Digi Power X - Earnings Call - Q3 2025

Transcript

Operator (participant)

Good morning, and welcome to DigiPower X Inc.'s Third Quarter 2025 Financial Results Conference Call. Please note that this event is being recorded, and a transcript will be available on DigiPower X Inc.'s website. At this time, all participants are in a listen-only mode. A brief question-and-answer session will follow the formal presentation. Unless noted otherwise, all amounts referred to during the call are denominated in U.S. dollars. Certain comments made during this call may include forward-looking statements, or forward-looking information within the meaning of applicable U.S. and Canadian securities laws. Such statements and information reflect current expectations, and as such, are subject to a variety of risks and uncertainties that could cause actual results to differ materially from current expectations.

Those risks and uncertainties include, but are not limited to, factors discussed in DigiPower X Inc.'s report on Form 6-K for the three and nine months ended September 30th, 2025, and the annual report for the year ended December 31st, 2024, as well as the company's other disclosure documents. Except to the extent required by applicable law, DigiPower X undertakes no obligation to publicly update or review any forward-looking statements or information. During the call, management may make reference to certain non-IFRS measures that are not separately defined under IFRS, such as EBITDA and adjusted EBITDA. Management believes that those non-IFRS measures, when considered in conjunction with IFRS financial measures, provide useful information for both management and investors.

Reconciliations between IFRS and non-IFRS measures are presented in the tables accompanying the press release highlighting DigiPower X financial results as of the quarter ended September 30th, 2025, have been filed and made accessible under the company's continuous disclosure profile on SEDAR+ at www.sedarplus.ca, and are also available on the SEC's EDGAR website at www.sec.gov/edgar. I would now like to turn the call over to Michel Amar, CEO of DigiPower X. Please go ahead, sir.

Michel Amar (CEO)

Good morning, everyone. Q3 2025 was a transformational quarter for DigiPower X, as we strengthened our balance sheet, accelerated our shift into AI infrastructure and delivered positive net earnings. Quarter highlights, working capital went from $500,000 in Q3 2024 up to $15 million this quarter. Net income is +$300,000 versus a $6.4 million loss last year. EBITDA, $1.9 million positive, adjusted EBITDA $0.8 million positive. Digital assets, BTC holdings up 143% to 97 Bitcoin. Ethereum, 1,000 tokens. Total digital currency value, $15.4 million, up 213% year-over-year. Warrants reduced from 8.8 million warrants to 2.6 million outstanding. Debt, no long-term debt. CapEx, $3.1 million in Q3, $9.5 million year-to-date, primarily tied to tier 3 AI data center conversion. Strategic and operational progress. Our first ARMS 200 tier 3 AI pod assembly began in Q4 2025, and will be online in Q1 2026.

Approved 60-MW load study in New York, enabling future AI expansions. First NVIDIA B200 cluster built with Supermicro on track for Q1 2026 activation. Advancing long-term colocation and AI compute agreement with multiple potential customers. Beginning January 2026, ARMS 200 modules will be deployed across all tier 3 sites. Energy revenue grew 112% year-over-year to $8.7 million. Cost of revenue and depreciation reduced by $9.3 million year-to-date. NeoCloud Z, our GPU-as-a-Service platform, launching January 2026. Financial position today, over $90 million in cash, BTC, ETH, and equivalents, the strongest liquidity in company history. Our liquidity equals more than 1/3 of our market cap. This capital fully supports the 2026 AI infrastructure build-out. AI transition across power assets. Phase I deployment, Q1 2026, 5 MW. Phase II, Q2 2026, 15 MW. Phase III, Q3 2026, an additional 30 MW.

By Q4 2026, we would have deployed 55 MW, which is 40 MW critical IT load. Current power availability, Alabama, 55 MW. New York Upstate, 141.7 MW for a total of close to 200 MW of power available for 2026. North Carolina, an additional 200 MW anticipated in 2028. DigiPower X is now firmly positioned as a next-generation tier 3 AI infrastructure company. With the ARMS 200 platform, the NeoCloud Z platform, major power capacity and tier 3 deployment on the way, we are building a scalable high-density AI compute ecosystem ready, to serve both emerging AI companies and enterprise customers. We've never been better positioned for growth.

Operator (participant)

Thank you. We'll now be conducting a question-and-answer session. Our first question comes from Jeremy Hayes. What are your expectations in AI revenues in 2026 and 2027?

Michel Amar (CEO)

Great question, Jeremy. Related to the scaling up of our first module infrastructure, ARMS 200 in Alabama, which we intend to complete quarter after quarter, up to 40 MW of IT load, mostly 75%-80% will be colocation. The colocation range is about $140-$150 a kilowatt-hour per month. To give you a little bit of a size of income, if we average 20 MW because we are scaling up, we estimate roughly about $50 million in revenues average for the year 2026 in colocation. The second part of AI revenues will be our GPU-as-a-service through our NeoCloud Z platform, which we built with Supermicro. We anticipate about 1,024 B200, B300 NVIDIA chips, GPUs online, generating with a scale-up, about $15 million.

The growth estimate for 2026 related to the execution of the infrastructure through our ARMS 200 system solution, should be roughly about $65 million. On top of that, we will have the existing Bitcoin mining, which should be even with last year, we have to add a fourth stream of income, which is our energy sales to the grid through our power plant, which to give you a sense for this year so far, we are at $8.9 million year-to-date selling energy, which for Q3 was about $3.1 million energy sales, which represents on this quarter, Q3, over 35% of our revenues.

Operator (participant)

Thank you. What are the current debts?

Michel Amar (CEO)

We do not hold any long-term debts. We are current. We have no obligations, no trigger that could hurt the company. We are completely debt-free. I think we have the lowest payable ever in our company.

Operator (participant)

Thank you. How is your current cash holding?

Michel Amar (CEO)

We are holding a little bit more than $90 million total. Most of it is in cash. We have a few holdings of crypto Bitcoin mining, about 115 or 116 Bitcoins. We have 1,000 Ethereum. Everything else is in cash. That accounts today, a little bit above $90 million, which allows us to really develop our modular system. We actually, this year, year-to-date, already spent close to $10 million. The first development of Q1 has already been paid for. The first ARMS 200, which will be assembled in the next few weeks in Alabama and that is rated three, will be on end of December. Tested the first week of January, along with the Supermicro team as we plan to have this ARMS 200 system on the partnership book of Supermicro, which will be a very interesting development for us.

Operator (participant)

Thank you. Our next question comes from the line of Patrick Gray. Can you please describe the ARMS AI-ready modular solution platform?

Michel Amar (CEO)

Great. DigiPower X, wholly owned subsidiary US Data Centers Inc., has developed a proprietary ARMS platform, ARMS for AI-ready modular system. We developed that platform early this year, I would say 10, 11 months ago with Supermicro. It's really an incredible product system that we can also develop for our own sites and for other people's sites. We can deliver this system within 180 days. It's built for all kinds of chips, primarily for the NVIDIA B200, B300. It can be customized for other servers or other chips. It is really a modular system that has three product lines, 1 MW, 5 MW, and 10 MW. We can expand it much faster than a hyperscale building that takes a couple of years. In our opinion, for 2026, this is an incredible demonstration of what we can do at the speed we can do.

Of course, the reason why we can develop it in a faster way than a hyperscale, is that we already have the high-voltage substations. We already have the connection with the grid power. Our power plants generate power as well, so we are connected already with power. All we need to do is to insert this modular system that we're rated three, ready for redundancy and uptime in standard in compliance with the tier 3. I think you're allowed to 1.6 hours of downtime a year, so we got rated three. 2026 is going to be a real transformational year for us, and we expect to enter our full cycle of developing AI.

Operator (participant)

Thank you. Can you please go into some detail on your relationship with Supermicro Computers Inc.?

Michel Amar (CEO)

We developed a very interesting relationship with Supermicro. They are a pretty large company. They have a very good relationship with NVIDIA. In order to leverage their human resources and resources, we chose Supermicro and we integrated their optimized service rack directly to our ARMS system pods. We designed it that way. Supermicro is doing a very big part of the job by supplying racks ready to go.

We also partner with them on all the software integration, and it's really very, very helpful. That allows us to have a very big team on our side with a very small GNA, because we have a very low SGNA and we are leveraging the Supermicro team. They are coming to our side actually end of this month, November 28. They will come and start to prep with our people the setup of our modular systems, both in Upstate and in Alabama.

Operator (participant)

Thank you. Our next question comes from the line of Anthony Sphere. Can you please describe the retail compute platform NeoCloud, who the potential customers or end users will be?

Michel Amar (CEO)

We developed that platform. Actually, Alec Amar developed it along with again, Supermicro. It's geared for the smaller players, not for the hyperscaler customers. It's for the AI developers or ML engineers start-up companies. There are plenty of startup companies now that need access to an infrastructure. It is hard for them to get access to a hyperscaler. Research institutions and universities, non-hyperscaler corporations priced out of AWS, Azure, or NVIDIA cloud capacity, HPC workloads for simulation, modeling, and data processing. It's kind of our retail service, GPU-as-a-service. We expect to develop that portal or that direct vertical GPU-as-a-service to be 20%-25% of our total revenue. The breakdown would be 75% colocation with longer contracts. We are in negotiation now, anywhere from five to 10- to 15-year contracts.

There is a lot of demand for 2026, early 2027. We are trying to get the best scenario or the most solid contract, 15-year contract. It's like a modified lease on AI infrastructure. Simultaneously, it is ready, it will be online in January. We are developing this platform, NeoCloud Z, backed by Supermicro. We will be able to integrate multiple customers, all kinds of customers for deep learning, for any kind of applications. That will result in a much higher margin. The colocation margins are pretty set and revenues. The direct GPU at service provides very high margins, so blended, we will have a very healthy margin in the AI business.

Operator (participant)

Thank you. This concludes our Q&A session. That concludes our call today. We thank you for your interest and participation. You may now disconnect your lines.