Cango - Q4 2025

Transcript

Operator (participant)

After today's presentation, there will be an opportunity to ask questions. Please note this event is being recorded. I would now like to turn the conference over to Mr. Paul Yu, Chief Executive Officer. Please go ahead.

Paul Yu (CEO)

Thank you. Hello, everyone, and welcome to Cango's fourth quarter and full year 2025 earnings call. 2025 marks a landmark year in our company's history. Our first year of transformation since pivoting to Bitcoin mining in November 2024. It was a year of accelerated execution, and we accomplished several critical objectives. First, asset restructuring and global deployment. Through a series of transactions, we relocated our asset from traditional auto finance business to our Bitcoin mining operations within six months. This helped us build a global distributed mining network. Second, leadership and management. To align with our new strategy, we have strengthened our board and management team with seasoned industry professionals. They have brought deep expertise and established networks in both digital assets and infrastructure, which has sharpened our competitive edge in the sector. Third, listing structure optimization.

During the year, we transitioned from an ADR listing to a direct stock listing. This move lays a solid foundation for us to access a broader range of capital market tools, reach a broader base of investors, and reduce holding costs for existing shareholders. Operationally, 2025 showed a clear execution discipline. Despite significant market volatility in the second half of the year, we maintained professional standards across core metrics, including hash rate scale, Bitcoin production, and miner uptime. In the fourth quarter of 2025, we recorded total revenue of $179 million, and produced 1,718.3 Bitcoin. For the full year, total revenue reached $688 million, with Bitcoin production totaling 6,595.6. As economy of scale took hold, we achieved strong revenue growth and posted positive EBITDA for the full year.

The net loss attributable to shareholders for 2025 was $622 million, mainly due to the following factors. First, some non-recurring transformation costs. This includes a one-time book loss of around $169 million from discontinued operations. A further loss of $257 million came from impairment loss from mining equipment and the company acquired and settled in equity triggered by us by the significant appreciation in Cango's share price between signing and delivery. Towards the end of fourth quarter, the price of Bitcoin and other cryptocurrencies declined sharply, driven by external macroeconomic factors and geopolitical tensions. This resulted in a fair value loss of net $96.5 million on our Bitcoin holdings and an additional impairment provision of $81 million on mining machines as a result of the downward price impact on their fair value.

In the early stages of our transformation, constrained by our CapEx capabilities, we adopted a co-location model to rapidly secure a large share of the Bitcoin network hash rate. We quickly built a hash rate of 50 exahash per second, capturing approximately 4%-5% of the global network. However, computation intensified globally and our cash cost per Bitcoin mined approached a high of $84,000 in the fourth quarter 2025. Recognizing further price pressure heading into 2026, we took prudent action. We reduced debt exposure, recovered liquidity, and began phasing out inefficient capacity. These steps have strengthened our balance sheet and enhanced operational efficiency as we enter the new year. In February 2026, we strategically sold 4,451 Bitcoin from inventory and used the proceeds to repay loans, reducing our overall debt.

We completed a $10.5 million capital injection from shareholders. Additionally, we signed agreement with Armada Network Limited and Fortune Peak Limited for new funding around totaling $65 million. We expect these steps to progressively strengthen our active base and mitigate potential market volatility risks going forward. On the operations side, we are optimizing our operations by phasing out older, high energy consuming mining machines. We're also gradually moving our computing power to regions with lower electricity price. While this will lead to a contraction in our total hash rate scale in the short term, it will effectively improve the energy efficiency of our overall fleet, lower cost per coin, and enhancing our resilience against dramatic, drastic market fluctuations. Finally, many of you have asked about our AI business transformation.

Our efforts to reduce existing debt, strengthen equity capital, and optimize Bitcoin mining operations have created the necessary flexibility to really make progress on AI. On that note, we have officially established EcoHash, a wholly owned subsidiary based in Texas, dedicated to high-performance computing and AI inference. Leveraging our accumulated experience in large scale deployment and management of distributed computing infrastructure, as well as our broadly partnered global energy network of Bitcoin mining sites, we will launch standardized modular AI computing nodes aiming to provide highly flexible and cost-effective solutions for long-tail AI inference demand. As of today, we are making steady progress on feasibility studies and preparatory work. Let me share a few updates. On the infrastructure front, we have initiated the first phase retrofit of our owned mining site in Georgia, USA, for standardized AI nodes deployment.

On the product side, our containerized GPU computing solutions also reached the leverage deliverable stage. Our objective is to leverage our existing accessible, skilled energy network to provide flexible and intelligent computing power to support the digital economy. In 2025, we demonstrated the speed of our transformation. In 2026, we will demonstrate our resilience and our ability to adapt and evolve. While the current macroeconomic environment presents challenges, we also see long-term opportunity. The logic behind our decisions is clear, proactive adjustment, disciplined execution, and commitment to the AI era. With that, I will turn the call to Michael Zhang, our Chief Financial Officer, to take you through the financials in more detail.

Michael Zhang (CFO)

Thanks, Paul. Hello, everyone, and welcome to our fourth quarter and full year 2025 earnings call. Before I start to review our financials, please note that unless otherwise stated, all amounts discussed are in US dollars. Total revenue in the fourth quarter were $179.5 million. For the full year, revenue reached $688.1 million. Revenue during the quarter from the Bitcoin mining business was $172.4 million, with a total of 1,718.3 Bitcoins mined during the period. The average cost to mine Bitcoin, excluding depreciation of mining machine, was $84,552 per coin, with all-in cost of $106,251 per coin.

For the full year, revenue from the Bitcoin mining business was $675.5 million, with a total of 6,594.6 Bitcoins mined during the year. The average cost to mine Bitcoin, excluding depreciation of mining machine, was $79,707 per coin, with all-in cost at $97,272 per coin. Revenue from our automobile trading business was $4.8 million in the fourth quarter and $9.8 million for the full year. Now, let's move on to our cost and expenses. Cost of revenue exclusive of depreciation in the fourth quarter was $155.3 million and $543.3 million for the full year.

Depreciation in the fourth quarter was $38.1 million and $116.6 million for the full year. General and administrative expenses in the fourth quarter was $9.9 million and $28.9 million for the full year.

Impairment loss from mining machine in the fourth quarter was $81.4 million and $338.3 million for the full year. Loss from change in fair value of receivable for Bitcoin collateral in the fourth quarter was $171.4 million and $96.5 million for the full year. Operating loss for the quarter was $276.6 million, with a net loss from continuing operations of $285 million in the fourth quarter. For the full year, the operating loss was $437.1 million, and net loss from continuing operations was $452.8 million. On a non-GAAP basis, adjusted EBITDA for the full year was $24.5 million. Moving on to our balance sheet.

As of December 31, 2025, we had cash and cash equivalents of $41.2 million. Our balance sheet also includes $663 million of receivables for Bitcoin collateral. In terms of operational assets, we carry our mining machines at a net value of $248.7 million of depreciation. On the liability side, we had $557.6 million in long-term debt. Together, these figures represent a core component of our financial structure as we close the fourth quarter of 2025. This concludes our prepared remarks. Operator, we are now ready to take questions.

Operator (participant)

We will now begin the question and answer session. To ask a question, you may press star then one on your touchtone phone. If you're using a speakerphone, please pick up your handset before pressing the key. To withdraw your question, please press star then two. If you would like to state your question in Chinese, you may do so, but then please also restate your question in English. Your first question today comes from Pingyue Wu with CITIC Securities. Please go ahead.

Pingyue Wu (Analyst)

Hi. Thank you, management team, for taking my question. This is Pingyue from CITIC Securities. My first question is, the company recently launched the EcoHash, a subsidiary focused on HPC and AI inference compute service. How do EcoHash position itself in a highly competitive AI compute market? What is the core logic behind your approach compared with traditional data centers? Thank you.

Paul Yu (CEO)

Thank you for your question. EcoHash is not designed to replace traditional hyperscale data centers. Instead, we focus on targeted opportunities within specific segments of the AI compute market. Hyperscale facilities are built for large-scale centralized model training workloads. Those projects require significant upfront capital and construction timelines that can span several years. By contrast, our initial focus is on AI inference and generative AI workloads. These use cases have distributed demand and are sensitive to latency. These use cases often require flexible deployment of compute nodes closer to end users, rather than relying solely on massive centralized facilities. Our approach centers on resource reuse and modular design. Notably, we can leverage our global energy network connected to our existing Bitcoin mining sites. From this space, we are deploying standardized and modular AI compute nodes that can be deployed much faster than traditional data center infrastructure.

This model shortens timelines, lowers upfront construction costs, and delivers compute capacity more efficiently. For now, EcoHash remains in early phase of model validations and technical integration. We are taking a measured approach. Our primary objective is to explore how we can fully leverage our existing energy infrastructure to participate in the rapidly growing AI inference market with a model that is asset light, quick to deploy, and able to deliver stronger capital efficiency. Thank you.

Pingyue Wu (Analyst)

Thank you. My second question is, the company sold more than half of its Bitcoin holdings in February 2026. This appears to be a notable shift from the mine and hold strategy highlighted in the third quarter. My question is, what drives this decision? Thank you.

Paul Yu (CEO)

Thank you for your question. Actually, we understand that investors are watching this shift very closely. From a financial management perspective, our shift from a pure Bitcoin accumulation strategy toward more strategic monetization reflects our focus on maintaining balance sheet strength in the current market environment. Given the heightened volatility in Bitcoin prices since late in the fourth quarter and into early 2026, we made a decision in February to monetize a portion of our Bitcoin holdings.

Michael Zhang (CFO)

The objective was to reduce financial leverage and further optimize our balance sheet, ensuring that the company remains well-positioned to then navigate potential continued market volatility. It is also worth noting that we are seeing a broader shift across the mining industry. In a cyclical environment with increasing volatility, maintaining excessive exposure to a single asset can introduce unnecessary balance sheet risk. As a result, a more balanced approach between long-term asset exposure and financial stability is becoming increasingly common across the sector. At the same time, the company is entering a critical phase in validating the diversification of our computing power business. We see AI computing as an important long-term growth driver. By making this strategic adjustment, we are also creating greater financial and operational flexibility to support continued development and scaling of our AI-related initiatives. Thank you.

Pingyue Wu (Analyst)

Thank you. Thank you very much.

Operator (participant)

Your next question comes from Ming Zhang with Minsheng Securities. Please go ahead.

Ming Zhang (Analyst)

Morning, this is Zhang from CITIC, and thanks for this opportunity. I have two questions. The first one is that we noticed that the company's net position remained relatively high at the end of the reporting period, and the Bitcoin prices has been volatile recently. I mean, if the price remains weak, how will the company fund the development of its AI business? You mentioned that a $10.5 million capital injection from the controlling shareholders and $65 million equity financing arrangement. How will these funds be allocated between the mining business and the AI initiatives in 2026? Second question is that, regarding your development of AI compute network, what is the expected timeline over the next year?

When could the business begin contribution in full revenue? That's my question. Thank you.

Michael Zhang (CFO)

Thank you. I will take your first question. We have taken proactive steps to strengthen our balance sheet, as I just mentioned. We recently sold 4,451 Bitcoins from inventory and used the proceeds to partially repay outstanding loans. This reduced our overall financial leverage and increased flexibility as we advance our AI initiatives. At the same time, we completed the closing of our $10.1 million capital injection and enter into agreement with Armada Network Limited and Fortune Peak Limited for an additional $65 million equity investment. Once this new round of financing is completed, the company's leverage ratio will decline further, resulting in a stronger balance sheet that better support the development of our AI business. For the AI segment, we intend to follow a disciplined and phased investment strategy.

Phase one is product and business model validation. During this stage, we will rely primarily on internal capital. We will conduct pilot infrastructure upgrades and deploy compute products at our own mining site in Georgia. Phase two begins once the model is validated. We plan to establish several backbone nodes in collaboration with selected partner mining facility. In these cases, infrastructure upgrades will be carried out jointly with site operators and project-level structured financing, such as GPU-backed financing, may be used to support expansion. Phase three occurs as the compute network gradually forms and begins generating stable operating cash flow. At that point, we expect to use a flexible mix of equity and debt financing to fund the next stage of growth. Thank you.

Paul Yu (CEO)

Regarding the AI timeline, our approach to the AI business remains measured and pragmatic. The initiative is still in the early stages, so our near-term focus is on validating the commercial models and re-evaluating unit economics. Given where we are in the pilot phase, it would be premature to issue specific revenue forecast at this time. Currently, our most tangible progress is taking place at our self-operated mining site in Georgia. We are conducting a small-scale pilot project to deploy the first batch of standardized AI compute nodes there. This will allow us to validate the technical architecture and gather operational data. The broader 1.2 GW energy network that we can access serves a long-term strategic resource pool. However, this represents a medium to long-term capacity option. It is not necessarily a commitment to immediate large scale capital expenditure.

For now, we are focused on gradually validating the model through the Georgia pilot, while ensuring that overall liquidity and financial stability remain intact. Thank you.

Operator (participant)

Thank you. Your next question comes from Marco Zhang with Gelonghui Research. Please go ahead.

Marco Zhang (Analyst)

Hi, this is Marco from Gelonghui. Thanks for taking my question. Congrats on your successful transformation last year. I have two questions here. First, you increased your hash rate from 3 exahash per second to 50 in 2025. Do you have specific hash rate expansion targets for 2026?

Paul Yu (CEO)

For 2026, our focus is efficiency rather than scale. Our goal is to maintain a healthy cash flow and strong risk resilience across market cycles. In 2025, we produced approximately 6,600 BTC, which validated the strength of our existing operational footprint. For 2026, we will implement a prioritized efficiency strategy. This starts with systematically phasing out older, high energy consumption mining rigs. We will also gradually relocate some of our hash rate to regions with more competitive electricity pricing. This optimization may result in a temporary reduction in total hash rate in the year, in the near term. However, it will greatly improve fleet-wide energy efficiency, lower cost per Bitcoin mined, and strengthen our resilience during periods of volatility. Our objective is to build a more resilient compute portfolio by phasing out inefficient capacity and freeing up liquidity.

We strengthen our balance sheet. This also preserve capital resources that may later support our ongoing AI transition. Thank you.

Marco Zhang (Analyst)

Got it. Thanks. My second question is for our modeling purpose, looking ahead, from your perspective, how should investors evaluate Cango's valuation framework in 2026 and beyond? Should the company be viewed primarily as a mining company or as an AI infrastructure provider? Thanks.

Paul Yu (CEO)

Thank you for your question. Bitcoin mining remains our foundation, while AI represents our incremental growth engine. Over time, we believe investors may increasingly evaluate our performance through metrics such as revenue per megawatt. Whether we are deploying power into Bitcoin mining or AI compute, the underlying principle is the same: converting energy into economic value. We will allocate resources toward whichever segment delivers stronger returns. In that sense, Cango is evolving into a flexible compute platform. We can dynamically allocate energy-backed compute capacity across different markets based on return potential. Thank you.

Operator (participant)

The next question comes from Kevin Dede with H.C. Wainwright. Please go ahead.

Kevin Dede (Managing Director)

Hi, Paul. Hi, Michael. Thank you so much for having me on the call. I'd like to quiz you a little bit more, Paul, please, on detail behind your AI pilot in Georgia. How long do you think it will take you to validate the model? Do you think you might be able to turn to live market revenue sometime within this calendar year?

Paul Yu (CEO)

Okay.

Simon Tang (CIO)

Hi, Kevin. This is Simon Tang, Chief Investment Officer here. I'll step in and take this question, if that's okay.

Kevin Dede (Managing Director)

You're perfect, Simon. Thank you.

Simon Tang (CIO)

Hi. Great to reconnect. In terms of the AI pilot in Georgia, because this is gonna be a modular containerized solution, so it should be relatively quick. We anticipate that from breaking ground to overall coming on stream, it should take somewhere between 4-6 months, and this is a relatively conservative estimate. Secondly to answer a second question, in terms of revenue generation within this year, yes, we do anticipate that there is going to be some sort of revenue generated from this business model this year.

Kevin Dede (Managing Director)

Okay. Simon, as you look at optimizing the Bitcoin mining fleet, how much of your 50 exahash would you classify as inefficient? How much capital do you think you'll be able to allocate toward replacing the fleet versus investment in AI infrastructure?

Simon Tang (CIO)

Got it. I think when we talk about inefficient, it's a function of both the mining machine model as well as the power price that we have in place for that particular site, right? It's very difficult for us to quantify at the moment, holistically how much of that we would classify as inefficient.

Overall, in terms of the general direction of this business, as Paul and Michael have alluded to earlier, we're looking at a variety of ways to increase the economics and outcome of this business, whether it be swapping some of these machines for newer models or whether it be moving them to some more cost-efficient sites or whether it be through renegotiating some of these contracts, which are either expiring or which are just generally being renegotiated as well. In terms of the capital allocation for this effort, in terms of new capital investment, it's going to be more geared on the AI side.

On the mining machine side, currently we do not have any significant plans for allocating more investment into procuring new machines.

Kevin Dede (Managing Director)

Okay. The auto business seemed to kick up pretty nicely in the fourth quarter, and I was hoping you could help me understand whether or not there was some seasonality there, how you expect this year, 2026, to progress. Do you think you should see an overall lifting in revenue there? And then, please give us some indication on where you are on profitability in that business.

Michael Zhang (CFO)

Thank you, Kevin, for your question. I think yes, we see that there is a very quick development of our overseas auto trading business overall. We do expect that there is still a significant growth related to that sector in the coming year. As Simon just mentioned, since we allocate the majority of our capital into the AI sectors, I mean, the AI initiatives. We do not expect that we will allocate further capital into the auto trading sectors. It's actually a type of internal growth.

Simon Tang (CIO)

I mean, the organic growth by our auto trading business line itself. I think, yeah, and also, it's also related to the demand side. You know that there is due to the geopolitical reasons and, actually, the price volatility related to the energy. I think it's very difficult for us to give a very clear view about, I mean, the performance, the financial performance for the automotive automobile trading business in the next year. Thank you.

Kevin Dede (Managing Director)

Oh, thank you, Michael, for taking my questions. I appreciate it. Paul, I'd like to offer my congratulations. It's really pretty amazing on how quickly you're able to transform the company, and I have no doubt that you'll be able to work out all the problems you may run into in addressing HPC and AI. Congratulations on all the progress and good luck in the future.

Simon Tang (CIO)

Thank you, Kevin. Thank you for your time.

Operator (participant)

Your next question comes from William Gregozeski with Greenridge Global. Please go ahead.

William Gregozeski (President and Director of Research)

Hi. Thanks. I just wanted to ask about how much of the Georgia facility is being allocated to the phase one pilot, and are you able to give some kind of rough sense as to how much money is being spent on that phase one pilot?

Juliet Ye (Senior Director and Head of Communications)

Hi, Bill. This is Juliet. Thank you for your question. I'll try to take this one. With regard to the mining site in Georgia, we're currently starting the retrofitting work for the site, basically because we're actually adopting a modular kind of like approach. We don't expect to turn like a major kind of hash rate or MW into AI at this stage. That one. It should be used as a showcase. We will say 1-2 MW to show the possibility to show the things we can do with our existing infrastructure.

In terms of CapEx, basically we've been running demo projects as we actually discussed in previous calls last year in terms of AI transition. We are thinking of a ballpark of around like $20 million for one megawatt including GPU. Just in case. It's still in the process of a feasibility study, so we will show more details including numbers when we have the LM site ready probably later this year as just mentioned by Simon. For the retrofitting work, it might take around like 4 to 6 months in a kind of like conservative approach.

I hope that answers your question. Thank you.

William Gregozeski (President and Director of Research)

Yes. Thank you.

Operator (participant)

Thank you. This concludes our question and our session. I would like to turn the conference back over for any closing remarks.

Simon Tang (CIO)

Thank you for joining us. Oh, we're good. Thank you very much. Thank you everyone for joining our earnings call today. Thank you.

Operator (participant)

Thank you. The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.