Chinese companies from traditional industries buying land in the US to mine bitcoin and sell AI computing power has become one of the most profitable pivots on the US stock market these past two years.

One US-listed Chinese company that started out in domestic auto lending pulled off exactly that transition, and its share price and profile climbed along the way. On August 11 last year, it paid $19.5 million in cash for a large mining farm in the United States.

Barely a year after the purchase, a fire sent two workers to a burn center.

The Fire on the Morning of October 22

At 10:45 a.m. on October 22, fire alarms went off at 881 Commerce Drive in Soperton, Georgia.

The site is a bitcoin mining farm in the town. The building caught fire suddenly, though the blaze stayed relatively small and local firefighters put it out shortly after arriving.

The fire did injure people: two workers on site were burned. Their injuries were serious enough that responders called in a helicopter to rush them to a specialist burn center in Augusta. Locally, only burns above a certain severity threshold warrant a helicopter transfer to a dedicated burn center.

As of October 23, when local outlet WSB-TV published its report, neither the workers' identities nor the extent of their injuries had been made public, and the cause of the fire was described only as "still under investigation."

The mining farm belongs to Cango, a US-listed company with Chinese roots that trades on the NYSE under the ticker CANG. Anyone who followed China's early fintech boom will recognize the name: Cango began in domestic auto finance, connecting car buyers with dealerships to arrange loans, and listed in the US in 2018.

How Does an Auto-Loan Company End Up Mining Bitcoin

The pivot was forced on the company, and it was executed decisively.

As regulators tightened their grip on China's lending sector in recent years, the auto loan business got harder and Cango's original core operation shrank sharply.

Looking for a way out, Cango made a bold call in late 2024: turn the ship around entirely and bet on bitcoin mining, deploying mining rigs across the United States, East Africa and Oman in one sweep.

The timing was right. Bitcoin prices rose, network-wide computing power climbed with them, and Cango's stock followed. To outside observers, it looked like a well-played comeback.

With momentum behind it, Cango stepped up spending. On August 11, 2025, it announced it was paying $19.5 million in cash to take full ownership of the Soperton, Georgia mining farm where the fire broke out.

Cango already had rigs at the site under a third-party hosting arrangement before the purchase. Buying the facility outright was about becoming the landlord itself.

How the 50 Megawatts Were Split

The plan at the time was clear: the farm has 50 megawatts of total power capacity, with 30 megawatts reserved for Cango's own bitcoin mining and the remaining 20 megawatts leased out for rental income.

Megawatts measure power capacity, and in the mining and data center business they are effectively a proxy for capacity — the power you have determines how many machines you can install. Fifty megawatts is not small, but it is well short of the scale of the largest miners.

Mining for yourself while leasing out the rest has become a fairly common playbook among pivoting miners over the past two years. The mining side rides bitcoin's swings, the leasing side collects contracted payments, and the two together make revenue steadier than mining alone.

An AI Business Tucked Inside the Mining Farm

The more valuable business sits inside the same building.

On September 3 this year, less than two months before the fire, Cango's EcoHash unit announced it had carved out a dedicated zone within the Soperton farm to launch commercial GPU computing services, with supporting infrastructure able to handle up to 3 megawatts.

In plain terms, part of the mining farm is being converted into an AI data center and leased to customers who need to train AI models.

Mining bitcoin for cash while building an AI computing power business for the narrative is precisely the combination US equity investors like best right now.

Three megawatts out of 50 is a small share, but it carries different weight. Mining is a cyclical business; GPU leasing runs on long-term contracts and offers a more predictable cash flow outlook. That is why pivoting miners like to foreground the AI half of the story in their earnings reports.

Machines and Money Travel Easily. Local Rules Do Not.

Start with what is known: there is no evidence so far that the fire was caused by mismanagement, and the official investigation is ongoing. How many machines were damaged, how long operations will be halted, and whether the newly launched AI business was affected are all absent from official statements and press coverage.

Even without conclusions, anyone who has run a mining site knows that operating a large farm is far harder than outsiders assume.

Plenty of people imagine mining is just plugging machines into power. In reality it is a capital-intensive manufacturing operation that punishes weak fundamentals. Large numbers of high-draw machines are packed into a single building, running around the clock, throwing off enormous heat and pushing wiring and grid connections to their limits.

That means cooling fans, aging wiring, sprinkler heads — any small detail going wrong can start a fire. Reports of fires at North American mining farms have surfaced every few weeks over the past two years.

More to the point, running physical facilities abroad means operating under an entirely different set of rules.

Many Chinese companies, drawn by cheap US power and land, have bought mining farms and built data centers in the United States. Machines and capital move easily; local fire safety codes, worker protection regimes, and insurance claims processes bear no resemblance to those at home.

Take this accident. Two workers were airlifted to hospital, and by standard US practice the Occupational Safety and Health Administration (OSHA) will likely open an investigation, followed by workers' compensation claims, insurance loss assessments, and the question of how to reassure residents of a small town. Handling all of that is typically far more complicated than repairing burned-out machines.

Moving from asset-light consumer lending to asset-heavy mining operations overseas, Cango has been fast and accurate in reading the shift and repositioning. But once that leap lands in physical industry on foreign soil, the real test is no longer bitcoin's price or the size of the AI wave — it is managing a building in another country and keeping a close eye on the wiring and the fire hydrants. That is the hard, unglamorous work no company going abroad can skip.