The Great Migration of Bitcoin Machines from China

China had been the largest center of Bitcoin mining for years, concentrating a significant portion of the global hashrate. It offered hardware manufacturers, experienced operators, and various types of cheap electricity — from seasonal hydroelectric power in Sichuan to coal in the northern regions. Farms could move within the country according to the rainy season and available price.

Data from the Cambridge Centre for Alternative Finance shows that China’s share was falling even before the nationwide crackdown: from an estimated 75.5 percent in September 2019 to 46 percent in April 2021. So the story does not begin with a single ban overturning a fixed mining map overnight. Mining was already spreading to other countries.

Estimates are based on mining pool data and geolocation of connections. They don't see every machine and can be skewed by VPNs or the methodology of the pools involved. The numbers are the best available map, not an exact count of all ASICs in the world. Every conclusion about shares, therefore, needs this boundary.

The concentration was still significant. A political decision in one country could affect a large portion of the global performance simultaneously. Critics saw it as proof of centralization, miners as an operational risk, and Chinese authorities as an energy or financial activity they wanted to restrict.

Mining performance wasn't owned by a single Chinese company or controlled by a single command. However, it was exposed to the same legal environment, electrical grids, and supply chains. Geographical decentralization turned out to be something other than the number of wallets or nodes: it was about how much hardware could be affected by the same authority.

In May and June 2021, orders and inspections were followed by mining farms shutting down. Operators lost revenue every day their machines stood idle. Smaller businesses sold equipment into a market flooded with second-hand hardware, accepting steep discounts. Larger companies searched for sites with megawatt-scale electricity connections and permission to operate for the long term.

Thousands of miners had to be labeled, packed, and transported across borders. Reuters agency reports described limited air and sea capacity, customs procedures, and a lack of prepared facilities. A mining farm isn't a warehouse with sockets; it needs management, transformers, ventilation, security, and maintenance.

Kazakhstan attracted with its proximity and affordable energy, the United States with a more stable capital and contractual environment, and Russia with cold temperatures and electricity in some regions. Each destination had its own risk. A low energy price could be offset by an unstable network, permits, or the same political vulnerability that miners were fleeing from.

The move also changed ownership. Some machines were bought by foreign operators, others were relocated by Chinese companies themselves, and part of the outdated equipment was no longer worth reinstalling. The hashrate that later returned, therefore, wasn't always the same performance under the same control at a new address.

The media shortcut about a digital exodus obscures the most interesting part: network statistics consist of physical things and working decisions. Each exahash in the graph represents a device, contracts, service, and a person who bears the risk that another jurisdiction will change the rules before the investment returns.

Each operator's decision looked different. A well-capitalized firm could rent containers, negotiate a new power contract, and wait months without revenue. A smaller miner might have had only one realistic option: sell the ASIC to a buyer who could afford the relocation. The sell-off of machines was therefore not merely a sign of panic but a mechanism by which hashrate moved to people with access to different infrastructure. Equipment prices, electricity prices, and expected rewards were all being recalculated simultaneously. The network had no central evacuation plan; relocation emerged from thousands of unequal decisions, some of which saved a business and others closed its final chapter.

According to data published by Cambridge, estimated global hashrate fell to 57.47 EH/s on June 27, 2021. With less computing power online, miners were finding valid blocks more slowly on average than the target interval of 10 minutes. Bitcoin kept running, but users could wait longer for confirmations until the next difficulty adjustment.

The protocol compares the actual time with the target every 2,016 blocks and adjusts the difficulty. When a large portion of the performance disappeared, the difficulty decreased. Each remaining hash had a higher chance of finding a block, and operation outside of China became more profitable. The rule created an economic incentive that helped replace the performance.

The adjustment does not happen as soon as machines switch off. The network waits for the end of the current 2,016-block interval; the shutdown does not restart that count. With less computing power, the remaining blocks can take longer to arrive, extending the wait for lower difficulty. Resilience does not mean the network feels no impact. It means there are rules already in place to bring it back toward its target pace, without an emergency decision by a central administrator.

On December 21, 2021, Cambridge reported estimated global hashrate of 193.64 EH/s, almost back to its level before the ban. By February 2022, the estimate had reached a new high of 248.11 EH/s. This recovery did not come from a new factory opening on a single day. It came from idle machines gradually returning to service, purchases of new equipment and mining expanding across more locations.

Rapid recovery is often presented as proof that the ban had no effect. Data say something else. Months of outages, forced sales, and costly migration were real. The network, however, had adjustable costs and a global reward price that motivated people to fill the released space.

Later data from Cambridge surprisingly showed a significant share of performance again attributed to China. Part of the mining apparently continued secretly or returned after the first wave. This is an important correction to the simple story of a complete ban: a legal ban and zero activity are not the same thing.

Geographical data from pools have limits, and hidden operation is, by definition, difficult to measure. It's impossible to precisely determine each machine or claim that all reported connections corresponded to the physical location. A reasonable conclusion is that mining significantly shifted and, at the same time, didn't completely disappear from China.

The new distribution raised more questions. Kazakhstan later faced energy problems and restrictions, and some American regions debated noise and consumption. Mining from stranded energy could utilize unsold capacity, but each project had its own economic and ecological conditions. The shift itself didn't guarantee a better energy profile.

The network's resilience, therefore, didn't look like the indestructibility of one center. It looked like the ability to lose a center, adjust difficulty, and offer sufficient reward to people elsewhere. This mechanism worked because machines were movable and the Bitcoin reward had a global market. Both conditions have costs and boundaries.

The great migration changed Bitcoin's map, but mainly clarified what the map means. Hashrate isn't a cloud without a state. It's an industry tied to energy, law, and logistics. The network survived the Chinese intervention thanks to the protocol, but also thanks to months of work by people who were able to unscrew, transport, and reconnect its physical base.

The return of hashrate did not mean the return of the same industry. Machines changed hands, some firms entered new markets, and regions took on the economic and political risk that China had previously concentrated. Location estimates are furthermore based on pool data and technical signals, not on a perfect census of every device. When later measurements again showed Chinese hashrate, it was not possible to determine precisely from them how many machines had circumvented the ban and how they were operating. The honest conclusion is therefore not that geography ceased to matter. It is the awareness that the mining map is fluid, incomplete, and yet essential for assessing shared vulnerabilities.