New Narrative in Mining: Cango Sells Bitcoin Reserves, How AI Computing Power Is Reshaping Mining Company Valuation Models?

In early 2026, a decision by Bitcoin mining company Cango drew widespread market attention: this mining firm, which held over 7,528 BTC at the end of 2025, sold 4,451 BTC in a single transaction in early February, cashing out approximately $305 million to pay down debt and support a strategic shift toward AI computing infrastructure. This move was not an isolated event but reflected a collective choice across the Bitcoin mining industry under current market conditions. When mining costs surpass the coin price, Bitcoin reserves—once considered core assets—are being redefined as adjustable strategic resources.

Why do mining companies reduce their Bitcoin holdings when prices are low?

Cango’s reduction in holdings directly responds to a harsh market reality: the economics of mining have fundamentally reversed. Industry data as of March 2026 shows that the total cost to mine one Bitcoin is about $87,000, while the market price is only around $67,000, meaning each Bitcoin produced results in a net loss of $20,000. For Cango, its average full mining cost (including depreciation) in Q3 2025 was as high as $99,000 per BTC, well above current market prices.

In this context, the logic of holding Bitcoin as a “store of value” is broken. Cango explicitly stated in its official announcement that the sale aimed to reduce financial leverage, strengthen its balance sheet, and fund strategic expansion into AI computing infrastructure. As of February 28, 2026, Cango’s Bitcoin holdings had decreased to 3,313.4 BTC, while its deployed hash rate remained at 50 EH/s, indicating the company is rebalancing its asset structure—from “holding digital assets” to “controlling physical computing power.”

What is the core driver behind the industry’s shift toward AI computing power?

There is a natural physical link between mining and AI computing: electricity and infrastructure. Cango’s strategic roadmap clearly reveals this mechanism: leveraging its globally connected grid infrastructure to provide distributed computing capacity for the AI industry.

The core of this transformation is the re-pricing of computing resources. Bitcoin mining revenue is compressed by coin price volatility, difficulty adjustments, and equipment depreciation, whereas AI data centers offer long-term contracts of 10 to 15 years, enterprise clients with investment-grade credit (like Microsoft, Meta), and stable, predictable dollar cash flows. Cango’s plan involves three phases: short-term deployment of containerized GPU nodes at existing sites to serve small and medium-sized enterprises; mid-term development of a software orchestration platform to integrate distributed resources; and long-term goal to become a mature AI infrastructure platform. To accelerate this process, the company appointed former Zoom tech expert Jack Jin as Chief Technology Officer for AI, whose experience in GPU cluster deployment directly supports the new strategy.

What are the costs and trade-offs of this structural transformation for mining companies?

The transition is not without costs. Cango disclosed in its February 2026 operational update that its average operating hash rate for the month was 34.55 EH/s, below the deployed 50 EH/s, due to “fleet optimization and related temporary shutdowns.” This reflects the pain of capacity adjustment during the transition from ASIC miners to GPU-based computing. About 31% of the hash rate was offline for upgrades, implying short-term revenue losses.

Deeper trade-offs involve a shift in corporate positioning. Cango was once the second-largest publicly listed Bitcoin miner globally, with a “HODL + mining accumulation” model that reinforced itself during bull markets: rising coin prices increased asset net worth, supporting further hash rate expansion. But the market environment in 2026 forces a reassessment of this model. Selling Bitcoin reserves means relinquishing potential gains from future price rebounds in exchange for current financial stability and cash flow for the transition. It’s a strategic choice balancing time preference and risk exposure—exchanging future uncertainty for current structural survival.

What does the collective “sell Bitcoin, shift to AI” trend among miners mean for the crypto market?

From a market structure perspective, the industry’s collective shift could have profound effects on Bitcoin’s supply and demand dynamics. Historically, miners have been the largest “structural sellers” of Bitcoin—they need to sell regularly to cover electricity and operational costs. As miners pivot to AI services, with stable dollar income from hosting contracts, they no longer need to passively sell Bitcoin and may even become buyers.

On-chain data already reflects this change. In early 2026, corporate treasuries holding Bitcoin showed three consecutive weeks of net reduction; Cango alone sold over 54% of its holdings in two weeks. While short-term selling pressure may weigh on prices, if this trend continues, the market’s largest “natural short-seller” is systematically exiting, which is a positive long-term signal for Bitcoin’s supply structure. The Hash Ribbon indicator shows that since late November 2025, the miner capitulation period has been one of the longest in history, often signaling market bottoms as capacity is cleared.

How might the future evolution of mining and AI computing power unfold?

Looking ahead, the relationship between mining and AI computing power could evolve into a dynamic balancing mechanism. MARA’s hybrid model offers a template: utilizing the same electrical infrastructure to flexibly switch between Bitcoin mining and AI GPU services. When electricity prices are low, compute power is used for mining; during peak AI demand, resources shift to GPU inference services. In this mode, Bitcoin mining becomes a flexible load, reducing when AI needs high capacity and filling the gap when AI demand subsides, thus optimizing electricity costs and maximizing revenue.

Cango’s approach is more focused on complete transformation. The company explicitly aims to become a “global distributed inference computing grid,” leveraging its existing 40 global sites and connected infrastructure as the physical foundation. The “edge” resources accumulated through Bitcoin mining—sites near cheap electricity but far from traditional data centers—are well suited for distributed AI inference deployment. The future of mining may no longer be just a “race for hash power,” but evolving into an operator of distributed computing infrastructure.

What are the potential risks and limitations of this transformation path?

The path is not without challenges. First, technical differences pose real hurdles: Bitcoin mining relies on ASIC chips, while AI computing requires GPU clusters and supporting orchestration software. Although Cango has appointed technical leaders, bridging the capability gap from mining operations to AI infrastructure will take time.

Second, market patience is limited. While selling Bitcoin improves the balance sheet, the company still faces cash flow pressures; analysts report a leverage-free cash flow of -$252 million. Capital-intensive AI infrastructure investments have long payback periods, and tightening financing conditions could strain liquidity.

Finally, competition is intensifying. Besides Cango, firms like Core Scientific and Bitdeer are also transitioning into AI. As more players enter, competition for quality electricity, GPU supply, and customer contracts will become fierce. Establishing a differentiated competitive advantage before “hash power oversupply” becomes critical for each miner’s success.

Summary

Cango’s strategic shift from Bitcoin mining to AI computing power exemplifies the cyclical adjustment and structural reshaping of the crypto industry. When the narrative of “holding as faith” faces economic realities, miners are redefining their core assets and business boundaries. This transition may exert short-term selling pressure but could ultimately reshape Bitcoin’s supply-demand structure—transforming miners from “passive sellers” into “infrastructure operators,” leading to a more mature market landscape.

FAQ

Q: Does Cango’s sale of Bitcoin mean a bearish outlook on Bitcoin?

A: Not necessarily. Cango’s reduction is primarily a financial restructuring—reducing leverage and freeing liquidity to invest in AI infrastructure. The company states it remains committed to mining operations and optimizing mining economics. This is an asset allocation adjustment, not a judgment on Bitcoin’s intrinsic value.

Q: How does the industry’s shift to AI affect Bitcoin network security?

A: In the short term, some hashrate will go offline, slowing the network’s total computational power; long-term, this is a healthy capacity cleanup. Inefficient miners unable to bear high costs will exit, leaving more efficient, professional operators, which can enhance overall network security.

Q: Can mining companies successfully transition into AI infrastructure?

A: Success depends on multiple factors: technical migration capability, capital support, and market competition. Cango’s advantages include its global grid-connected infrastructure and phased implementation plan. However, the transition is a long-term effort requiring ongoing progress in technology deployment and customer acquisition.