The sixth-largest bitcoin miner by hash rate sold its entire active treasury across February and March, slashing mining costs by 19 per cent while securing $75 million in fresh capital to convert its global infrastructure to AI compute.
Cango Inc., the sixth-largest bitcoin miner by hash rate, sold 6,451 bitcoin worth approximately $442 million across February and March, applying the entire sum to retire crypto-collateralised loans as the New York-listed company accelerates a pivot from mining digital currency to selling computing power to the artificial intelligence industry.
The liquidation — disclosed in Cango's March operational update filed on 8 April — dwarfs the periodic treasury sales that other miners have made to cover operating costs. This wasn't trimming around the edges. Cango moved 4,451 BTC in February alone for roughly $305 million in USDT, then sold another 2,000 BTC in March to bring its outstanding loan balances down to $30.6 million. The company ended March with just 1,025 bitcoin in its treasury, valued at around $73 million — a residual position for a firm that, six months ago, was accumulating aggressively.
The sell-off accomplished two things simultaneously. It cleared the debt that had been secured against Cango's bitcoin holdings, removing the liquidation risk that leveraged mining companies face when prices drop sharply. And it freed up the company's 40-plus mining sites worldwide — representing 27.9 exahashes per second of self-mining capacity and another 9 exahashes under leasing agreements — to be repurposed for AI workloads without the overhang of crypto-denominated liabilities.
Cango's mining costs fell 19.3 per cent during the same period, from $84,552 per bitcoin in Q4 2025 to $68,215 per bitcoin, a reduction the company attributed to "simplified production methods prioritising margin stability." With bitcoin trading around $71,000 at the time of the March sales, those margins were thin — a few thousand dollars per coin before accounting for depreciation, overhead, and financing costs. The economics explain the urgency of the pivot better than any strategy slide deck could.
The AI transition is being funded by more than just bitcoin proceeds. Cango's executive team pledged $65 million in equity, and DL Holdings subscribed to a $10 million convertible bond, bringing the total fresh capital to $75 million. The plan, as outlined in the filing, is to use Cango's existing power contracts and physical infrastructure — the data centres, the cooling systems, the grid connections — to provide distributed computing services to AI companies that are desperate for capacity. It's the same playbook that several North American miners have floated, but Cango is executing it at a scale and speed that few have matched.
Cango is not alone in retreating from the HODL strategy that defined the post-2024 halving miner playbook. MARA Holdings and Bitdeer have both been selling bitcoin in recent months; Strategy (formerly MicroStrategy) posted $14.5 billion in unrealised losses on its bitcoin position in Q1. The halving compressed margins across the industry, and the subsequent price decline from October's $126,000 all-time high to the current $70,000 range has turned what was a profitable bet into a balance-sheet liability for overleveraged operators.
The market has punished Cango for the transition. The company's stock is down 72 per cent year-to-date, reflecting investor scepticism about whether a Chinese-founded auto finance company that pivoted to bitcoin mining in 2024 can successfully pivot again to AI compute within a single fiscal year. The scepticism is not unreasonable; Cango's track record of rapid strategic shifts suggests a management team chasing whatever sector is generating the most capital markets enthusiasm at the moment. Auto loans, then bitcoin mining, now AI — each pivot has arrived precisely when the previous business model's economics turned unfavourable.
But the underlying asset — cheap power access and physical data centre infrastructure — has genuine value regardless of what workloads it serves. GPU clusters for AI training and inference require exactly the same inputs that bitcoin miners consume: electricity, cooling, and space. If Cango can secure the hardware and the customer contracts, the infrastructure conversion is technically straightforward. The question is whether the company has the operational expertise and the client relationships to compete in AI hosting against established players like CoreWeave, Applied Digital, and the hyperscalers themselves.
Cango's 37 exahashes of total operating capacity made it responsible for 2.82 per cent of bitcoin's global network hash rate. That share will shrink as the company redirects facilities to AI, which in turn concentrates mining power further among the remaining large operators — a dynamic that runs counter to the decentralisation ethos that bitcoin's architecture was designed to promote, but one that the industry's economics have made inevitable.
The company's March filing left open the possibility of retaining some mining operations alongside AI compute, a hedged approach that acknowledges the difficulty of timing a permanent exit from an asset class as volatile as bitcoin. For now, though, the numbers tell a clear story: Cango sold nearly all its bitcoin, paid off its loans, and raised fresh capital. The HODL era, for at least one major miner, is over.
