管理層發言
Hello, and welcome to today's webcast on HIVE Digital Technologies financial results for the 6 months ended September 30, 2025. My name is Nathan Fast, Director of Marketing and Branding at HIVE, and I'm pleased to be your moderator for today's call. Before we get started on Slide 2, I would like to briefly note the disclosures for today's presentation. Except for statements of historical fact, this presentation contains forward-looking statements within the meaning of the U.S. Private Securities Litigation Reform Act of 1995. Words such as expects, believes and similar expressions identify these statements. Actual results could differ materially, and we disclaim any obligation to update them, except as required by law. For a full discussion of risk factors, please refer to our most recent SEC filings at sec.gov. In addition to discussing results that are calculated in accordance with GAAP, we will also reference certain non-GAAP financial measures, including adjusted EBITDA, adjusted net income and free cash flow. Management uses these metrics to evaluate operating performance and believes they provide investors with additional insight, and they are presented for supplemental purposes only and should not be considered in isolation from GAAP results. Reconciliations to the nearest GAAP measures are included in the appendix to this presentation and in the press release and Form 8-K furnished to the SEC. On the next slide, I'm pleased to introduce today's presenters, Frank Holmes, Executive Chairman; Aydin Kilic, President and CEO; and Darcy Daubaras, Chief Financial Officer. I would now like to hand the presentation over to Mr. Frank Holmes from a macro recap of the quarter. Frank?
Thank you, Nathan. Let's get into my macro recap, and I'm going to try to weave in some of the most recent past couple of weeks' sentiment, unpacking the drivers of short-term negative sentiment in data centers and AI infrastructure, which I believe is just setting up for a phenomenal buy for us as an industry as a whole. I'm going to try to explain this in this presentation and why HIVE is so uniquely positioned in this macro theme. Next, please. Well, HIVE's journey to really accelerate its growth in Bitcoin mining was stalled after Putin invaded Ukraine and sources of hydro and geothermal electricity around the world all of a sudden went shut down or very small. There was no scaling ability until the opportunity a little over 18 months ago when I reached out to the President of Paraguay and flew down to meet him because they have surplus electricity. I'm going to try to walk you through sort of this journey, and that this President is really quite unique. President Santiago Peña is a former rugby player. He is 47 years old, studied at Columbia University in New York, worked in Washington with the IMF and has had an incredible journey in Paraguay in building the country and becoming recognized. And every time I go down, I find something new that I love more about the country. The country is very unique and very pro‑America. So to give you an idea of all those negative news about Latin America, with Venezuela and some of the drama of the Panama Canal and then we have, for sure, Colombia's President. I think it's really important to look at how Communism and Socialism has penetrated deeply into Central and South America, but not in Paraguay. And I think that that's what's really important for investors and why we felt so confident this is where HIVE would scale its Bitcoin operations, sourcing green energy, surplus energy and really also helping out the overall economy grow, and I'll show you some of the other unique things we've done in schools that are nearby us. President Santiago Peña was the sole Latin American Head of State present at the 2025 Gaza Peace Summit. So this is something else that's really important, he's pro-Taiwan, and one of the largest suppliers of food commodities for Taiwan. For a population of 6 million, they produce food that can be exported to feed about 60 million. Next, please. So here's our President of the country, Gabriel Lamas, who's an electrical engineer, worked for the government's utility company. As you can see, he's very tall compared to the President who's like 6'2, Gabriel is like 6'4, 6'5 and he's just been a great leader of maximizing the whole scaling, the concept of scaling and coordinating and working with our CEO, Aydin Kilic, who'll speak in a few minutes. And Luke Rossy, who's our Chief Operating Officer; and Bill Gray, our CTO. Gabriel has done a phenomenal job accelerating the scaling from 6 to 25 — the journey to 25 we've announced for 24, and we're very close here to 25. So we feel very confident of hitting that. You can see here, Gabriel was recently with them. And there's a picture when I first met the President a little over 18 months ago. The economy is one of the strongest and steadily growing economies in Latin America. So that's another real positive sign for the country. Next, please. As a money manager, I have always said that government policies are a precursor to change. Peña's policies for fiscal development are outstanding and he has pushed to make sure that every child is fed and then upgrading the school system, but we're helping with that. We're helping, especially in rural areas. This story I'm trying to point to is that Paraguay is powering HIVE's high velocity ascent to 25 exahash. What's interesting is that yesterday — sorry, 2 days ago, November 12 — that is Hayes Day because in 1870, there was a tragic war, a trilateral war with Uruguay, Argentina and Brazil trying to take over this massive green belt country, and they wiped out genocide all men and boys to the age of 10. It was tragic, and it was President Hayes that came in and resolved that they kept about 60% of their land. To this day, the country celebrates November 12, Hayes Day, and the largest department in the country is called Presidente Hayes. So there's a very strong bond with America that when I talk to investors, the average American is not aware of. Next, please. So part of that journey in addition to rapidly scaling and building out data center campuses across the country, building a buzz is also creating lots of local jobs in construction. We've shown you a visual of lighting the streets in Valenzuela, a city near us. But then it's also for the kids, and this is a thank you from a former leader, a mayor of Valenzuela for lighting up the streets and making the streets safe at night for women and children. Next, please. So when we went out and saw the schools, which are very close to us, I'm talking about 0.25 mile, 0.5 mile away. We started in this process of upgrading these schools and putting in air conditioning, new electrical work, painting, new chalkboards, infrastructure and you can see here a PlayStation for kids so that they can go and play and enjoy these facilities and also for embracing data centers. There's a big, great disconnect even in very advanced countries like Sweden that take it for granted all the apps on your mobile phone and Netflix and Spotify. They think it's all free, and they don't realize no data centers, no digital realm that we all live in and we rely on. So this is a visual of relating this to the children of Paraguay. Next, please. You can see the picture, this is the key people here. You can see Gabriel Lamas is in the middle, and he's kneeling down because he's so tall. But behind is myself, and you can see to my right, Luke Rossy, and you can see to the left, is Aydin Kilic, our CEO, who's also an electrical engineer like Gabriel Lamas is. And beside Gabriel is Craig Tavares. Craig is our President of Buzz, which is building out our AI business. And you can also see Johanna Thörnblad, who's the President for Sweden. And you can see one of our independent directors and other key people that have been very much involved in the growth and acceleration. Investment banker Jamie Brown is there. So there are a lot of key people that are helping drive this. Behind the visual is the largest, biggest hydro dam in the Western Hemisphere. It's 5 miles long. It's epic in size, producing over 10 exahash of electricity, 5 for Brazil, 5 for Paraguay. Paraguay is a small population. They've been selling some of their electricity to Argentina, which has been only really horrific for their fiscal management of the country. So it's a small country, and that's what's good about Bitcoin mining. Not only do we build this infrastructure — you need substations from this high-voltage electricity, and we build those. Then we build a campus of data centers, and then we end up hiring 50 young engineers that there's no real job opportunity for them. Now there is with HIVE. So it's really, for us, an exciting part of our journey of creating this from little kids' education to creating career paths for young engineers. Next, please. So on that journey of going to Latin America, I really want to show you that this year has been a phenomenal year for Latin America, the 40 biggest market cap stocks versus the S&P. They've crushed it. They've really outperformed the overall S&P 500. I think that's important for investors to recognize. Next, please. So on this big build-out, I call it an arms race. China has significantly surpassed America in building data centers. But before that, they built power. The power came from hydroelectricity. They got the water from the glaciers in the Himalayas and rerouting water from Tibet down to creating dams after dams and building out a huge grid. America has to get caught up, and it is. Under the stewardship and leadership of President Trump, as the CEO of NVIDIA says impressed him the most about Trump is that he embraced the need for AI and power and data centers. Here is a classic where Bitcoin miners were the first to go. This is in West Texas, Sweet Abilene, Texas, where Crusoe was using flare gas, surplus energy from solar, wind and flare gas. Today it is the largest CapEx spend in America and I'm told in the world right now in fast-tracking a $500 billion AI data center vision from former flare gas to a Bitcoin miner. Bitcoin miners were the early pioneers of building the necessary infrastructure wherever there was surplus or additional or wasted electricity and creating tax revenue for the local economy, creating jobs for the local economy. This particular site created 2,500 construction jobs. When that's all done, it will be 500 jobs and a big boom to the local community. Next, please. So Bitcoin mining — the goal is up 2.8% of the global hash rate. We really believe the blue-sky opportunity in Paraguay is probably a gigawatt of electricity. That's the potential when you use the word blue sky and not to go into the moon. Paraguay, no doubt, just makes the perfect place for us to expand our footprint. Backed by the most consistent and stable executive leadership team in the sector, I'm very proud about HIVE. HIVE's CEO is the longest standing. CFO is clearly the longest standing. The turmoil in the industry has been just phenomenal. We've been able to do this and have operating income no matter how volatile Bitcoin has been. We have always been able to squeeze out operating income and have huge expansion, which we've done this year in growth, the most efficient Tier 1 Bitcoin mining data center operator, that's really important for you to recognize spanning 9 time zones in 3 countries and 5 languages. We have the lowest industry SG&A per Bitcoin miner, best-in-class for uptime and efficiency. We know a lot about Bitcoin mining, and we also know in depth and breadth more than these other Bitcoin miners that are going into the space of AI. We've been doing it for 3 years. I believe we're quite successful. It's much more complicated. It's much more detailed and substantially more expensive per megawatt per chip to go build out a Tier 3 HPC or high-performance computing data center for AI. Cambridge University has said it's 33:1 from a Bitcoin mining setup. What Bitcoin mining does is it sources electricity, builds out the necessary infrastructure and then you go as further things are developed in an economy such as dark fiber optics to move the data along. It takes time for that in many emerging countries. For Paraguay, I have to agree with Senator Rubio saying that Paraguay has the potential to become the data center capital of Latin America. We hope to be a significant part of that HPC build-out as we are doing what we're doing now in Bitcoin mining. So we have this proven track record. We're now scaling in AI, and let me walk you through a few comments on that. Next, please. So November 10, HIVE reports October production, 289 BTC, achieves the 24 exahash and completes Paraguay expansion while fueling Canada's Tier 3 and AI data center growth. Next, please. We've shown this growth by buying a former Bitcoin mining data center within the city of Toronto near the airport. So uptime is all the time. Going from Tier 1 to Tier 3 will be much faster than starting from scratch. This is a perfect ideal location, and we feel that this will be a 2-tier payback. We've already built out 10,000 customers around the world, but the unique relationship we have with Bell Canada is transformative besides going from Tier 1 Bitcoin data center to Tier 3, which is much more expensive because you need much more air conditioning and cooling infrastructure. When you think of these H200s and Blackwells, these chips are more intensive in consumption of electricity. So you need much more air conditioning, HVAC, and more support. Those dynamics are important for investors to realize; besides headlines, it takes time to do it. We've done it, and we're now expanding rapidly. Next, please. So when we look at our suite of assets around the world, one of them is that we bought an asset besides Toronto. We bought it in Northern Sweden, Boden near Facebook, and will convert a data center there like we're doing in Toronto. That process has already started. We also have New Brunswick on the border of Maine, and this is a beautiful campus of data centers. We made a press release recently, we bought more land, secured the position to build out our Tier 3, and it will be the biggest pure HPC data center complex in Canada at this time. We're very excited about this. It's a tremendous blue-sky opportunity. As we're going right now for this particular asset, first is to tie up the land. We have the electricity and now to get the engineering, complete drawings so we can start the construction in the first quarter of next year. Next, please. So Buzz high performance overview, HPC. Data center locations today are downtown Montreal and Stockholm, Sweden. We have GPU clusters. We have over 5,000 GPUs and AI cloud services built — purpose-built AI cloud with managed services in agentic refinery. That means that you can provide more services, especially when you have a 600-person sales force at Bell Canada that needs to get caught up quickly, and we're the strategic partner for that. We've already shown what we can do. We're doing $100 million in revenue, and now we're going to scale it over the next 12 months fivefold. Blue sky — we're going through a transformation of Tier 1 to Tier 3 HPC data centers. When we look around the world right now, the back of the envelope is 90 megawatts. If you start looking at other assets we're working on securing, it's much higher. The long-term vision and opportunities for Paraguay are immense. We're excited about this growth opportunity. Unlike other people, we've done it, and we know what it is to build data centers. We know what it is to scale Tier 1 data centers, and we know what it takes to build Tier 3, and we are scaling. Next, please. This was a very important transaction for Buzz HPC in Canada because Bell Canada is the largest telecom in the country. This partnership is to help build sovereign AI infrastructure through the Bell AI Fabric initiative, deploy advanced NVIDIA hardware and networking, deploying NVIDIA Ampere, Hopper and Blackwell GPUs with Quantum-2 InfiniBand across Bell's fiber network and data centers. Canada is a huge opportunity because they're way behind, highly educated, and they've just been slow but now elevated with the AI minister. This is full tilt for the country doing everything to get caught up. The initial deployment and national reach has been a 5-megawatt GPU facility in Manitoba with plans for national expansion, like we talked about Eastern Canada. Data sovereignty, security and sustainability have become very important issues for many of these countries. HIVE has its head office in Texas reporting for Nasdaq and doing financial reporting under GAAP. But the intellectual capital and the Buzz business are Canadian companies that would qualify under Canadian sovereignty and wouldn't be a challenge to the degree that people are worried about privacy or foreign influence. I think this is important to recognize how uniquely HIVE is positioned for this. Next, please. There's our growth. February was $13 million. May was $20 million. We're hoping to get these others deployed as fast as possible to get us up to $40 million in the fourth quarter, which is going to have this run rate. Our year-end is March. Then going into '26 based on all the projections we have from hooking up our suite of NVIDIA chips, looking at those contracts, it looks like a very conservative run rate of $100 million annual run rate with extremely attractive margins for growth. Next, please. So HIVE's top institutional investors are Citadel Advisors, Schwab Crypto Thematic ETF, Amplify Investments, Tidal Investments and Charles Schwab as a discount broker. I remember when a few years back, it used to be Robinhood. Now it's showing up with Charles Schwab, and Charles Schwab bought TD Waterhouse. It's great because they are the biggest broker now. It's great to see broad retail in addition to institutions. We met with the fund managers of most of these companies this week in a conference that Aydin Kilic, our CEO, and I attended in Miami put on by Cantor Fitzgerald called crypto, infrastructure, AI and energy, a big theme going forward. Next, please. We celebrated 4 years on Nasdaq, and it was a major home run. We had 100 people out for it. Ministers from Paraguay and New York City came out for it. I saw the President of Bermuda, auditors, a huge crowd, and we want to thank everyone for helping us celebrate that. There's Aydin hitting the bell and that's my daughter. You can see Craig Tavares, the President of Buzz, and his daughter. It was a home run for us all to be participating. What's happened short term before I pass you on to other executives is sort of AI adoption. There's been some important research from Morgan Stanley about framing the AI boom. I'm a big believer the AI boom is real. Sentiment drives a lot of these markets short term. You can get clusters calling bubble. There are so many analysts who say in bubbleology but do not really understand what is driving the secular market. We're due for this correction, which we're getting, which only sets it up for a more attractive buy. If we were valued at other data centers, and if you apply those valuations to HIVE, HIVE would be a much higher valuation, just to give you an idea of the great opportunity as we continue to scale. It's important to recognize much of this noise is short term, and it gives opportunity as you see some other companies go through corrections. It's part of delays and disappointments on the infrastructure build-out that some are experiencing, but it's not going away. This is an arms race, just like the push by NATO in Europe and Canada and the U.S. of huge military spend, huge rebuilding of military with AI, with GPU chips, and they need data centers. The digital economy will continue to need and grow. We have experienced phantom demand in Bitcoin mining when China shut down mining and everyone from China started making bids for energy all over the world, but they were just phantom attempts to secure some energy. My history of these cycles is that we are still in the early innings of adoption. When you look at a pure data center and the multiples they trade at to revenue and to cash flow, it makes some Bitcoin miners attractively priced. My biased position is HIVE. When I look at multiples to revenue as we power forward with our twin-engine strategy of both Bitcoin mining growth and AI, we have a big growth profile for Bitcoin next year. We're conservative in how we depreciate our ASIC chips faster over 2 years because every 4 years the supply halves. Technology is improving energy efficiency dramatically. Ten years ago an ASIC could be much less efficient; now efficiencies are much better and will continue to improve. By the time of the next halving, I think efficiencies will improve further. AI and technology improvements will allow more energy-efficient data centers and ASIC chips. That means less consumption for the global Bitcoin network and still profitable business and significant role in building the railway tracks for high-performance computing growth. Thank you.
And thank you, Frank, for that executive overview. Frank, of course, is the Co-Founder and Executive Chairman of HIVE. I'm going to get into a bit more of a deep dive on our accomplishments for this quarter and our strategy for 2026. Let's hop into it. So it was a record-breaking quarter for HIVE. $87 million of revenue. Of that, $82 million from Bitcoin mining and over $5 million from our HPC business. So we're a $20 million ARR for our AI cloud business, which is phenomenal and a $42 million gross operating margin for the quarter. That translates to a $31.5 million adjusted EBITDA and a $23.8 million EBITDA with $38 million of depreciation this quarter that leads to a net loss of $15.8 million, which, of course, is on a noncash basis. But moreover, what I'm really proud of is that we realized an 18% ROIC this quarter on an annualized basis. And finally, as of September 30, we had 210 Bitcoin in the treasury, although in addition to that, we have 1,992 Bitcoin pledged. So between our pledged and what we have in the treasury, that's about 2,200 Bitcoin. It's been a phenomenal quarter, and I'm really proud of how our balance sheet deployment to scale the business was, in my opinion, a great success. Let's go to the next slide. We are very serious and very dedicated about having best-in-class ROIC as the numbers this quarter once again affirmed, and it comes from disciplined capital allocation. We always focus to get the best ROI. ASIC purchases are the biggest variable in realizing an ROI and then from their free cash flow investments in the Bitcoin mining business. We're experts in that. Pound for pound, we're the best in the industry at that. We also lead with best-in-class uptime. We run our ASICs to their entire economic lifecycle to maximize free cash flow and profit on our Bitcoin mining business from the investments we make. By the way, we also lead the sector with corporate G&A, and we have the numbers to back it up. Next slide. The overview of the global business. Currently, we're mining approximately 10 Bitcoin a day with 25 exahash capacity installed. This was fueled by transformative growth in Paraguay, where we grew our business from $100 million fiscal 2025 to currently $400 million ARR. Additionally, we have 5,000 GPUs between Canada and Sweden operating AI cloud, and that is currently at $20 million ARR as per our quarterly financials. The really exciting news is our target to grow to over $200 million ARR between both AI cloud and hyperscaler colo. Finally, we talked about our capital deployment strategy prioritizing ROIC. A double-click on our BTC pledge: we have great downside protection as well as upside. Almost 2,000 Bitcoin that we pledged — 1,234 Bitcoin were pledged at $87,000. So with Bitcoin today at approximately $97,000, there's over $12 million of upside in our currently pledged Bitcoin. Additionally, we pledged 758 Bitcoin at an average of about $115,000, which is about $14 million of downside protection from that component. So collectively, if you look at the downside protection of $14 million and the upside potential of $12 million, even with Bitcoin having corrected where it is, the value of our pledge has been accretive to the tune of about $25 million, which I think is phenomenal. It shows we're methodical and thoughtful. We've used our cash flow from operations, our HODL strategy and earlier this year some of the ATM to scale the business to get to that $400 million ARR, and we have some very exciting things in the pipeline for the AI and HPC business with Buzz. Let's hop to the next slide. So the dual-engine strategy of growth: cash flow from the Bitcoin mining business helps us to scale Buzz HPC, where we have 3 AI data center expansions going to Tier 3 plus. At a glance, the Bitcoin mining business: 25 exahash, a blended fleet efficiency of 17.5 joules a terahash. We're at $400 million ARR, approximately 50% operating margin after electrical costs and it's a footprint of 440 megawatts of green energy. The exciting news we recently announced is an additional 100 megawatts that ANDE and the government of Paraguay has approved for HIVE. We're excited about that. That gives us the potential to get to a capacity of 35 exahash for the end of next year, and that would bring our fleet efficiency down to 15 joules a terahash if we were to upgrade with the latest generation machines. We only buy ASICs when they satisfy our ROI targets. We're focused on building infrastructure. The incremental cost to bring that 100 megawatts on is only $250,000 a megawatt. Recall that the all-in purchase price for the site and even to build our Valenzuela site is $400,000 a megawatt all-in. That's land, substation and buildings. The good news is that all the civil work is already done for the additional 100 megawatts at Yguazú. So all we do is drop in the transformers for the substation and then the hydro mining infrastructure, $250,000 a megawatt, very incremental and attractive cost to bring online capacity. With that capacity, we have optionality. We can either install next-generation ASICs — we always shop for ASICs at about a 1-year ROI — or we can look at other strategies to monetize that 100 megawatts of green energy. We have ordered the long lead items, namely the substation, which are expected to arrive in Q1 2026. Standby for updates. If we were to do 35 exahash, the current mining economics, that will be about a $550 million ARR by end of Q4 2026. Moreover, Buzz has a very exciting year ahead. We've recently announced an additional 6,000 Blackwell GPUs, B200s for H1 and B300s for H2 will come online in 3 data centers, bringing our current 5,000 GPU cloud to a total of 11,000. We'll double the size of our GPU AI cloud by end of 2026. With the conversion of New Brunswick, we bought additional land so that campus is now over 30 acres and is currently an operating Tier 1 70-megawatt data center. We'll convert that to Tier 3 plus and that will provide hyperscaler colo. Add that up, it gets us to a target of $225 million for the end of 2026 for Buzz. Collectively, our target for the end of 2026 can be $750 million for both the dual-engine strategies, of course prioritizing ROIC. On Bitcoin mining, we will scale if ASIC investments allow for attractive ROIC. On Buzz HPC, we'll be bringing online these GPUs in our partnership with Dell for the first phase of our Bell partnership. Next. At a glance, the global footprint for Buzz (just HPC): we currently have approximately 5,000 Hopper H100s and H200 series GPUs, about 4,200 A-Series GPUs in Sweden and Quebec. The additional 6,000 Blackwell GPUs will come on in 3 tranches: one through partnership with Bell in Manitoba, one through the Toronto data center we recently closed on, which will be converted and liquid-cooled, and our facility in Boden, Sweden, which will also be Tier 3 liquid-cooled. Toronto and Sweden are HIVE- and Buzz-owned and operated; we own the land and buildings and will convert those to liquid-cooled. Those will come online in Q4 of next year and each facilitate 2,000 NVIDIA Blackwell Series GPUs. This shows the math and footprint of scaling from 5,000 GPUs to 11,000 GPUs. With New Brunswick conversion to Tier 3 colo expected to deliver possibly 50 to 55 megawatts, market rates in that sector would be about $85 million of additional ARR. This gives a high-level glance of how we incrementally scale our $20 million ARR to over $200 million target for end of next year. Next slide. Beautiful photo of one of our H200 clusters operating. To clarify on GPU business: we're doing both GPU AI cloud and targeting hyperscaler colo in New Brunswick. If we focus on the GPU cloud business, the growth is $20 million going to $140 million, reflecting growth from 5,000 to 11,000 GPUs by bringing on 6,000 Blackwell GPUs in those 3 facilities. We have a mix of long-term contracts from 3 to 12 months and on-demand; we have over 10,000 unique users through partner networks. We're focusing on liquid-cooled Tier 3 data centers as NVIDIA racks get denser. Blackwell B300 clusters could be as dense as 180 kilowatts per rack, and rumors are higher for future generations. We're excited about the partnership with Bell. Next slide. This quarter we announced the Bell partnership; Bell is Canada's largest telecom provider. The team has executed operating agreements. The first cluster of B200s operating in Dell servers will arrive at the end of this year and go live January 2026 bringing incremental ARR and forming part of the sovereign data strategy, Canadian-owned meeting residency requirements. Buzz is a Canadian company with a great pipeline of data centers in Canada through Bell and the Toronto data center. Craig Tavares and Mario Sergi and our HPC team have advanced that initiative. Next slide. By the numbers on a timeline basis, this shows current $20 million ARR. As we bring clusters of B200s into the Bell partnership — 1,000 for Q1, another 1,000 by Q3 — each cluster of 1,000 GPUs on long-term rental basis. We're forecasting long-term 2-year contracts and using semi-analysis market rates on a dollar per GPU hour basis. Long-term 2-year contracts is about $20 million per cluster of 1,000 GPUs. So bring on 1,000 GPUs in Q1 gets us to $40 million ARR. Another 1,000 in Q2/Q3 gets us to $60 million ARR. Q4 next year is the big ramp because of conversions of Toronto and Boden — each will bring on 2,000 more Blackwell GPUs. By Q4, we target B300s. A cluster of 4,000 GPUs would add about $80 million ARR. That's how you get to $140 million ARR for AI cloud business with GPUs. Separately is hyperscaler colo with New Brunswick conversion from 70-megawatt Tier 1 into Tier 3 plus data center. That incremental revenue would be about $85 million based on $130 a kilowatt a month. This is the growth strategy for Buzz for 2026 with both AI cloud and hyperscaler colo. We have the land, power and expertise in running GPU clusters and we've been doing it for the last 2 years. Next slide. We're vertically integrating the tech stack. HIVE and Buzz have expertise in building and operating data centers and GPU clusters, and with Bell and sovereign initiatives in Canada, we intend to sell managed services to AI enterprises. This means custom endpoints allowing for agentic AI, inference, training and tuning. This platform we're very pleased to announce. Next slide. Buzz HPC cloud has launched and been pressure-tested; it's scalable and allows us to sell sovereign compute for clients in Canada. This opens the sovereign market in Canada — Canada is home to the Vector Institute and Geoffrey Hinton. Craig Tavares and the team are working with strategic partners in Canada with Bell, Dell and research institutes. We have exciting news in the pipeline. Moreover, this platform enables higher margin AI enterprise solutions. Next slide. The Buzz cloud platform has been ranked and awarded bronze by SemiAnalysis. That's the industry benchmark. On our first go-around with SemiAnalysis ClusterMax, we received bronze. Other industry peers are in underperforming category. We believe we're close to silver next year, which would put us up there with AWS. Craig and the team have done an amazing job. Next slide. Blue-sky potential integrating the cloud platform to sell managed services for AI enterprise clients would be a much higher margin business. Instead of renting GPUs for bare metal at long-term rates as the base case, the bull case is monetizing GPUs via the Buzz cloud platform for AI enterprise clients in Canada at a higher dollar per GPU hour. The base case used about $2.30 per GPU hour; the blue-sky case uses about $4.50 per GPU hour. At $4.50, you get about $40 million of ARR per cluster of 1,000 GPUs. So if clusters are monetized through the Buzz cloud platform, the current $20 million ARR could go to $60 million ARR and subsequently grow further. If you add hyperscaler colo conversion of New Brunswick, that brings you north of $300 million ARR target. The blue-sky and base case midpoints are where we expect to land. If our AI cloud business did $250 million ARR, it would be a multibillion-dollar valuation given margins. On monetizing GPUs, we have vendor financing, lease-to-own, so we are not required to raise capital to bring on the cluster of GPUs; $360 million for 6,000 Blackwells can be financed so we can be mindful about ROIC. CapEx is mainly to finish conversion of Toronto and Boden data centers. Those two facilitate growth of 6,000 GPUs. New Brunswick conversion to HPC would be more capital-intensive. Stay tuned. Next slide. Bitcoin mining side: we're mining approximately 10 Bitcoin per day even with difficulty at 152 trillion, currently operating a footprint of 440 megawatts of green energy. We have a proven track record of scaling. We've maintained best-in-class uptime and efficiency and lowest G&A per Bitcoin mine. Our Paraguay efforts are led by Gabriel Lamas and Carlos Torres. I'll be flying down there in a couple days for my quarterly site visit to walk every data center and inspect operations. We do this with disciplined capital allocation prioritizing ROIC. Next slide. Thanksgiving came early. We targeted hitting 25 exahash and complete the 300 megawatts by U.S. Thanksgiving and we hit it a couple of days ago. We put out a press release around November 11. Thanksgiving came early. We installed 25 exahash ahead of schedule. We've demonstrated expertise. We wanted to scale when accretive, and finding green energy is more challenging. Very happy the team delivered: shout out to Luke Rossy, our COO; and Bill Gray, our CTO. The whole team, Darcy, Gabe, everybody has done a phenomenal job. Next slide. Scenario analysis: Bitcoin corrected to just under $100,000. With Bitcoin at different prices you can see our current annualized mining margin is about $200 million with Bitcoin at $100,000, $300 million at $125,000 and almost $400 million at $150,000. This is mining margin after electrical cost. Regarding hyperscaler colo deals, Cipher announced a deal with AWS at $102 a kilowatt which equates to about $0.14 per kilowatt hour. With Bitcoin at around $102,000 this week, that's about $0.12 per kilowatt hour with our S21+ Hydros. You're not far off on a dollar per kilowatt hour basis compared to hyperscaler colo. Institutions like long-term stability and fixed prices over 12 to 15 years to underwrite. Bitcoin mining has upside because when hash price was $55, mining business was doing $0.15 per kilowatt hour, so even more top line. Our bottom line operating cost is about $0.05. That's why we can target 1 to 1.5-year ROI in Bitcoin mining whereas hyperscaler colo ROI is longer, around 8 years depending on CapEx. We do both. We're targeting HPC with AI cloud at about a 2.5-year ROI on GPUs and aware institutions prefer long stable cash flow of hyperscaler colo which is why New Brunswick is in the 2026 pipeline. Next slide. For the research analysts, hash rate and power online: Phase 3 in Yguazú, the additional 100 megawatts gets us to a global capacity of 540 megawatts; 440 megawatts today is operating with hydropower. We are not committed to immediate capital spend to bring on 10 exahash of ASICs. We will do so if economics and ROI satisfy our 1 to 1.5-year target after cost. There are other ways to monetize extra 100 megawatts. Stay tuned for exciting 2026 developments. Next slide. By the numbers for 35 exahash on an ARR basis: currently at $400 million ARR doing 50% margin after electrical cost ($200 million margin) for 25 exahash at 17.5 joules/TH. Scaling to 35 exahash with latest ASICs could reach $550 million ARR and around 60% margin ($320 million cash flow) based on current mining economics. Next slide. ROI principle: Bitcoin mining is an intersection of computer science and electrical engineering. We're building high-voltage, high-energy infrastructure. Revenue is combinatorial mathematics: number of hashes per block is difficulty times 2^32, Bitcoin price fluctuates, difficulty fluctuates. Over a 4-year horizon it trends down as more hash comes online. We always aspire to get 1-year ROI after electrical costs; once achieved, you're free cash flowing. Two levers: dollar per terahash ASIC acquisition price and dollar per kilowatt hour operating cost. Cheaper ASICs and lower power price extend economic life and free cash flow. We run ASICs for their entire economic lifecycle. Next slide. By the numbers: we lead the sector in cash return on invested capital. We realized 18% annualized ROIC this quarter. We aim to be beacons of truth in an industry with hubris. HIVE has been standing strong as the longest public crypto miner, navigated two Bitcoin halvings and an Ethereum merge; we've built an ASIC miner with Intel. We're numbers people and focus on ROIC. Next slide. We have the lowest G&A in the industry. We are a high-performance culture: 9 time zones, 5 languages. We brought 300 megawatts online in 6 months — almost an exahash per week; might be an industry record. Under 10% of total revenue is reflected as G&A and peers are in the 20% to 30% range. We scale with a lean team. Next slide. Relative valuation signals a strong opportunity. On an EV to exahash basis we should be about a $4 billion enterprise value. Many peers have AI and HPC strategies; so do we, and we think we have one of the best AI and HPC strategies with upside in 2026 as we bring online 6,000 Blackwells and pursue New Brunswick hyperscaler colo. We have the best uptime, lowest G&A per Bitcoin mine, best ROIC and right now the best value. Next slide. We invest in communities: we refreshed 18 schools in rural regions by our data centers. We refreshed primary schools with paint, air conditioners and playground equipment. We sponsor the hockey team in Boden and invest in communities according to local needs. Next slide. Embracing tradition: we were invited to the Presidential Palace for a YPO event with President Santiago Peña as host. We wore traditional scarves and had a great executive team turnout. Next slide. Paraguay is building ties globally, Peña has done a phenomenal job having relationships globally including with the U.S. We had strategy sessions with Peña in September and shortly after were granted an additional 100 megawatts. It's about strategic alignment with government and utilities. Next slide. Directionally, Latin America is emerging with HPC on the horizon; OpenAI and an Argentinian energy company discussed a Stargate for Latin America — Argentina is adjacent to Paraguay and buys power from Itaipu Dam. We believe there will be potential for HPC in Latin America. Next slide. With that, I'll hand the presentation to Darcy Daubaras, the longest-standing CFO in the Bitcoin mining industry, to give an overview of financial results. Darcy, over to you.
Thank you. Good morning, everyone. And thank you for joining us today. I will be walking through the Q2 results for the 3- and 6-month period ended September 30, 2025. We are providing certain non-GAAP measures in our presentation today. The company believes that these measures, while not a substitute for measures of performance prepared in accordance with U.S. GAAP, provide investors an improved ability to evaluate the underlying performance of HIVE. These measures do not have any standardized meaning prescribed under U.S. GAAP and therefore, may not be comparable to other issuers. Further details are found in the management discussion and analysis for the 3- and 6-month periods ended September 30, 2025. Moving on to the first slide. HIVE ended September 30, 2025 quarter with 236.8 million common shares, 2.7 million options, 11.4 million RSUs and 5.2 million warrants outstanding. On the next slide, let's start with the key highlights for the quarter. For Q2, we generated $87.3 million in revenue, driven by digital currency mining and high-performance computing services, and delivered $31.5 million in adjusted EBITDA. Production for the quarter was 719 Bitcoin equivalent, which is up from 406 in the prior period, supported by stable operations, strong uptime across our sites and the execution of our Paraguayan expansion. These numbers reflect disciplined cost management, a focus on efficiency and the benefit of our diverse global footprint. Now let's take a look at how this operational performance translates into our balance sheet strength. We take pride in maintaining a healthy balance sheet. As of September 30, we held $22.6 million in cash, $24.4 million in digital securities and $17 million in receivables and prepaids. That totals $136.7 million in digital assets against $55.5 million in current liabilities. We also maintained $25.7 million in strategic investments. This strong liquidity has allowed us to manage market cycles, invest in expansion opportunities, specifically our HPC and Paraguayan opportunities, and avoid over-leveraging the company. With that context, let's look at how our earning metrics have evolved. Shifting to our gross operating margin on a year-over-year basis: comparing this quarter to Q2 last year, our gross operating margin, calculated as total revenues minus direct operating and maintenance costs and high-performance computing service fees, increased to $42.4 million in the most recent quarter compared to $400,000 in Q2 of last year. In this most recently completed quarter, we are reporting a basic loss of $0.07 per share compared to a net income of nil per share reported for Q2 last year. Looking at revenue increases year-over-year, we generated total revenue in the second quarter of fiscal 2026 of $87.3 million versus $22.6 million in the previous year's second quarter. The revenues compared to the same quarter in fiscal 2025 can primarily be attributed to the expanded hash rate from the Paraguay expansion. Our gross mining margin, which is revenues minus direct operating and maintenance costs and high performance computing service fees, increased to $42.4 million or a 49% margin in the most recent quarter compared to $400,000 or only 2% margin in the prior year comparable quarter. That's a direct result of optimizing our mining fleet, the Paraguayan expansion, and improving operational efficiency. Zooming into the last two quarters, comparing current fiscal Q2 to prior Q1, we generated revenue of $87.3 million versus $45.6 million in the previous quarter. The increase was impacted by increased exahash capacity with the Paraguay expansion, an increase in the price of Bitcoin resulting in higher revenue and our continued focus on HPC revenues. Our gross operating margin increased to $42.4 million or a 49% operating margin in the most recent quarter compared to $15.8 million or 35% margin in the prior quarter. The increase was due to higher BTC pricing and improved fleet efficiency. I want to remind stakeholders that our net income is comprised of operational earnings or cash flow plus investment earnings, which includes realized and unrealized earnings and often noncash charges. Adjusted EBITDA in this quarter ended September 30, 2025 was $31.5 million versus adjusted EBITDA of $12 million in the September 30, 2024 period. I will highlight that adjusted EBITDA is a non-GAAP figure. For this completed quarter, we experienced a loss of $15.8 million compared to a net income of $35.0 million in the previous Q1 quarter. The net loss this quarter was driven by noncash losses on strategic investments and changes in the fair value of derivatives linked to Bitcoin used as deposits on equipment purchases. Quarter-over-quarter, adjusted EBITDA in Q2 fiscal 2026 was a profit of $31.5 million versus $44.6 million in the previous Q1 quarter. Q2 fiscal 2026 experienced net loss of $15.8 million compared to a net income of $35.0 million in the previous Q1. Q2 fiscal 2026 was a solid quarter for HIVE. We delivered strong revenue, expanded margins and maintained a robust balance sheet. Our operational discipline, fleet expansion and cost control measures position us well to compete and capture growth opportunities. As always, I want to thank our stakeholders and encourage them to continue to follow our expansion efforts in both Bitcoin mining and high-performance computing operations. There's exciting things to come, so stay tuned.
Thank you, Darcy. That concludes the presentation for today. We will now begin the question-and-answer portion of our call. Our first question comes from the line of Darren Aftahi from ROTH.
分析師問答
Congrats on the progress. Two if I may. Just can you talk about what the assumed spend on CapEx per megawatt is to either retrofit or greenfield your data centers to Tier 3? And then on the AI cloud strategy, I think, Aydin, you talked a little bit about financing and then leasing the GPUs. Do you guys plan to buy these at the end of that cycle? And then what's the assumption on useful life you're assuming on GPUs?
Yes, you bet, Darren. So the site in Boden is $25 million to build out to convert and Toronto is $40 million. We've had those facilities quoted and we're ready to commence the conversion. The price per megawatt is different because each facility has different existing infrastructure. Boden was already designed as a GPU facility; that's the one you saw on the site tour last year. Toronto is the facility we recently purchased. So again, that's $25 million and $40 million, respectively. That provides the footprint for 6,000 GPUs. Regarding the buyout: it's really lease-to-own. The buyout is de minimis, so at the end of the term we'll effectively have ownership of the GPUs for a very nominal amount. On useful life, broadly what we've seen is that after 3 years GPUs are worth about half of what you bought them for, conservatively. I think a useful life conservatively is 5 years. I back that up by noting we bought 38,000 Ampere GPUs (A40s, A6000s, A5000s, A4000s) announced in July 2021. Many of those GPUs (over 4,000 A-Series) are still running four years later. So that's why I put useful life at 5 years. It could be longer. We sold about 34,000 of those Ampere series GPUs last year at approximately 90% of their original face value, which was driven by global demand. So as a conservative guideline, 50% after 3 years and a useful economic life of 5-plus years is reasonable. In our past disclosures, the sales demonstrated strong residual value, but I'm giving a conservative estimate for modeling.
Next question coming from the line of Mike Grondahl from Northland.
My question on those 6,000 GPUs that you're putting out the next couple of quarters: how many of those are contracted today? Meaning what kind of visibility do you have on those? And then could you talk a little bit about the average term? At one point I thought I heard 3 months to 12 months and then another point 2 years — just the demand you're seeing.
Yes, Mike. The existing GPUs we have online today are in the 3- to 12-month range. We put out a press release this morning announcing the order of the first cluster of 500 Blackwell B200s, and that's the first cluster we expect to be delivered before Christmas this year. For the cadence of the 6,000 GPUs: the first 1,000 will be online by calendar Q1 2026, the second cluster of 1,000 by Q3 2026, and the additional 4,000 in Q4 2026. Q4 is driven by retrofit timing of Boden and Toronto facilities. The 6,000 GPUs are not all live yet; we're bringing them online in that cadence and will provide more color on contracted GPUs as they go live. Regarding contract term assumptions, the base case modeled longer-term commitments (up to 2-year contracts) because institutions favor stability; the existing on-demand and shorter-term work is 3 to 12 months. The realized ARR will likely land in the midpoint between the base case long-term contracts and the higher-margin cloud enterprise case. I hope that helps.
Next, we'll go to the line of Chris Brendler from Rosenblatt.
Congrats on the results, very impressive progress. I just want to ask a question on the Bitcoin mining business: the improvement in gross margins here — I know the network has gotten more difficult and challenging from a competition perspective — but your gross margins improved significantly quarter-over-quarter. With a full quarter run rate of Phase 3 complete now, I would expect gross margins to continue to improve. Just wanted to get your take on that, that your hash costs are going to continue to decline and your gross margins, all else being equal, should continue to improve now that you're fully up and running on Phase 3.
Yes, you've got it, Chris. The fiscal Q2 we reported had an average operational hash rate for the quarter of about 16.1 exahash and that included growth from July, August and September. We ended the period at a higher number as Phase 2/3 came online. Today we're at 25 exahash and we reached that milestone last week. We completed 300 megawatts, operational 24 exahash and installed 25 exahash, with some short-term seasonal curtailments in Canada. A few days ago we hit full 25 exahash operational as well. More hash rate coming online with our latest generation 15 joule per terahash S21+ Hydro machines will bring the average hash cost down and Paraguay's cheaper electricity further helps. Those two factors are contributing to improved gross mining margins. For modeling, the quarterly average operational hash rate for the quarter ended September was about 16.8 exahash; we ended the quarter much higher. Monthly production reports are helpful for analysts; we continue to provide those and encourage analysts to use them. We've mined every quarter profitably since 2021. We upgraded our fleet intentionally to mine with positive gross mining margin through to the next halving. We're monitoring macro conditions, but in the meantime we're mining profitably.
My follow-up: tremendous amount of growth over the last year and big plans for next year between HPC and Bitcoin mining. Phase 4 is on the table for Paraguay. How do you think about balancing the capital needs between HPC and Bitcoin mining? Is there a preference at this point or are you still looking to grow both at the same rate?
Good question. We have the additional 100 megawatts in Paraguay awarded to us. Our global operating capacity is 440 megawatts with 300 megawatts in Paraguay; the additional 100 megawatts is awarded at an existing site (Yguazú) which was originally designed as a 300-megawatt site. Incremental cost to bring that 100 megawatts online is roughly $250,000 per megawatt since civil work is done; it's attractive. The major capital item for Buzz is GPUs: a cluster of 1,000 Blackwells with InfiniBand is roughly $60 million, so 6,000 GPUs is about $360 million. But vendor financing and lease-to-own significantly reduces the need for raising equity or corporate debt to finance GPU procurement. That means the ramp for the first 2,000 GPUs can be achieved via vendor financing and the Bell colo arrangement, which requires no significant CapEx. The only CapEx we must finish is Boden and Toronto conversions at $25 million and $40 million respectively. New Brunswick conversion is more capital intensive and we'll provide updates. We are balancing capital deployment across accretive opportunities and prioritizing ROIC. Where vendor financing alleviates CapEx, we will use it; where we need to build infrastructure, we will evaluate accretive investment opportunities.
Time for a few more questions — two more. Next question comes from the line of Fedor Shabalin from B. Riley.
Thank you very much for putting out monthly updates. That's helpful. Could you please provide an update on the timeline for the New Brunswick data center, including whether construction will be phased or completed at once? Additionally, should the colocation agreement not proceed hypothetically, is there a path to fully utilize the capacity for AI cloud demand? How long would it take in that case, including timing for arranging financing for GPU procurement?
Fedor, the New Brunswick site is attractive given proximity to the border of Maine and its opportunity as a hyperscaler colo project. Institutions value hyperscaler colo for stability even though ROI is longer. Our $85 million target contemplated hyperscaler colo conversion of New Brunswick. Converting a 70-megawatt Tier 1 to Tier 3 plus is capital intensive and typically underwritten with long-term contracts. For GPU-driven AI cloud demand, if the colocation agreement did not proceed, there's still demand potential to utilize capacity for AI cloud; it would require additional conversions and procurement cadence similar to other conversions. For the 6,000 GPU plan, our immediate ramp is focused on Bell partnership, Toronto and Boden conversions, and vendor financing supports GPU procurement so that significant outside capital is not required for those phases. The New Brunswick conversion remains in our 2026 pipeline and we will update the market on sequencing. The projections we provided for HPC revenue assumed New Brunswick as hyperscaler colo in the base case; we will provide more detail on timing as engineering and commercial discussions progress.
Final question comes from the line of Joe Vafi from Canaccord.
Terrific progress. Great to see it. Can we drill down a little on Itaipu and Paraguay — clearly becoming a very strategic location for HIVE and very differentiated. You mentioned the Stargate initiative in Argentina. Could you drill down: how much power do you think is available to HIVE out of Itaipu? And what's the President of Paraguay thinking about AI and data centers as the industry develops?
Joe, great question. The Paraguayan leadership, including President Santiago Peña, are internationally educated and very business savvy. Peña studied at Columbia University and the ministers and leadership want to position the country for success; they're attuned to the data center industry. OpenAI's discussions in Argentina highlight regional interest. Paraguay has the Itaipu Dam — the largest hydro dam by production — and the question is how to monetize those resources. We grew our footprint from 300 to 400 megawatts in Paraguay, which means U.S. dollars paid in a timely manner for the government. We believe power pricing in the region will be attractive directionally. Itaipu operates binationally; there's a large capacity (about 14 gigawatts nameplate) and when we visited production varied by day — we observed around 10.7 gigawatts on a given day. The infrastructure's construction costs have been paid off and the question now is how pricing and allocation evolves post‑CapEx. We're working closely with government and utility partners, incubating a vision for data centers and HPC in Latin America with Paraguay as a catalyst. We're engaged in initiatives to improve grid infrastructure and contribute to national development. Stay tuned for updates as we advance these initiatives.
Excellent. Thank you, Joe. Thank you to all of our analysts and our attendees. We appreciate you joining us this morning. That concludes our Q&A session in our Q2 2026 earnings call. Thank you for joining. Thank you for your support. We look forward to speaking to you again soon.