Cango - Earnings Call - Q3 2025

Transcript

Speaker 2

Please note today's event is being recorded. I would now like to turn the conference over to Paul Yu, CEO. Please go ahead.

Speaker 4

Hello, everyone. Welcome to Cango's third quarter 2025 earnings call. This quarter marks the one-year anniversary of our strategic transformation into a Bitcoin miner, an important milestone for the company. Today, I will reveal our third quarter results and share how Cango continues to create long-term value in a rapidly changing market environment. During third quarter, we remained focused on our core mining operations, further strengthening Cango's position as a skilled and operationally disciplined Bitcoin miner. This is clearly reflected in our financial performance. In the third quarter, total revenue reached $225 million, up 60.6% sequentially. Operating income was $43.5 million, and net income was $37.3 million. Today, Cango operates a deployed hash rate of 50 exahash globally, positioning us among the leading miners worldwide.

In the third quarter, we produced 1,930.8 Bitcoins, averaging 21 Bitcoins per day, up 37.5% in total output and 36% in daily production compared with the second quarter 2025. Leveraging our asset-light model, we built a competitive global footprint across the Americas, the Middle East, and Africa in just one year. In our mining operations, we continue to execute our strategy of which prioritize hash rate optimization over expansion by refreshing older, less energy-efficient models to the T21 and S21 series and disciplined operations. We significantly improved average operating hash rate from 40.91 exahash in July to 44.85 exahash in September and further to 46.09 exahash in October, with efficiency surpassing 90%. In August, we also acquired a 50-megawatt mining facility in the state of Georgia, lowering per-unit operating costs and building dedicated energy infrastructure to support our long-term strategy.

The current market environment remains volatile with significant fluctuations in Bitcoin prices. Cango is closely monitoring these dynamics and will continue to manage our deployed output and explore partnership models to mitigate market risks and enhance operating stability. While consolidating our core business, we also clarify our long-term strategy. Building a global distributed AI compute network powered by green energy, we will use Bitcoin mining as a practical on-ramp towards our energy and compute ambitions, following the sequence of from Bitcoin mining to energy exercise and from operational depth to AI compute deployment. In the third quarter, we executed our first roadmap with strict financial discipline, launching small-scale pilots with clear technical and IRR thresholds across both energy and AI compute.

Our clean energy projects in Oman and Indonesia are now underway and are expected to be commissioned within the next one or two years, providing strategic support for subsequent AI infrastructure development. In AI compute, Cango is taking a differentiated approach. Instead of building large centralized data centers, we focus on flexible distributed compute units. In practical, this will integrate dispersed GPU resources into standardized compute pools and break them into smaller units tailored to the needs of small and mid-sized enterprises. This approach is enabled by two core advantages: our distributed operational expertise and our global energy footprints, allowing us to execute a unique asset-light plus edge-first strategy.

In terms of governance, we have assembled a new leadership team with deep experience in digital infrastructure and finance and completed the transition from an ADR listing to a direct listing on the NYSE to enhance transparency and reduce shareholder transaction costs. These initiatives provide strong support for our next phase of development. Lastly, let me briefly update you on our legacy business. Our used car export platform, AutoCango, delivered strong performance this quarter with revenue of $3.3 million, up 90% sequentially. The platform remains asset-light and continues to scale, connecting buyers from Africa, the Middle East, and Eastern Europe with quality vehicle inventory from China. With that, I will now turn the call over to Michael Zhang, our Chief Financial Officer, to take you through the financials in more detail.

Speaker 3

Thanks, Paul. Hello, everyone, and welcome to our third quarter 2025 earnings call. Before I begin the review of our financials, please note that starting this quarter, we will begin reporting US dollars, which better reflects the profile of our revenues and profit following the investiture of our China asset in May 2025. Unless otherwise specified, all amounts discussed are in US dollars. Total revenue in the third quarter of 2025 was $224.6 million, up 60.6% sequentially. Revenue from the Bitcoin mining business in the third quarter of 2025 was $220.9 million, with a total of 1,930.8 Bitcoins mined during the period, up 50.9% and 37.5% respectively on a sequential basis. The average cost to mining Bitcoins, excluding depreciation of mining machines, was $81,072 per coin, with oil cost at $99,383 per coin. Revenue from our automotive training business was $3.3 million in the third quarter of 2025.

Now let's move on to cost and expenses during the quarter. Cost of revenues exclusive for depreciation in the third quarter of 2025 was $162.6 million. Depreciation in the third quarter of 2025 was $35.4 million. General and administrative expenses in the third quarter were $6 million. We recorded operating income of $43.5 million and net income of $37.3 million in the third quarter of 2025, compared with an operating loss of $1.2 million and a net loss of $9.5 million in the same period last year. On a non-GAAP basis, adjusted EBITDA for the third quarter of 2025 was $80.1 million, compared with $1.2 million in the same period last year. Moving on to our balance sheet. As of September 30, 2025, we had cash and cash equivalents of $44.9 million. Our balance sheet also reflects $660 million receivables for Bitcoin collateral.

In terms of operational asset, we carry our mining machine at a net value of $365.7 million after depreciation. On the liability side, we had $405.1 million in long-term debt owed to a related party. Together, these figures represent the core components of our financial structure as we close the third quarter of 2025. This concludes our prepared remarks. Operator, we are now ready to take questions.

Speaker 2

Thank you. If you would like to ask a question, please press star and one on your telephone keypad. If you are using a speakerphone, we ask that you please pick up your handset before pressing the keys. If at any time your question has been addressed and you would like to withdraw your question, please press star then two. At this time, we will pause for just a moment to assemble our roster. Today's first question comes from Emerson Zhou with Goldman Sachs. Please go ahead.

Speaker 6

Thank you, management team. I have two questions. Number one, given the current Bitcoin prices, will the company consider selling Bitcoin holdings to fund new business expansion or manage market risk or support operation needs? The second question is, you mentioned that equipment operates improved energy efficiency, but we see October's operational hash rate, which was 46.6 exahash, which is still below the deployed hash rate of 50 exahash. What are the main factors behind this gap, and when do you expect full utilization? Thank you.

Speaker 3

Thank you for the question. I think I will take the first one. This quarter, we continue to follow a mining holding strategy, retaining all mined Bitcoin as part of our strategic reserve. We've seen a heightened volatility recently, driven by tight market liquidity and increased uncertainty around the U.S. rate cut paths. However, we believe the fundamental thesis for Bitcoin as a core reserve asset remains intact under a broader macro backdrop. We will take a flexible approach across debt, equity, and other financing channels to support our development of new initiatives. Our Bitcoin reserve will also provide a meaningful liquidity buffer and optionality for structured financing if needed. Thank you.

Thank you for your question. I'm going to answer the question regarding the operation efficiency. After completing the acquisition of 18 exahash in late June, we reached full-scale operations at 50 exahash for the first time in July. During the initial integration phase, we experienced temporary downtime due to cross-state machine relocations and ongoing power system commissioning at the newly acquired site. This factor created short-term pressure on uptime. Our operations team responded quickly and through system-level optimization, uptime has now stabilized above 90%, which is considered industry-leading and demonstrates the strength of our operational capabilities. It is important to note that external factors such as extreme weather and grid curtailment periodically affect miner availability. This is an industry-wide reality, and achieving 100% uptime is not feasible. Among comparable industry peers, uptime above 90% is regarded as a strong performance benchmark.

Going forward, we will continue enhancing efficiency through upgrades to our intelligent operations and maintenance systems, while replacing low-efficiency hardware where appropriate. Thank you.

Speaker 2

Thank you. Our next question comes from Qingyue with Citech Securities. Please go ahead.

Speaker 0

Thank you, management team, for taking my question. This is Qingyue from Citech Securities, and I have two questions. The first question is related to the debt structure. The company mentioned converting short-term debt into long-term debt. Can you elaborate on the financial benefits of this shift, and what is your current cost of debt? Secondly, my question is related to AI CapEx. Some capital-intensive data center operators have undergone significant value reset. Some people are questioning whether AI CapEx is entering bubble territory. Given that you are now entering into the AI infrastructure space, how do you view this risk? Thank you.

Speaker 3

Thank you for the questions. I will take the first one. Through this optimization of our debt maturity profile, our liability will now primarily compose our long-term borrowings. This better aligns our capital structure with our strategy of building Bitcoin reserves through self-mining, enhancing balance sheet stability, and reducing financial risk. At present, we plan borrowing costs to remain in the 7%-8% annualized range, and this level is expected to remain stable following the maturity structure adjustment. Thank you.

Regarding the second question, it is true that the market is reassessing returns on AI investments, particularly for hyperscaler data centers with high leverage, heavy CapEx, and long contract cycles. At the demand level, AI inference and industry-specific applications are still expanding rapidly. The demand mix may evolve, but long-term compute demand is not disappearing. Entering later gives us the benefit of observing market shifts and avoiding high leverage extension at the tail end of the previous cycle. Our advantage lies in a lighter asset and leverage structure and a more distributed edge-oriented operating footprint. We evaluate and monitor AI project investments, potential returns, and cash flow profiles at every stage. This allows us to dynamically adjust course, optimize capital efficiency, and preserve strategic feasibility at all times. Thank you.

Speaker 2

Thank you. Our next question comes from Joey Chai with Guojin Securities. Please go ahead.

Speaker 7

Thanks, Manjulu, for taking my question. I have two questions as well. The first one, Bitcoin has pulled back sharply from its all-time high in October. How does this affect your operating pace for Q4 and 2026? With your current cash position and BTC holdings, how long can you operate under extreme market conditions? Do you have a worst-case plan? The second one, the Joder site is self-owned. Does this contradict the asset-like model? Will future expansion favor owned sites or leased sites? Thanks.

Speaker 3

Thank you for your questions. Yes, we conduct frequent internal stress tests. Thanks to our asset-like models and operational flexibility, we can dynamically adjust and even shut down high-cost sites under extreme scenarios to reallocate hash power and control operating expenses. We have the flexibility to adjust our BTC holding strategy as needed.

Regarding the Joder site, it's important to clarify that Joder site acquisition is not a strategic pivot, but rather an upgrade of our SLI model. We chose to acquire this site because it aligns with our long-term needs around securing low-cost power, gaining great stability, and deepening our infrastructure operation capabilities. Looking ahead, we will continue to follow a balanced model of lease-first with selective strategic acquisitions. Leasing will remain our primary path for rapid expansion and geographic diversification. We evaluate potential acquisitions against strict criteria: power cost, scalability for AI-grade data center upgrades, and regulatory stability. We believe owning a portion of K sites is essential to maintaining long-term cost advantages and supporting our strategic transition. On capital allocation, we prioritize efficiency over sheer scale.

All cash flows will be directed first to initiatives that strengthen our structure cost advantages, such as acquiring sites to lower power costs or upgrading underperforming compute equipment. At the same time, we remain disciplined in managing our asset structure, evaluating and monitoring leverage and financial discipline through rigorous financial and operational metrics. Thank you.

Speaker 2

Thank you. Our next question comes from Kevin Dede at HC Wainwright. Please go ahead.

Speaker 1

Hi, this is Daniel Marine on for Kevin Dede. We're curious how Cango feels it's best to address the HPC market, whether that be through cloud compute or a PowerShell model. If the recent pullback in Bitcoin could at all accelerate this, how are you guys thinking about timelines with those two ventures? Thank you.

Thank you for your question. In AI compute, Cango is taking a differentiated approach. Instead of building large centralized data centers, we focus on flexible distributed compute units. In practice, this will integrate dispersed GPU resources into standardized compute pools and break them into smaller units tailored to the needs of small and mid-sized enterprises. This approach is enabled by two core advantages: our distributed operational expertise and our global energy footprint, allowing us to execute a unique SLI plus edge-first strategy. Thank you.

Speaker 2

Thank you. That's all the questions we have time for today, and this concludes today's conference call. We thank you all for attending today's presentation. You may now disconnect your lines and have a wonderful day.