The　growth　of　the　manufacturing　industry　is　the　engine　of　rapid　economic　growth　in　developing　regions.　Characterizing　the　geographical　distribution　of　manufacturing　firms　is　critically　important　for　scientists　and　policymakers.　However,　data　on　the　manufacturing　industry　used　in　previous　studies　either　have　a　low　spatial　resolution　(or　fuzzy　classification)　or　high-resolution　information　is　lacking.　Here,　we　propose　a　map　point-of-interest　classification　method　based　on　machine　learning　technology　and　build　a　dataset　of　the　distribution　of　Chinese　manufacturing　firms　called　the　Gridded　Manufacturing　Dataset.　This　dataset　includes　the　number　and　type　of　manufacturing　firms　at　a　0.01　degrees　latitude　by　0.01　degrees　longitude　scale.　It　includes　all　manufacturing　firms　(classified　into　seven　categories)　in　China　in　2015　(4.56　million)　and　2019　(6.19　million).　This　dataset　can　be　used　to　characterize　temporal　and　spatial　patterns　in　the　distribution　of　manufacturing　firms　as　well　as　reveal　the　mechanisms　underlying　the　development　of　the　manufacturing　industry　and　changes　in　regional　economic　policies.