Cango Posts $452.8M Loss in First Year as Bitcoin Miner

Cango Inc posted a $452.8 million net loss for 2025 — its first full year as a bitcoin miner — a headline number that looks brutal until you realize the company is already executing its exit from the business. The NYSE-listed firm generated $688.1 million in revenue during the year, but transformation costs, equipment impairments, and fair-value swings on bitcoin collateral obliterated the bottom line. Now, months into 2026, Cango has sold a massive chunk of its digital asset reserves and is steering hard toward artificial intelligence infrastructure.

Revenue vs. Reality: What the 2025 Numbers Actually Show

Mining generated $675.5 million — more than 98% of Cango's total 2025 revenue. On the surface, that looks like a functioning operation. The problem is what happened below the revenue line.

Total operating costs and expenses hit $1.1 billion for the year. In the fourth quarter alone, expenses reached $456 million, driven by an $81.4 million impairment loss on mining machines and — the real gut punch — a $171.4 million loss from changes in the fair value of bitcoin collateral. You can mine all day, but when your collateral is repriced against you at the wrong moment in the cycle, the arithmetic gets ugly fast.

CFO Michael Zhang addressed the results in a statement on Monday, attributing the annual net loss to non-recurring transformation costs and market-driven fair-value adjustments. He added that the company worked on reducing leverage through an adjusted treasury policy and securing equity infusion to manage volatility and fund new investments. That framing — non-recurring, transformation, one-time — is doing a lot of work in that sentence. The $452.8 million figure is the figure, regardless of what accounting bucket you assign it to.

Cango Inc entered the bitcoin mining sector in November 2024, acquiring 32 EH/s of computing power and launching operations across North America, the Middle East, South America, and East Africa. The company still runs its original international automobile trading business, which contributed a symbolic $4.8 million in Q4 revenue — less than 1% of total quarterly mining output. The auto business didn't carry the company into mining, and it's not carrying it out.

How Does a First-Year Bitcoin Miner Lose $452 Million?

What caused Cango's net loss in its first year of bitcoin mining?

Cango's bitcoin miner net loss for 2025 came from three converging forces: the enormous capital cost of building a global mining operation from a standing start, impairment charges on equipment as the competitive landscape shifted post-halving, and bitcoin's price volatility feeding directly into fair-value adjustments on its collateralized holdings. None of these are unique to Cango — every first-year miner faces them — but for a company still running its inaugural year, they hit harder than for an established operator with amortized infrastructure.

Production-wise, the operation was actually running well. Cango mined 6,594.6 BTC in 2025, averaging 18.07 BTC per day. In Q4, production ticked higher to 1,718.3 BTC — 18.68 BTC per day — though costs rose in lockstep. By the close of December 2025, cumulative production since the company's late-2024 inception had reached 7,528.4 BTC. For a miner that didn't exist before Q4 2024, that's a real number.

The honest reading of 2025, though, is that Cango spent the year building infrastructure it was already planning to repurpose. The transformation costs baked into the $452.8 million loss aren't purely operational growing pains — they're partly the price of a strategic about-face that was in motion before the fiscal year was even finished. When you're already planning the pivot, every dollar spent on mining equipment has a shorter useful life than the depreciation schedule assumes.

The 2024 halving is the backdrop to all of this. It cut block rewards in half at exactly the moment Cango was deploying capital to enter the sector. The company bought in at elevated competitive intensity, built out a multi-continent operation, and still couldn't make the unit economics work at the bottom line. That's a useful data point for anyone tracking the mining industry's profitability trajectory through 2025.

Selling Bitcoin to Buy AI: The Pivot That's Already Underway

Following the close of fiscal 2025, Cango sold approximately $305 million worth of bitcoin last month — cutting its digital asset reserves by 60%. The stated purpose: repay debt and bankroll its push into AI compute infrastructure through a subsidiary called EcoHash. CEO Paul Yu was direct about the direction.

Yu described EcoHash as leveraging Cango's expertise in scalable computing and energy networks to deliver flexible, cost-effective AI inference solutions. Initial site retrofits are already underway, he said, with products ready for deployment.

Call it strategic foresight or call it a company getting out while the exit is still viable — either way, Cango is not waiting around. The 60% reserve reduction is not a minor portfolio adjustment. That's a decisive move that tells you more about where management thinks the returns are coming from than anything in the earnings press release.

The logic behind the pivot isn't hard to follow: miners already have large-scale power infrastructure, cooling systems, and physical facilities optimized for dense compute loads. AI inference workloads — particularly the kind that hyperscalers and enterprise clients are increasingly outsourcing — fit that physical profile. The capital cost of retrofitting is lower than building from scratch, which is the thesis every miner-turned-AI-company is currently running.

We are advancing our pivot to become an AI infrastructure provider. Through EcoHash, we are leveraging our expertise in scalable computing and energy networks to deliver flexible, cost-effective AI inference solutions. With initial site retrofits underway and product ready for deployment, we are positioned to execute with focus and strategic discipline in the new era.

Cango Joins a Wave of Miners Swapping Bitcoin for AI Capital

Cango is not doing this alone. A growing cluster of publicly traded miners are monetizing bitcoin treasuries to fund data center and AI infrastructure expansions — and the moves are accelerating across the sector.

Core Scientific expects to sell substantially all of its bitcoin holdings in 2026, with the majority of sales anticipated in Q1, to fund AI and high-performance computing expansion. The company held 2,537 bitcoin worth $222 million at year-end 2025 and had already sold roughly 1,900 bitcoin for approximately $175 million in January, according to its earnings call. That's a fast liquidation pace.

MARA Holdings — the largest public bitcoin miner by coins held — widened its treasury policy for 2026 to permit the sale of accumulated reserves. The revision was disclosed in a 10-K filing with the SEC this month and marks a departure from the company's long-standing approach of treating mined bitcoin as a permanent long-term asset. MARA held 53,822 BTC worth $4.7 billion as of December 31, 2025. When the most aggressive bitcoin hoarder in the public markets opens the door to selling, you have to take seriously the idea that the industry's accumulation thesis is softening.

The pattern here — mine bitcoin, accumulate reserves, sell reserves to fund AI — is either a rational capital allocation response to post-halving profitability pressure, or it's a sector-wide bet on a narrative that may not deliver the margins everyone is modeling. If AI inference turns out to be as competitive and capital-intensive as bitcoin mining, these companies may find themselves in exactly the same position in 2027 that Cango found itself in for 2025: strong revenue, ugly net loss, scrambling for the next pivot.