Bitcoin Mining Hardware Has Consolidated Into an Oligopoly. That Is Now a Grid Issue.
Bitcoin was first mined in 2009 on a standard laptop CPU. Within a few years, miners had migrated to graphics processing units for their superior hashing efficiency, then to field-programmable gate arrays, and by 2013 to application-specific integrated circuits purpose-built for Bitcoin mining. Each hardware generation rendered the previous one unprofitable, because the proof-of-work mechanism adjusts difficulty to maintain a ten-minute block interval regardless of total network computing power. Better hardware does not reduce the network’s energy consumption; it raises the difficulty floor, and the total energy expenditure grows with investment.
The economics of ASIC hardware have produced a highly consolidated industry. Three manufacturers — Bitmain, MicroBT, and Canaan, all founded in China — account for an estimated combined market share exceeding 85%. Their machines define the efficiency frontier at any given moment. In 2025, the Antminer S19 XP Hyd operated at 255 terahashes per second on 5,304 watts, while the MicroBT Whatsminer M50S ran at 126 terahashes per second on 3,276 watts. Both figures represent substantial efficiency improvements over prior generations, but the network’s total power demand has continued to rise because the efficiency gains are immediately absorbed by higher difficulty levels and more deployed machines.
The mining pool structure has followed the hardware toward consolidation. Individual miners working alone cannot compete effectively with pooled operations that spread equipment costs, electricity contracts, and computational variance across large facility deployments. Large mining pools operating ASIC-filled data centers now dominate the Bitcoin network’s hashrate, with the largest facilities drawing hundreds of megawatts from regional grids.
All three major hardware manufacturers started U.S. production operations in 2024 in response to tariff pressure on Chinese imports. That localization does not resolve the supply chain concentration; it relocates some of the manufacturing while leaving the design, firmware, and intellectual property in the same hands. For a load that now measures in gigawatts across American grids, the hardware dependency on a small number of Chinese-founded firms is a supply chain risk that the electricity sector has not fully priced.