constitute one of the world’s largest place-based digital economy experiments. This paper provides the first county-level evaluation of the program, asking whether the designation shifted local economies from secondary toward tertiary activity—a structural margin relevant to, but distinct from, energy transition. Using panel data on 1592 counties (2011–2023) and a matched staggered difference-in-differences design, we combine exact two-way fixed-effects estimation with the heterogeneity-robust estimators of Callaway and Sant’Anna (2021) and Sun and Abraham (2021), formal joint pre-trend tests, randomization inference, relative-magnitude sensitivity bounds, and spillover falsification. Our central result is a delayed shift in the nominal value-added structure toward services in the 2017 adoption cohort: the effect is statistically indistinguishable from zero for three years, then builds to +4.9, +6.7, and +10.6 percentage points in post-treatment years four through six (all p<0.001), is robust across never-treated, matched, distant, and growth-trajectory-matched control sets, and survives a conservative relative-magnitude sensitivity bound (breakdown M¯≈3.5) despite a formally rejected parallel-trends test. Three qualifications discipline the reading. First, the shift appears in value-added shares but not in (incompletely measured) employment shares, consistent with a price- and productivity-driven value-composition change rather than labor reallocation. Second, per capita income did not rise—it fell relative to poorer, faster-converging controls—so we make no growth claim. Third, the single-province 2016 (Guizhou) case is not separately identified: its sign flips across control sets, and it is statistically indistinguishable from province-level placebos (p=0.27), so we set it aside. Because sectoral recomposition bears on energy intensity (under constant sectoral intensities) but not on energy sources or absolute use—and because data-center construction is itself energy-intensive—we make no energy-transition claim; the net energy effect is sign-indeterminate with current data.