Earnings Labs

HIVE Digital Technologies Ltd. (HIVE)

Q4 2025 Earnings Call· Thu, Jun 26, 2025

$2.32

-5.51%

Key Takeaways · AI generated
AI summary not yet generated for this transcript. Generation in progress for older transcripts; check back soon, or browse the full transcript below.

Same-Day

+0.56%

1 Week

+29.05%

1 Month

+20.11%

vs S&P

+16.29%

Transcript

Nathan Fast

Management

Hello, everyone. Welcome to today's webcast on HIVE Digital Technologies financial results for the quarter and year ended March 31, 2025. My name is Nathan Fast, Director of Marketing and Branding at HIVE, and I'll 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 www.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 for a macro recap of the quarter. Frank?

Frank Edward Holmes

Management

Thank you, Nathan. So let's do a macro recap, starting the big picture, the transformative year, the most incredible year in HIVE's history. I -- when it first went public, become the first crypto mining company, and it's really quite impressive how the team is gelling. And let me walk you through why this is a transformative year for HIVE. Next. Next visual really is important for investors to understand the concept that they say that Bitcoiners buy the dip, stack the coins and hold on for dear life. Well, clearly, what you can see here is the daily volatility of HIVE is greater than Bitcoin. And MicroStrategy, now called Strategy, is double what Bitcoin is on a daily basis. And over 10 days, it's even more so. So it's really important when you look at HIVE, and we have lots of people like to try to trade like Bitcoin, et cetera; it's almost like you want to buy, if it's up 5% in a day, that's not the best day. But if it was down 5%, the math says it's time to buy. And that's really contrary to how most people think, except where I found that the Bitcoin ecosystem has the sort of concept that talks about it so often everywhere, everyday, Reddit, X, LinkedIn. Next, please. Well, I mentioned this is a transformative year. But to make it a transformative year, you need great leadership. And HIVE has expanded as senior leadership. And two of the key people that are helping scale our business is Craig Tavares. As Craig has a tremendous experience in Canada and building data centers, in particular, to run Buzz, be the President, which he is the Chief Operating Officer. And so we take a look in the past 6 months, the run…

Aydin Kilic

Management

All right. Thank you, Frank, for that intro. Let's have an executive overview of the year-to-date as well as some highlights from recent milestones. So it's been a stellar year for HIVE. The headline is $56 million of adjusted EBITDA. We're trading at a $350 million market cap. And by the way, we have over $60 million of Bitcoin on the balance sheet, 610 Bitcoin in the treasury $56 million of adjusted EBITDA on $115 million total revenue. So I think HIVE is an incredibly attractive buy, especially with a strong catalyst to growth this year. Our revenue was $105 million from Bitcoin mining and $10 million of HPC AI revenue. We did hit $20 million ARR in May. So this will continue to grow as we expand our HPC strategy, $25 million in gross operating margin. So it has been a strong year, considering we navigated the halving. This is -- our fiscal year is starting from March 31. So this year are effectively all post-halving economics. So I want to point that out. A lot of our peers had a December 31 year-end, ours is March 31. So this is effectively full post-halving mining economics. And by the way, 53% net cash and Bitcoin per share. We're trading at $1.75 today. And we've realized 22% ROIC in the last 12 months. So I think we've done a phenomenal job navigating post-halving economics, which are bearish, of course. And as the industry gets ready to torque up, I think it's going to be a phenomenal year for HIVE. Let's jump into it. So again, how do we realize these numbers? We prioritized ROIC in our capital deployment strategy. We've got prudent stable growth across cycles bear and both. We mined over 1,400 Bitcoin in this fiscal year. By the…

Darcy Daubaras

Management

Thank you, Aydin. Good morning, everyone, and thank you for joining us today. This reporting period marks a significant milestone for our company as we have transitioned our financial reporting framework from IFRS to U.S. GAAP. This change aligns with our strategic objectives, enhances comparability with U.S.-listed peers and supports our potential growth ambitions in U.S. capital markets. As this is our first earnings release under U.S. GAAP, I want to emphasize that while the core fundamentals of our business remain unchanged, certain financial metrics and disclosures may look different due to accounting presentation differences. We remain committed to transparency and will continue to provide clear, consistent reporting as we move forward. As this is the first time reporting under U.S. GAAP for HIVE, I want to highlight three key U.S. GAAP adjustments that will be seen in these newly presented statements compared to what we've had in other years. Those are functional currency, digital assets and leases. So functional currency change, HIVE changed its functional currency from Canadian dollars to U.S. dollars for the parent company. This was driven primarily by a shift in financing activities to U.S. capital markets. This change aligns with U.S. GAAP, which prioritizes financing indicators. The adjustment affects several areas, including warrant and convertible debt classification. One of the bigger ones is digital assets. Under U.S. GAAP, HIVE now records digital assets at fair value with gains and losses recognized directly in profit or loss. This contrast with IFRS where only losses were recorded in profit or loss, and gains flowed through other comprehensive income unless reversed, this will greatly assist the markets in comparability between ourselves and our peers. The third one is leases. While leases remain on the balance sheet under both standards, U.S. GAAP requires different income statement treatment. Rent expense…

Nathan Fast

Management

Are there any questions guys?

Bill Papanastasiou

Management

Yes. It's Bill Papanastasiou from KBW. Can you hear me?

Aydin Kilic

Management

Yes. I got you loud and clear.

Bill Papanastasiou

Management

Congrats on the sequential margin expansion and strong execution at Paraguay. I was just hoping to get some more color on the recent data center acquisition. Seems like a nice bolt-on addition to the portfolio. Any chance you happen to have a timeline of when you expect to get the liquid cooling installed and when the site will ultimately come online?

Aydin Kilic

Management

Yes. So that site, 7.2 megawatts would yield about 5.5 megawatts of compute, assuming 1.3 PUE and something like a 9- to 12-month retrofit with liquid cooling to bring that compute online. And it effectively would triple the megawatt footprint of HPC compute currently today. And yes, that's kind of the outlook for it. So it's very exciting.

Bill Papanastasiou

Management

Awesome. And apologies if I missed this. I can't really see the slide deck on my end, but was there any discussion in terms of the type of hardware that you guys are considering? I think there may have been some mention of NVIDIA Blackwell in the past. Just curious, how you're weighing, deploying kind of latest versus next-generation equipment?

Aydin Kilic

Management

Yes, exactly. That's a great question. So it will be a function of what -- yes, the slides were off, but it's a function of what will be available on the market at that point. So for example, today, you can order B200, but there's talk that in Q1 2026, you'll have B300 available. So it will sort of be an economic analysis because as you know, when the newest model comes out, we've heard that B200 fetching over $5 an hour right now. And so if there's a B300 that's out, typically, you'll want to gravitate towards getting the latest model GPU that's available. So that's why we haven't specified that. It's more of sort of an exercising capacity and analysis knowing, for example, a [ 1,000 ] cluster of Blackwell, so 1,028 GPUs, is a little over 1.5 megawatts, a little bit more power intensive than, say, an H200 cluster, which is a little under 1.5 megawatts. So yes, once -- as we sort of get closer to locking in the PO, then we'll provide more specificity on exactly which model GPU that we intend to operate.

Bill Papanastasiou

Management

And then just lastly, you're making all this progress on bitcoin mining gains, getting buybacks in [ ASH ] before the end of November and now scaling to the $100 million kind of annualized run rate on AI, HPC. Is there a particular revenue mix that the team is trying to achieve? And how could that differ going forward once you kind of execute on both of those businesses and scale?

Aydin Kilic

Management

That's a good question. I wouldn't call it so much of a revenue mix because we're positioning Buzz as a pure-play HPC company that will sort of be 100% HPC AI revenue and then HIVE moreover be -- I mean that's a wholly owned subsidiary of HIVE. And so how do we unlock shareholder value for that entity, we will evaluate. And of course, HIVE is operating Bitcoin mining data centers globally. So we're sort of bifurcating, I think it's perhaps a good way to think of it rather than looking at it as a mix. But as you could see from last year's financials, it was about 90-10. If we did get to next year and we hit $100 million on Buzz and $400-plus million on HIVE, like even at $55 cash price of 25 exahash, you're well over $400 million of ARR just on the Bitcoin mining. And so it's still in 80/20 mix if you did want to kind of have a ratio based on those projections.

Bill Papanastasiou

Management

Awesome. Appreciate all the color, and congrats on all the progress.

Aydin Kilic

Management

Yes. It's been an awesome year so far -- look forward to continued momentum.

Nathan Fast

Management

Thanks for those to questions, Bill. Our next question comes from the line of Mike Colonnese from H.C. Wainwright. Mike?

Michael Anthony Colonnese

Management

Congrats on all the progress in Paraguay and the recent hash rate ramp, great to see. Just following up on Bill's question around the 7.2-megawatt Tier 3 data center in Toronto, can you just share what the cost of acquisition was and what the related cost to do the full retrofit will be?

Aydin Kilic

Management

So we haven't -- good questions, Mike, we haven't publicly commented on that yet. So we'll be providing that color in the near future. So stay tuned.

Michael Anthony Colonnese

Management

Got it. Fair enough. I appreciate that. And as it relates to monetizing the 5.5 megawatt of critical IT load there once you're fully up and running, how should we think about that as it relates to customer demand? Are you looking at more on-demand type deals or contracted revenue streams? Just trying to think through the revenue of that additional $80 million that's projected to come online as it relates to timeline and underlying contracts.

Aydin Kilic

Management

Yes. So right now, I mean, we've got 5,000 GPUs to about 5,100 actually, 4,300 A-Series cards and 344 H100s and 504 H200s. And those are, for the most part, rented, I think it's about 80% utilization, and that includes some GPUs that are on longer-term contracts, for 6 months and longer, and some are on aggregators. So it's -- we're looking to -- with the sovereign strategy, as we sort of discussed that more in our most recent press release, we're really looking to embrace enterprises in Canada and working with the government as well as researchers to take on more long-term contracts, and that's sort of what we're alluding to with the sovereign strategy. As you know, there's a large AI grant in Canada. And so that's one of the initiatives. And so working -- we've got -- and by the way, we haven't opined on the price of the data center yet for competitive reasons, but we'll be providing more color on that at the right juncture. So yes, we've really sort of seen strong demand, and that was how we cross that $20 million threshold. And so if you think about it, with the, say -- it looks like a B300 cluster and say that's 1.8 megawatts for 1,028 GPUs, you effectively can fit 3,000 B300s in that cluster. And so we're actually renting out 5,000 GPUs now. So if you think of the quantum of GPUs, we're already ranking out a lot. The A-Series cards, of course, are much less power intensive than the Blackwell. So we're confident it's sort of within our ramp in footprint. You just kind of have more power-intensive cards that drive more revenue per unit. But in terms of the total quantity of cards, bringing on that additional 5.5 megawatts of IT load is actually less than what we have online in terms of number of GPUs. Is that helpful?

Michael Anthony Colonnese

Management

That's very helpful. I appreciate the detail. And just last one for me. Could you remind us of the cutoff point as to when you can repurchase those pledge Bitcoin with Bitmain?

Aydin Kilic

Management

Yes, it's to the end of this year, a little bit past the end of this year. We didn't specify who the vendor is with the pledge, though. That's only a clarification.

Nathan Fast

Management

Thank you, Mike. Next, we'll go to Tom [ Senske ] from Cantor. Thomas, please unmute.

Unidentified Analyst

Management

This is Tom on for Brett. First, I just wanted to touch on the regulatory situation in Paraguay. I know, as of June of last year, they put through a tariff against crypto mining companies. I wanted to see if there was any impact there and how the cost of power through your Paraguay sites compares to the facilities elsewhere.

Aydin Kilic

Management

Yes. So things are sort of in a holding pattern until 2027, but we're actively evaluating the landscape but also working closely with ANDE, which is a national power company. I was just there 2 weeks ago and met their CTO. We're going down again in September for actually a YPO event. The President of the country is in YPO's, and he's actually hosting a YPO event. So we sort of embrace the government and utility company and just the regulatory framework in Paraguay at different levels. We meet with governors and mayors when we go down there. And so I think there's an education process. Actually, I just met with the Ministry of Industry and Commerce when I was there a couple of weeks ago as well. And so I think as they understand the industry more, they we're optimistic they'll sort of embrace it more, and we have a long-term view that we expect with them embracing it to sort of have a more attractive -- power pricing trending in the right direction, if you will, is probably a good way to put it. And yes. So I would say that even with the citation of the June tariff, the power prices there are still more attractive than a lot of jurisdictions we've looked at. So yes, it's good. And I think that the revenue or so I should say, the gross mining margin reflecting the hash rate from Paraguay, stay tuned, it won't be too long because it's this current quarter, which will be reported in about 6 weeks. So yes, you'll see -- you'll sort of see the gross mining margins from this current quarter as we brought on those 5 exahash for period-end June, which I believe is reported in about, yes, 6 weeks from now.

Unidentified Analyst

Management

Awesome. That was very helpful. And then just one more, if I may, on this general CapEx throughout fiscal 2026. As we think about ramping to 25 exahash and also potential expansion within HIVE cloud business, how should we think about cadence of CapEx throughout '26? And how much of that should we attribute to the potential expansion of the cloud business versus getting to 25 exahash by Thanksgiving?

Aydin Kilic

Management

Yes, you bet. So the 25 exahash is fully funded. As you know, there was 610 Bitcoin on the balance sheet, as of our May production report, right, so that's sort of more current than the fiscal year-end. So that's approximately a little over $60 million. And so yes, we've effectively fully paid for all the ASICs that have been arriving, and that will continue to arrive through to 18 exahash. And so that's why we have shipments arriving weekly and deposits are in place. You sort of deploy your capital and you make those final payments for ASICs right before they ship, right? So the summer inventory that's all fully paid for and that's shipping. And so final deposits for the fall inventory, right, the third phase going from 18 to 25 exahash, those will be made in the fall. So it's all fairly near term. And I believe on the -- and that's on the ASIC side of things. And again, we're using the strategic upside strategy for that. And then on the construction, I believe, it was discussed that we actually had a sort of VTB arrangement, where $31 million of the $55 million purchase price was actually spread out over 6 months with $5 million monthly payments, 5 and change, right, 31 divided by 6. And so those continue from, I believe, April when we close the deal into about October. So that's nice and spread out. So yes, that addresses the Bitcoin. And as for the cloud we -- as mentioned, we haven't discussed the purchase price yet, although it was very attractive, I can -- I believe I can say that confidently. And the -- as with any construction project, right, you'll have -- you'll preorder some equipment and then the rest of the CapEx will sort of be spread out over that 9-month term. So we can provide more color on that. But really, as with the GPU business, you don't actually need to pay for the GPUs until you receive them on like crypto mining. And in fact, we have net 30 with -- I think it's fairly well known, we work with Dell and Super Micro as preferred OEMs. And we have net 30 payment terms. So if the GPUs don't arrive in 9 months, we don't need to pay for them until 10 months, right? So the CapEx on the HPC specifically to Toronto data center, I would say, is sort of further down the road? And in the interim, we're fully funded for the 25 with ASIC effectively arriving weekly between now and the fall with 25 exahash on the horizon.

Nathan Fast

Management

Next, we'll take a question from Fedor from B. Riley. Fedor, the floor is yours.

Unidentified Analyst

Management

Yes. Frank and the team. First of all, congratulations, strong progress you're making on both HPC and Bitcoin mine in France. And going back to your slide with Robert De Niro, Frank, I'm hoping institutional crypto adoption continues at the same pace we saw with catholicism spreading for Latin America. But anyway, I wanted to dive deep a little bit kind of follow-up question on Toronto HPC data center. Stepping back to bigger picture, it would be helpful if you could share us scale in the road map for this data center particularly? Kind of what does growth look like from here? And how you're thinking about expansion of composition of your fleet at this location? And also, how should we think about financing in terms of preferred options, maybe equity or debt split for the overall HPC segment?

Aydin Kilic

Management

Yes. So the HPC can be financed in different ways. For example, there's -- as you mentioned, there's debt some -- we've not done this, but just sort of providing color on how others in the industry have approached it, sometimes you can do not quite a vendor financing, but like it's no secret, Dell has DFS, which is Dell Financial Services. So they'll actually sort of lease you the hardware, the GPUs, right, which are the most CapEx intensive. And by the way, like just rough numbers, GPUs, latest-generation GPUs broadly are roughly about $30 million a megawatt, right, for HPC. And so there are different mechanisms to -- I mean you can buy them outright in cash, which is what we've done in the past for the GPUs that we have. And then as far as the data center construction financing, again, we sort of haven't disclosed the exact budget acquisition cost yet. It will be attractive and competitive, which is why we love this acquisition. And we're really excited about bringing it to the market. And as I mentioned earlier, Fedor, I think part of your question was scale or capacity. So 1.3 PUE, it's effectively 5.5 megawatts of IT load, which effectively, I think it's on one of the slides in my section, the exact page number. I don't recall. But it's effectively 3.5x of what our current capacity is, right? So that's really exciting. And then I believe I was addressing Mike's question just in terms of capacity of number of GPUs. So if you have like an H -- well, I mean, people would be buying H200 instead of H100 now, but that's -- those are a 100 cluster, again, 1,028 GPUs or 128 nodes with the [ HPs ] per node. And it's actually N minus 1. So you always have to have 1 node for switches. So you'd have 127x -- anywhere, just over 1,000 GPUs. That's about 1.3 megawatts for H-Series. And Blackwell B300 is expected to be a little more power-intensive, closer to 1.8. So you just sort of take your IT load -- so you always think your sort of gross utility load, right, at 7.2 megawatts, divide that by 1.3. And then if you want to say, well, how many Blackwells could I fit in there, you sort of divide by, conservatively, 1.8 megawatts. And again, it depends as can be B200, B300. So that all matters as well and -- or if it's going to be in a H200 cluster, then divide by 1.3 megawatts and then multiply you GPUs by the prevailing market rate on what they're renting per hour, like H200s are renting at $3 an hour right now on demand. So yes, I hope that was helpful.

Unidentified Analyst

Management

This is super helpful. And my second is about your customer profile and contracted dynamics. Can you kind of paint a picture of what your average HPC client looks like primarily serving on-demand, smaller scale customers? Or do you have a mix that includes some large enterprise clients as well? And kind of what's the typical duration you've seen on your HPC agreements, at least mostly short- term flexible arrangements? Or are you starting to lock in longer-term commitments with scale in Toronto? And kind of looking forward, how are you thinking about the evolution of your customer base? Is it fair to assume you're actively pursuing larger capacity contracts and trying to move up market to bigger enterprise clients, who can commit to, say, more sustainable and longer-duration agreements?

Aydin Kilic

Management

Yes. Generally, the answer is yes to the sort of numerous questions. That's sort of underlying by the sovereign strategy, and that's what we're getting out by working with governments and research institutes throughout Canada, viewing data really is a sovereign asset and working with those enterprises and research institutes that want to keep that compute with in the country. We actually have -- obviously, I can't say who, but there is a sort of a former Google deep mine guy that's training a foundational model on a 6-month contract right now on some of our GPUs. And so yes, we're absolutely targeting more longer-term contracts. We've done some compute with researcher out of NYU Stern, Colombia as well as Berkeley. And so they're all sort of doing very novel things. Some are -- I mean, in most of those specific cases, it's foundational model work. One, I believe, is actually a vision -- like a computer vision. So yes, it's a lot of really cool, fun stuff. And then, of course, the balance of the compute is with aggregators, some aggregators are really flourishing. And the -- it's also like having a -- like when you're working with the end users, like we've got a great team and so it's giving in the -- making sure the users have a good experience with the GPUs. It's not just the raw compute but making sure that they have a good experience as well. That's sort of another key thing because at the other end, is an actual user that's implementing and utilizing this compute. So it's just key that overall they're having a good user experience, and that's sort of a key takeaway if you talk to NVIDIA, that matters a lot to them as well. And so that's where we aim to please, and we're really excited about the direction that Buzz is going.

Nathan Fast

Management

Time for a few more questions. Dylan from ROTH, if you would kindly unmute yourself and proceed.

Unidentified Analyst

Management

Just to start on the Canadian data center, would there be any reason that those megawatts would come on in tiers? Or would you expect to energize all 7.2/5.5 critical IT load at once?

Aydin Kilic

Management

Yes, that would mostly be a function of the retrofit process Dylan. And so as I sort of commented on earlier, you'd kind of bring that online in roughly a 9-month process, and that would be -- you could do it all at once, absolutely because you're sort of retrofitting. And by the way, I should comment, one important thing is it's actually a working active data center today. But if we want to bring that to do liquid cooled, that's where the retrofit comes in.

Unidentified Analyst

Management

Got it. And just as a follow-up, where do you stand with some of your existing sites? Are you still considering potentially retrofitting any of those to HPC AI? Or are they going to stay Bitcoin mining for the near term?

Aydin Kilic

Management

Yes. Yes, our site in Sweden, we did a trip out there with some analysts, and I believe Darren saw it, we call it Little Boden. It's sort of -- there's 2 data centers that we have that are across the street for each other. And the smaller one, it's roughly 6 megawatts. We've identified that as a near-term candidate as well for HPC conversion. It was originally built as a GPU data center. So it's got a really good level of finish. And so yes, we've identified that one as well, and we'll be providing more color on that. So stay tuned.

Nathan Fast

Management

Thanks, Dylan. Our next question comes from the line of Mike from Northland.

Unidentified Analyst

Management

If you could just talk a little bit about the demand you're seeing for that 5.5 megawatts. Is it basically already spoken for? Or do you still kind of have to source the demand?

Aydin Kilic

Management

Yes. I mean right now, Mike, we just announced the data center. So we'll be providing updates on its conversion and bringing it to market. So we're doing -- as I sort of, in an earlier question, pointed out, it's roughly 3,000 Blackwell B300s if you were going to go that route, and we're already managing and renting out 5,100 GPUs. So although it effectively tripling our footprint when you just look at the number of GPUs because they're, again, more profit dense and power hungry as time marches forward. By the way, like the GB200 130 kilowatts per rack, and that's only 72 GPUs, right? So the GPUs directionally are getting more power intensive. So we're actually managing more GPUs today than the GPs that we would be bringing online. So we would provide guidance under that, but we're -- it's a great bit what we really like about this site, it's functioning today as a data center. And so it's operational. And it's a bite-sized deal that we're very confident that we could bring to market, and it will fit very well within our demand pipeline. And we are entertaining numerous -- like our H200s are overcommitted right now in terms of there's numerous groups or end users, if you will, that want to rent them with fixed contracts. And so that over demand, which is great, will tie in nicely to future capacity that we bring online.

Nathan Fast

Management

Mike, thanks again. Time for one final question. We will go to the line of Chris Brendler from Rosenblatt.

Chris Brendler

Management

And congrats on the results. I wanted to ask just a quick question since we're running a little long here on the mining gross margin, a nice sequential improvement I'm calculating here. Obviously, the Paraguay coming online, but it feels like that was only a partial quarter of impact. Can you give us any sort of sense on how much the mining margin can improve in the fiscal first quarter as you fully ramp the site and maybe a little bit of an outlook for the rest of -- or the beginning of, I guess, of fiscal '26? Because you've got a lot of low-cost power coming online, as more efficient machines as well, and we've also seen a recent improvement in the network hash rate. So any color there would be great.

Aydin Kilic

Management

Yes, totally. Great question, Chris. So actually, I did comment on this in a press release not too long ago as sort of in my quote section. And it's a really, really good question because it sort of highlights the improving fleet efficiency. So this quarter, right? Like again, I'm extremely proud of the team and I was still thrilled when I went down to Paraguay a couple of weeks ago to see the completed first 100 megawatts. And just everyone -- and by the way, like thanks, everyone, great show, like 60 people dialed in to this call. And I know all the analysts that follow the story so closely and kind of such deep insight. As you guys all know, we brought on 5 exahash 5 weeks in a row, right? It was extremely momentous. And so -- but to that end, all of that hash rate growth is in the last 5 weeks. So that's all in this quarter, right? So two things. First of all, the fleet efficiency, a big portion of that 5 exahash was S21+, air cooled units at 16.5 joules per terahash, right? And so our global efficiency sort of this current quarter is now 20 joules a terahash that will drop down to 18.4 joules per terahash when Phase 2 comes online and then 17.5 joules per terahash when Phase III is completed this fall. So you're sort of having this graduated improvement because the hydros that we're bringing online in Phase II, like Phase II is going to be all S21+ hydros. And by the way, like I think we put some really nice photos in the recent press release that commemorates 11 exahash, those containers are on site now, and they're being installed, which is super exciting. So that will…

Chris Brendler

Management

Looking forward to it. It's going to be interesting to see how much it improves in that fiscal first quarter. Thanks so much. Congrats again.

Aydin Kilic

Management

You bet. Thank you.

Nathan Fast

Operator

Thank you, Chris. That concludes our Q&A session and our Q4 and full year 2025 earnings call. Thank you, everyone, for joining. We look forward to speaking to you again soon.