Cango’s Shift Toward Distributed Compute

Cango Inc. (NYSE: CANG) is being priced like a leftover China-inc auto-finance story that lost relevance years ago. Yet by early 2026, the company had already crossed into a different category entirely: a scaled Bitcoin miner with material cash generation and an active capital allocation strategy that leverages its treasury to fund a global AI compute grid. The market’s problem is that it keeps reading the old label on the jar. The opportunity, if there is one here, sits in that gap between label and contents.

What CANG Actually Became After the Pivot

The cleanest way to understand Cango is to start with what it is not. It is not a growthy used-car export platform, even if AutoCango had a nice quarter. The earnings call itself framed that legacy segment as small and opportunistic, with Q3 revenue around $3.3 million, a rounding error next to the mining line. The center of gravity moved. The company’s Q3 2025 revenue was $224.6 million, and $220.9 million of that came from Bitcoin mining. That mix matters because it tells you where operational attention, capital allocation, and managerial incentives now live.

The second reset is geographical and reputational. The proceeds from divesting the China car-finance operations and the subsequent redeployment into mining and energy infrastructure were not cosmetic. They were a narrative severance. A stock can trade at a discount for years if the market thinks it is exposed to a regulatory regime it cannot price. The company’s strategy, as articulated by Paul Yu, effectively attempted to remove that variable and replace it with something the market can benchmark, namely hashrate, uptime, cost per coin, treasury value, and an infrastructure roadmap.

What makes this pivot unusual is speed. In roughly a year, the business scaled to a deployed 50 EH/s footprint, with a global spread across the Americas, the Middle East, and parts of Africa. That is not a “we are exploring” statement, it is already an operating reality, and it shows up in the income statement. I see that speed as the first underappreciated signal of capability because in mining, execution is often the only moat you get before energy economics kick in.

The Mining Engine: Profitable Today, Imperfect by Design

Cango’s Q3 numbers were not just “crypto up, revenues up.” They were the kind of operational block-and-tackle that usually separates real miners from promotional ones. The company produced 1,930.8 BTC in Q3, averaging 21 BTC per day, and generated operating income of $43.5 million with net income of $37.3 million. Adjusted EBITDA came in at $80.1 million. Those are meaningful figures for a company that the market is valuing like it is barely alive.

The obvious pushback is margins. On the last earnings call, the CFO disclosed an average cost to mine Bitcoin excluding depreciation of $81,072 per coin, and an all-in cost around $99,383 per coin. That is not best-in-class compared to the lower-cost peer set that owns more of its power. But the strategic tradeoff is not hard to interpret. 

Cango prioritized strategic velocity through an asset-light hosted approach to reach scale quickly, then planned to backfill the margin structure through selective site ownership and energy projects. The Georgia 50 MW facility acquisition is a good tell here because management explicitly described it as an “upgrade” of the asset-light model rather than a contradiction, with ownership used where it improves stability, power economics, and AI-grade upgrade optionality.

There is also an execution nuance that investors tend to misread. Deployed hashrate was 50 EH/s, but operational hashrate was lower during integration, running up from roughly 40.9 EH/s in July to 44.9 EH/s in September and then to about 46 EH/s in October. Management attributed the gap to relocation downtime, commissioning at new sites, and the unavoidable reality of weather and curtailment. That explanation matters because it frames operational variance as industry-normal rather than deterioration. If the company can keep uptime above 90%, which it claimed, then the remaining question becomes primarily economic, not operational.

In other words, today’s mining business is a functioning cash machine with a calculated cost structure designed to secure market dominance at unprecedented speed. The bull case assumes that the next phase is not more scale, it is margin conversion.

The Second Act: Edge Compute as the Real Re-Rating Mechanism

Cango’s thesis has little to do with Bitcoin mining suddenly gaining institutional favor. Mining on its own rarely earns durable multiples. What matters here is how management is positioning mining as a transitional layer rather than the destination. Cango is not pitching itself as the next hyperscale data-center operator, and that distinction is critical. Instead of competing with AWS or Azure on centralized capacity, it is targeting flexible, distributed compute units that aggregate dispersed GPU resources and repackage them for smaller and mid-sized enterprises. That may sound subtle, but it places Cango in a part of the compute stack that hyperscalers are structurally disincentivized to serve.

Large cloud providers thrive on scale and uniformity. Smaller, fragmented deployments with uneven demand and location-specific power constraints are messy and capital inefficient for them. Cango is deliberately leaning into that messiness. I think the company’s real edge is not brand or software, but operational discipline combined with energy access. It already runs a globally distributed infrastructure footprint through its mining operations, and it is actively building energy projects that can later power AI workloads at lower unit costs than pure-rent GPU clouds. In that sense, mining is not the business model. It is an operating school.

This is where the energy projects matter. The initiatives in Oman and Indonesia are often dismissed as “green narrative,” but I see them as the economic bridge between mining and compute. They are expected to come online within one to two years, which is early enough to influence how investors think about future cost structures. The Georgia facility is even more important in the near term. It demonstrates that a site can operate as a mining asset today while being structurally compatible with AI compute later. Management’s emphasis on pilots with strict IRR thresholds also stands out. In a sector where the phrase “AI infrastructure” often suspends financial discipline, this suggests a more controlled capital allocation mindset.

If edge compute begins generating even modest, visible revenue, the framing changes quickly. Mining becomes the funding layer, energy becomes the moat, and compute becomes the multiple driver. That sequencing is not promotional. It is simply logical, and it is the sequence management itself has outlined.

In early 2026, Cango successfully optimized its treasury structure, significantly enhancing liquidity and effectively neutralizing historical financial leverage. This strategic de-leveraging, reinforced by a recent US$75.5 million capital injection—including a combined US$65 million personal commitment from our Chairman and Board Director—underscores a profound internal conviction in the company’s trajectory. With a robust, high-liquidity foundation, Cango is now uniquely positioned to prioritize its capital allocation toward the rapid scaling of its global AI distributed compute grid.

Cango’s transition into the AI infrastructure sector has achieved significant tangible progress. The company has established EcoHash Technology LLC in Dallas as its global operational hub and appointed Jack Jin (formerly of Zoom) as CTO to spearhead its AI technical architecture. Notably, Cango's modular GPU compute nodes have already achieved operational break-even in early pilot projects, validating the commercial viability of its “plug-and-play” infrastructure solution.

Takeaway

Cango does not need the market to fall in love with Bitcoin mining. It needs the market to recognize that the company already built scale, already generates real earnings, and now holds a treasury that can act as both floor and financing option. The upside case depends on whether management can translate distributed operational competence into a distributed compute product with actual customers. If that happens, the stock stops trading like yesterday’s label and starts trading like tomorrow’s platform.