On The Brink with Castle Island - Karim Helmy and Brandon Bailey (Galaxy Digital) on Standardizing Miner Accounting (EP.264)

Episode Date: November 29, 2021

Galaxy Digital Mining Associate Brandon Bailey and Research Associate Karim Helmy join the show to cover a new accounting methodology for Bitcoin miners. In this episode:  Galaxy's new miner margins accounting methodology Brandon and Karim origin stories in mining How the randomness-based miner fingerprinting works How Karim thinks about miner depreciation schedules Why the Bitcoin e-waste paper understates the expected lifetime of Bitcoin ASICs S9s are still a quarter of the Bitcoin network Why Karim and Brandon created the new accounting framework The differences between marginal, direct, and total cost of production for BTC miners Does the sell side capably cover Bitcoin miners? What you would expect to see for marginal and total cost of production for established Bitcoin miners Why miners focus on marginal cost rather than total in their storytelling Major bottlenecks preventing the addition of new hardware Why hashrate may not converge to price in the near term Why the chip shortage advantages incumbents in Bitcoin mining Why Kazakhstan is scaling back their Bitcoin mining Why mining manufacturers do not get priority access to the best foundry capacity Why established miners will be able to mine Bitcoin at favorable rates for the near and medium term Why of the accounting identities Galaxy would prioritize when evaluating public mining companies How Galaxy arrived at a 3 year depreciation period for the average mining unit What advice Brandon and Karim would give you public market analysts evaluating mining companies See the full write up and sample model here.  Sponsor notes:  This episode is brought to you by Withum, a top 25 accounting firm with a cutting-edge Digital Currency and Blockchain Technology practice. To learn more, visit  withum.com/crypto OurCrowd analyzes companies across the global private market, selecting those with the greatest growth potential, then brings them to you. Get started at OurCrowd.com/otb

Hello and welcome back to On the Brankwith Castle Island. I'm Nick Carter. This episode is brought to you by our crowd and with them. More on them later in the episode. So today I'm sitting down with Brandon Bailey and Kareem Helmi, both of Galaxy Digital. And in my view, they're two of the sharpest analysts and researchers in the mining space. They've written a piece called How Much Is It Cost to Mine a Bitcoin, which is critical of accounting methodologies used to evaluate mining companies. They attempt to to standardize some of the accounting treatment for miners. In particular, they develop three different ways to evaluate cost of production, marginal, direct, and total cost of production, which take into account different portions of the mining life cycle. This is a really important contribution to the

debate, and I think will really help standardize the way that people evaluate mining companies. If you are an analyst, active in this space you owe it to yourself to read this paper, which I will link in the show notes. Let's dive right into the episode. Brought down by bad mortgage investments, Lehman, which has 25,000 employees, will be liquidated. The federal government loans American International Group, AIG, $85 billion. This is a different kind of market, and the Fed is asleep. The federal government is stepping it to stabilize Fannie Mae and Freddie Mac, the two mortgage giants that have been threatened by the housing crisis.

The Bank of England has pumped 75 billion pounds more to Britain's ailing economy with a new round of Conchristit easing. You print a couple trillion dollars, and all of a sudden people's start to worry. So out of this worry, we have something called the Bitcoin. Bitcoin. Welcome back to On the Brink. I'm Nick Carter. This is the mining mini-series. We've done a lot of mining episodes. They're very popular. We have two of the experts here, honestly, and I don't say that lightly. I'm sitting here with Kreme Helmi, formerly Coin Metrics, now Galaxy Digital, Research Associate and Brandon Bailey.

mining associate at galaxy cream infamously known for inventing the metric which determined the prevalence of different types of hardware on the bitcoin network is that fair to say uh yeah well first of all thanks for having me on um i think i think uh just popularizing it and and coming up with a good implementation kram is being humble but uh yeah certainly an important metric that we now use to determine what kind of hardware is active on Bitcoin. And Brandon and Cream have written a really cool little paper and publish a model entitled, How Much Does It Cost to Mine a Bitcoin? Turns out this is the least trivial and most complicated question ever.

I see it in the press all the time. You know, X miners mining Bitcoin at, you know, $9,000 a coin. And, you know, if you, you know, if you. you raised your eyebrows at that, you're right because these metrics are not standardized. And, you know, this is the accounting podcast. So we're going to talk about accounting concepts today. So if you, both of you don't mind just introducing yourselves and, and telling us briefly about how you came to, to work in mining. Sure. So Brandon Bailey, the way that I got into mining. It honestly started early back in 2021. I ended up going down the mining rabbit hole by

buying my first machine off compass and just kind of yolode into it. And I really wanted to, you know, as I was doing that, I really wanted to kind of better understand like how to think about mining, how to, you know, kind of underwrite it per se. And at the time, a lot of the tools out there were more so just kind of these standard calculators that only show the profitability at a static point in time. And so I started taking it upon myself to kind of build my own spreadsheet, the kind of like underwrite my own mining operation, looking at it under various different kind of multiples of metrics. And as I was doing that work, just got, you know, more entrenched into mining, thinking about it.

I was working in real estate at the time and just saw how similar it was to real estate just from underwriting perspective, looking at cash flows and whatnot. So continue to just kind of get more into mining. I see what's happening with the China ban. I see how North America is being set up to really benefit from this in a large way. And seeing what was happening with equity and debt capital markets just felt like it was a great opportunity to try to work in this industry as it started to go through its boom phase and saw that Galaxy was hiring and looking to grow their team and reached out to Amanda Fabiano

who leads the team through a Twitter DM and center the model I had been working on. And the rest is kind of history from there. That's awesome. I love how oftentimes people get hired through Twitter. that's how I originally made my way into the industry. I sent a cold DM to Chris Berniske back in the day that kicked everything off. Cream, what about yourself?

Yeah, so I got into crypto, caught the bug while I was in college, saw actually one of your tweets promoting a job at Coin Metrics. Is that right? That is, yeah. You're not just saying that. you're not just flattering me by no i haven't bookmarked because um i was like uh you know it's a it's a historically significant moment for me uh i wanted for christmas this year i'd like you to print that out

and mail it to me i will i will and um i expect some fun dice in which i can do that but uh i yeah i saw that um i was like whoa this nigg guy is smart and he's saying this team is really smart. And I didn't realize you were pumping your bags, right? Oh, big time. All of my tweets are sponsored. But yeah, so I started talking to the team there. Started out as a contractor. One of the gigs that I was doing for coin metrics was around mining flows. So trying to figure out how much of the Bitcoin supply was held by miners and

how much was held by pools. I think there's going to be a recurring theme here, but it's not as trivial as it sounds. And yeah, then from there, started kind of looking at hardware fingerprinting a little bit as just like an intermediary thing that was on the way to doing better minor flows. Yeah, that kind of spun off into its whole other animal.

We managed to fingerprint the Ant Minor S9 just using publicly available data that's store on chain. And then, you know, finally built out the Bitcoin miner flows. While I was at Coin Metrics, I met I'm at both Alex Thorne, who's my current boss and Amanda, who runs the mining team at a Galaxy and join Galaxy about six months ago. I just love how the Fidelity Mafia has their tendrils everywhere now, with both Alex and Amanda being ex-Fidelity.

I feel like you guys are an honorary part of the mafia as well. You know, one of my greatest regrets was that we couldn't keep you at coin metrics, but that's okay. We had to let you, you know, spread your wings and blossom. You've been doing some great research galaxy. So, you know, I'm glad that we could chat. Just quickly on the minor hardware fingerprinting, just remind us how it works, because I think it's so interesting

that you have this sort of public data set of random, that tells you about what kind of hardware is active? Yeah. So the general idea is that the nonses that are stored in Bitcoin's blockchain, so the proof of work in proof of work consensus is supposed to roughly follow a uniform random distribution. In reality, most hardware doesn't actually sample uniformly because true randomness is a really difficult problem from both the hardware and a software perspective.

And, uh, or sorry, uh, high fidelity like pseudo randomness is a really difficult problem. Um, so what you can do is you can kind of like exploit the non uniformity in how some hardware samples, um, uh, non says to, to try to figure out like how much of that hardware is out there as a proportion of hash rate. Um, and, uh, we were able to do this most easily for the S9, because at one point, like the S-9 and its related hardware were so dominant on the network. There were about three quarters of the hash rate. And it left a very clear visible pattern on,

you know, like on chain effectively, if you kind of like plotted out the nonce values over time. And, you know, we're hoping to be able to do this with other types of hardware as well. It's a little trickier and you kind of have to adopt a different workflow. But we were able to just like pretty much using incidental evidence, just like release date and the thickness of the apparent like non-uniform patterns to kind of like fingerprint out the S-9. And it's related hardware that use the same chip.

It kind of reminds me of like the forensic accounting that like the IRS will do to like spot tax frauds. if you guys are familiar with that, because human beings are really bad at coming up with like random data. And if they look at your invoices and it's, you know, particularly uniform in the distribution or it's too random also, you know, that's sort of indicative that something is amiss. So if it's either too uniform or too random, that kind of looks like you're scrambling the numbers and coming up with them yourselves as opposed to that being sort of business derived numbers. So it sort of reminds me of that. And it's also relevant today's conversation because the nature of the hardware informs your depreciation period. Am I correct

in that assumption? It does. And that's actually one of the biggest things that we've kind of like seen in our new project that we'll be talking about. But it's it's just that like, assessing depreciation periods on hardware is is pretty difficult um for a few reasons the biggest is that you know like the industry is relatively young uh bitcoin just passed its 13th birthday um and uh so you know hardware has kind of like changed pretty rapidly in that time just from going from like CPU mining to GPU mining to a brief FPGA dominant period uh and now now you know A6, and then within the A6, the curve has just kind of, like, been pretty steep in terms of efficiency gains.

So there isn't really a good, like, just a, you know, a nice, like, line that you can regress along and see, like, how long it'll take for a machine to depreciate. Right. Yeah. And the S-9s, I think, have lasted longer than literally anyone would have thought. And so, yeah, now we're trying to gauge, like, how to assess how to mine a coin and and you know one of the biggest components of that that's always underestimated is depreciating compacts right and it's so relevant i don't know if you guys

read this divreys paper quote quote paper about the e-waste the bitcoin produces um it they used a one point i believe 1.29 year depreciation period for bitcoin mining hardware which you know as we all know is a crazy thing to do, at least in 2021, that's for sure. And they used it to drive this headline figure of, you know, Bitcoin transactions produced X many grams of e-waste, with the assumption being that the hardware is being trashed periodically. And so you can sort of, you know, construct a ratio where you assign that almost moral responsibility

for the production of e-waste to specific transactions. I know it's crazy. And they used a 1.29 year. I'll double check this depreciation period with the idea. They derived that figure by looking at one of these laws, like rights law,

I think it may have been for either of you are familiar with that, which has to do with, I believe, economic gains in computing hardware, you know, over time and some rate of increase in efficiency of computing hardware and putting transistors on a chip. So kind of related to Moore's law.

And they just sort of naively assumed that the hardware was improving at that rate. And they picked this extremely brief depreciation period. But as we've seen, as you say, Bitcoin miners are lasting almost longer and longer. And, you know, the S-9 came out in, what, 2016, and it's still north of, what, 20, 25% of the network today? Yeah, it's about a quarter of the hash rate, which is insane, if you think about it, right? That's unbelievable.

I think we're still going to have some S-9s kicking next year. It seems more than likely, at which point there'll be, you know, half the network, half the history of the network will have been the S-9 era. And sure, not the S-9 dominant era, but it's not trivial by any stretch. Yeah, and this affects everything. So it affects the accounting. No question about that.

The D in a bit da, sensor depreciation. Can't forget that one. It affects your estimates of the energy consumption of Bitcoin, which is so interesting because the different types of miners have different ratios of hash rate to electricity consumption. So you have to have a precise estimate there to derive your estimate of energy consumption or carbon emissions. And, you know, your e-waste calculus, if you want to do that analysis. And so it just matters. And it doesn't receive enough attention, my opinion. So we're giving it attention. So on to the sort of the main course. So tell us about the motivation

for this paper. I mean, I often see headlines in the mainstream press about miners, and it'll have some preposterous figure, like this miner is mining Bitcoin for $10,000 a coin. And that always felt incomplete to me because it's like, well, why wouldn't everyone then obviously do this and close the gap and compress the minor margins? And as it turns out, basically those figures are wrong, or at least incomplete. So tell us a little bit about the sort of the motivating factor here. Yeah, absolutely. I think one of the biggest goals that we hope to achieve with this paper is just kind of introducing a set of standards into the industry for how to calculate some of these different metrics and especially the cost to mine a coin, which as you mentioned is one of the most frequently talked

about metrics when it comes to minors. So what we really wanted to do is just kind of introduce the framework or formula for how to do this math so that we could hopefully kind of have an apples-to-apples way of comparing a lot of these publicly traded mining companies. And so what we proposed effectively were three different ways of doing this, and you can kind of think of it as three different tiers. So at the first tier or the first level, we call it the marginal cost of production. And so if you're looking at, if you're looking at this from a filings perspective, what this really includes is the cost of revenues or the cost of goods sold divided by the Bitcoin mined over that period. And what that's reflective of is sort of the cost of electricity for that miner. So I think

a lot of the figures you see where people are quoting, you know, it only costs us $10,000 to mine a Bitcoin. I think they're just purely talking about, you know, their cost of electricity. But additionally, what we have at the second level, which is what we call our direct cost of production, it includes that same cost of revenues, but the depreciation expense, because what you paid for the machines themselves is a real cost that we think should be taken into account when you're at least looking at this cost of mine a coin. So again, that kind of gives you the second layer of cost that get factored in. And then the very last one we have is just your total cost of production. So that includes your full expense load. So your cost of revenues or cost of goods sold plus your depreciation expense, plus any overhead.

So that would be like your selling general and administrative, what you pay your employees, any kind of consulting fees or legal fees, you know, that you end up paying for your normal course of business. So we take that total and then again divide it over the Bitcoin mind. to come up with that figure. So effectively what you have is three different tranches of expenses that you can now use to compare these miners, at least in apples to apples basis, and kind of see where they're either heavy or light as it relates to these specific tranches of cost. Gotcha. So we have marginal, direct, and total cost of production.

Yep. And, you know, have you seen the sell side actually cover Bitcoin mining. I mean, there are a number of publicly traded mining stocks in the U.S., in Canada, in Australia now in the UK. We see them. There's more going public. Every week, there's more public listings. So there's a vibrant ecosystem of Bitcoin miners. They're all kind of doing the exact same thing, so they should be easy to compare. Is there an ecosystem of sell siders covering this stuff? Yeah, there's, there's, there's definitely a few Wall Street investment firms that are covering the space. But, you know, we still haven't seen some of the, the bigger attention from, I'd say, banks like JP Morgan or Morgan,

Morgan Stanley, Goldman Sachs, et cetera, kind of come into the space. So while there is some Wall Street coverage, you know, it's still pretty small and narrow in view. But to that end, I think that, you know, there's just been an overall lack of kind of determining what are the appropriate metrics. And I think that there's also some education that still needs to be done on just mining economics and kind of understanding some of the nuances operationally with Bitcoin miners and how that might show up in some of their financials or at least their filings. And so we want to kind of leverage our expertise on the mining side and help to work in tandem with Wall Street and others out there to try to set the standard or at least

try to come up with some metrics that we think would be most relevant for for assessing the operational performance of these companies. Let's take a quick break to talk about our sponsors. Now, on this show, we only really like to work with sponsors that we actually know and can recommend. And both sponsors today are great examples of that. So our first sponsor for this episode is Witham. They're a top 25 accounting firm with a cutting edge digital technology and blockchain practice. So wherever your company is, from pre-seed to IPO, they have tailored solutions just for you. They've helped some of the largest companies in the crypto industry with audit, tax, and advisory needs. And they've also helped some of our portfolio companies, too.

To learn more about their advisory, audit, and tax services, go to on the brink. dot link slash with them. That's on the brink. Dotlink slash with them. All around the world, tech companies are innovating and driving returns for investors. Our crowd analyzes companies across the global private market, selecting those with the greatest growth potential and brings them to you. From personalized medicine to cybersecurity to robotics, quantum computing, and more, in state-of-the-art lab, startup garages, and anywhere in between, our crowd is identifying innovators so you can invest. when growth potential is greatest early. Our Crowd's accredited investors have already invested over $1 billion into growing tech companies,

and many of their members have benefited from the 46 IPOs or sale exits of their investments. Now you can truly diversify your portfolio by investing early in innovative private market companies at OurCrowd. Join the fastest growing venture capital investment community at OurCrowd.com slash OTB. That's our crowd.com slash OTB. So you guys also published a pretty handy spreadsheet here, which an analyst could use in theory, given certain known variables and then possibly public mining,

quarterly disclosures. Maybe what about just walking us through the model in terms of showing us the contrast between your marginal and your total cost of production, to show us the difference there. I think that would really help. People appreciate it. Sure.

So just kind of to paint the picture of what the spreadsheet looks like, we really tried to, I really tried to set it up in such a way that as far as inputting assumptions into the model, you'd be able to look at a 10Q, for example, and kind of see the same exact construct as far as line items go and the flow of things that. you might see and be able to plug that information directly into the spreadsheet. So once you've put in those inputs, it'll kind of auto-calculate the three different metrics we define, so marginal cost production direct in total. And I think what's most important to understand if you're going to compare companies in

the space where you might see some variance is specifically related to whether or not the company is hosting or if they own their own information. infrastructure. That's one nuance. If a company is hosting, there is a chance that their depreciation expense could be a little bit lower relative to a mining company that owns their own infrastructure because they don't have to depreciate the property plant and equipment effectively. That's something that could show up there. So, you know, that miner might have a higher direct cost of production relative to someone that is hosting. On the flip side, you might see a mining company that owns their own infrastructure with a lower marginal cost of production

because they have control over their electricity expense. It's all going to be dependent upon when they sign that hosting agreement. But if you're one of the newer mining companies that just recently entered into a hosting contract, you're going to see that their marginal cost is likely higher because hosting rates have certainly moved, have trended up higher, just given the margins in the space. And then lastly, I'd say total cost of production would give you some insight into how heavy they are for payroll. Are they, you know, overcompensating maybe their executive management team? You know, what does that, you know, type of setup look like?

And it'll provide some insight into that overall overhead structure. And so, you know, given today, you know, you're looking at, you know, a Bitcoin price of, I don't know, $58,000. you're looking at fees that are 3 to 5% of block reward. I don't know what the hash rate is, 130 X a hash, something like that. Given all those things, where would you expect to see, I know there's obviously variance here, but where would you expect to see the marginal and total cost of production for your kind of standard established large public Bitcoin miners?

What should we expect to see there? Yeah, I would say for the marginal cost of production, you should expect to see a range between about $12,000 and $15,000 on average. For total cost of production, I would say it should range between $30 and $35,000. So, you know, still they're showing strong margins, like no question about it. Absolutely. There's still sort of an arbitrage to close there in terms of hash rate converging to price. so to speak.

Absolutely. I think one of the other interesting things to point out about these metrics as well is that Bitcoin's price, you know, it really impacts your margin, right? So your marginal cost production margin. So, you know, what a discount effectively are you able to mine Bitcoin at? But in terms of impacting, you know, the actual marginal cost of production or direct cost of production, that's more so influenced by your overall, like a public mining company's share of the network cash rate.

So if they're losing, say, you know, if there are a percentage of the network cash rate shrinks, you could expect that their cost of production will go up, you know, when you're looking at that next quarterly filing. So it's not a static number. It's constantly evolving as the dynamics of the mining industry from a network perspective. are changing. And I'll add to that, it's up quite a bit over the last recent period just with hash rate soaring really, right? And this is something that we've kind of, to your point earlier, about like companies reporting like sub 10K cost to mine a coin. Part of that is that they were telling

the story of marginal cost rather than total. The other part is just like over the last month conditions have gotten quite a bit steeper as hash rate has come back online. And this all quite ties into the overall margins, where, you know, like, this is a physical process and people, people tend to neglect that. We've seen, like, quite a few constraints really act as bottlenecks over the course of the last year. First, there were, you know, like the legal constraints in China. Well, before that, even there was constraints around hardware coming online, brought on partially by COVID and partially just by like, look, our global supply chains have been over-engineered.

Then we saw, you know, restrictions in China knock off half the network. That's slowly built back up, but that was definitely slowed down by hosting constraints. And now we're kind of back to, you know, a soft combination of both hosting constraints and hardware. So, you know, I personally don't think we're going to see a total. convergence of hash rate and price and, you know, immediately at the very least because of the physical nature of this arb. Right, right. And that's actually, you know, I'm not really a big mining investor. Certainly, you know, that's not my mandate. But that's my largely uninformed outsider's hypothesis is that you expect this arbitrage to eventually close. You expect

hash rate to catch up to price because why should there be a perfectly competitive industry where there are grossly inflated margins? That seems like an aberration that seems like it shouldn't exist. That said, though, obtaining hosting capacity, the right voltage, electrical infrastructure, physical installation, obtaining the chips directly, and the political license to operate, those are all key bottlenecks, right? Absolutely. Of those, and correct me if there's any, I'm missing, which would you say are the most significant constraints

that sort of retard hash rate growth to the level, you know, to the level that quote unquote should be? I would say, I would say that shipping is definitely an issue. I think originally, you know, it was could you get the machines? It's since pivoted into whether or not you could get the racks space, finding a way to get your operation up and running. I think from what we're hearing, just the timelines for getting things like transformers and other important equipment that goes into building out a data center or hosting facility,

those timelines are extended and those costs are rising. It's also really important to think about how long it takes to build out. Typically it could be anywhere from eight to 12 months, assuming no construction delays. if you're setting up a new data center. And with supply chain issues, delays are increasing, I think all of those things are kind of playing a role into sort of the slightly delayed delivery schedule that we're seeing for some of the machines that I think some of the publicly traded mining companies have ordered.

That's causing, I think that's partly causing some of that bottleneck. Yeah, building stuff in America is, really expensive just at every stage of the process really and and the lower regulation jurisdictions that we've seen like Kazakhstan have actually started to scale back mining just kind of in line with having like other energy crises in the area so I think like we're going to definitely continue to see construction costs be a pretty large barrier it's affecting everything from like you know home builders to airlines to I guess now miners right so that's not going to go away anytime soon

but I think people do tend to underweight that hardware is still a limiting factor like ultimately there is still a threshold if all the profitable hardware were plugged in immediately it would still be profitable to mine bitcoin today at at the energy prices that most of these like efficient and public miners are still mining at. So part of it is just like this art can't go away until more hardware gets built. Right. And by hardware, you mean like actual electrical hosting infrastructure, like genuine facilities? Or do you mean Bitcoin mining units themselves? I mean the mining units.

So that's interesting. So the quote-unquote supply chain disruption is affecting the margins of Bitcoin miners, partly because just literally building things in America is hard and getting more expensive buildings. And two, and whatever specialized parts, you need to step up the voltage up and down. And two, because as far as I understand it, Bitcoin miners are not your like tier one clients at the foundries. And they're not necessarily getting prioritized access. to allocation at the foundry level. Is that correct? Yeah, that's true. The new bit main machines are five nanometer, you're sorry, use five nanometer chips. So I think like the non-preferential treatment

that they've gotten, or sorry, the reputation that manufacturers have traditionally had of being like very fickle customers, which has led to them being like generally deprioritized. is easing slightly. But there's still heavyweights who are, you know, like taken out a lot of the overall manufacturing capacity. And kind of like the most notable example of this is Apple recently booking like the entirety of TSM's 7 nanometer capacity. So you're seeing that.

You're seeing like auto manufacturers are also getting scaled back because they're also not tier one clients. So yeah, I mean, it's, it's, It is a global chip shortage and, you know, we're feeling the follow-on effects more than anything, but it is affecting every aspect of our increasingly digitized world. Yeah. I mean, as a complete aside here, I think shortage implies that there is like a decline in output in some sense, which I don't believe there actually has been. It's more that it's difficult to increase output to measure it with the increase in demand that we've seen for chips partly because everyone became extremely online

as a part of COVID and also not to go to my Bitcoin talking points here but we printed trillions of dollars and people use that to buy digital products and so that stimulus caused demand to increase So yeah, that's just me netpicking the people that, you know, claim that it's all supply side as opposed to, you know, demand side on the shortage. There's definitely elements of both. Also, cars have just gotten, like, for example, more complex and require more chips. But there are actually supply side shortages. One of them is just like, if I remember correctly, there's, you know, a shortage of,

sufficiently pure water in Taiwan to manufacture the wafers. And there's other components that are just like there's not enough of them to be able to max out plant capacity right now. So the punchline is that mining, and it's interesting that we're talking about supply shortages and scarcity as actually working to the benefit of the mining industry, because that's, you know, the ultimate outcome is that U.S.-based miners that are in market and establishes. will continue to mine Bitcoin at favorable rates for the foreseeable for the medium term. Absolutely. Yeah, there's definitely been a premium you could say for for hash rate that's that's already plugged in.

And as a result of these sort of supply chain issues are bottlenecks, I think it's creating, you know, this golden era golden kind of period of time for Bitcoin mining where it's just hyper profitable to be a participant in the space. Right. I think the analogy I would draw would be like in the gold mining space if like, you know, all of the whatever diggers, you know, somehow 50% of them were removed and junked. And then it became extremely difficult to buy more specialized mining equipment. The few miners that were active and in market would profit enormously from that. Absolutely. So returning to the accounting identities you've developed, I'm guessing that for you, the marginal cost of mining is not the most important one. Would you prioritize from an analytic perspective the total cost? Would that be the chosen variable for you? I would actually probably hone in a little bit more on the direct cost of production. I think the two big outputs or considerations that I think about are what, you're paying for electricity, so your cost of electricity, and then also, you know, what's your

capex look like? So how much are you actually paying for the machines to, you know, to mine with? So I'd probably hone in the most on the direct cost of production. When you look at the depreciation expense, effectively, it's kind of giving you a view into the historical capital expenditures of these public mining companies. And so for a certain sense, it kind of kind of can tell you on a relative basis, you know, who at what price didn't these companies acquire their machines at? And how does that compare to, you know, the peer set or the comp set? So a company that has really, really low direct cost of production, so low electricity expense, and they also have paid a pretty favorable rate for their machines, I would be

most keen on or interested in. And then another metric that we didn't really talk about in this paper, but I think is interesting because we're constantly looking at these things is kind of EBDA minus CAPX. And so that's more of a forward-looking kind of view into what they're paying for future machine orders. And it also gives you a view into how they're funding their growth. of how much money are they spending to pay for the growth of their operation? And so, you know, I think it's a basket of different kind of metrics that could be appropriate when you're looking at these companies. I think a lot of these metrics will also evolve over time, just as we're seeing the debt

and equity capital markets kind of emerge. I think looking at return on equity, return on assets, return on invested capital, those kind of metrics will prove to be really, really important to just demonstrate, you know, which of these companies are able to generate the most returns from, you know, the capital they've raised and from the machines they've purchased. And will Galaxy be performing this analysis for your cohort of large publicly traded miners or is it just a toolkit that you want to make available to the street to do, you know, to do this analysis? Yes. So for Galaxy, you know, for the mining team specifically, you know, we're really

trying to build this one-stop shop for financial services related to Bitcoin mining. And so for us, we're tracking all of these metrics. I think that there's a lot of insight that can be gained from tracking these things and analyzing them over time. And we're personally hoping to be able to use them to help our clients. And our clients are a lot of these publicly traded mining companies. So for us, you know, we have advisory and consulting services. We do minor financing. You know, there's a range of other opportunities and offerings that we have. So we look at this data collection and kind of insight to the financial performance of these companies is being able to try to offer them or help them, you know, achieve their goals financially. So that's

that's really the focus of it for us. There may be some things that we put out to the public, but ultimately we're doing a lot of this analysis internally. But you guys went to the effort of open sourcing your model here, which is very handy. Thank you for doing that. So was that just kind of a gesture of good faith? Or would you actually like these to become sort of standard metrics in the analyst community? Yeah, we would absolutely love to have these become standard metrics in the analyst community. I think in terms of thinking of these metrics, I think we would be more open to kind of open sourcing these things. Like I think this is more of a community thing.

I think setting standards should be an open source and community effort. It's more so if you want our analysis that we've already done the work for, that's a different thing. But we wanted to provide the tools for others to be able to do it on their own. So lastly, on the depreciation. period, you guys defaulted this model to a useful life of three years. And, you know, my understanding is that, well, so there's economic depreciation, then there's actual physical, sort of mechanical deterioration that these machines experience. I mean, they have only so many cycles that they can run before they start to fail. I don't know what that is based on the vintage. But I would imagine,

you know, a unit can be obsolete in either by physically breaking down or because hash rate grows such that it's just not competitive anymore and it's just no longer an efficient former hardware on the network. So telling me about this three years, is that just kind of the back of the envelope thing

or, you know, is that an assumption that you are willing to defend? Like, how do you think about the current timeline for new units being produced today? There's a saying among statisticians that all models are wrong and some models are useful. So in this case, look like we are not going to die on this hill of defending a three-year depreciation schedule. It is certainly wrong. And what's more to the extent that any depreciation schedule is correct, it shouldn't be linear. Because these machines tend to have like a long-tail life.

where they decay more slowly. Right. And this is something we've kind of seen in the S-9s, which are still trading, like, around $500. And, yeah, so look, like, we don't think it's right. But you need to pick a number. We think that this is relatively on the conservative side. So that's kind of advantageous

if you're trying to model the profitability of an operation. And it thankfully happens to be the median that's used by the public companies that we looked at. So, you know, like the median company won't have to change this one variable. But we've also seen quite a wide range, and it does very much impact the final result of the calculations. So we'd like to kind of just see some sort of standardization around, like, moving to something

that's defensible and normalizable. And we think, like, you know, three years is a good enough number, linear is an easy enough calculation. Right, but it's interesting because you say they have like an initial burst where they're cutting edge and they might dominate. And then there's a longer, you know, there's like a retirement community for miners almost. I remember seeing a talk from Rochiorall at Compute North talking about life cycle mining. And so, you know, we get into a lot of complexity here. But, you know, your S-9s, you might want to put them in a different type of energy.

asset than your S-19s, right? You might put your S-9s at a more interruptible energy asset, which has a cheaper all in cost because you don't have that uptime expectation, whereas your, you know, flagship S-19s or whatever the latest ones are from what's minor, you might put those in your 99 or 100% uptime, you know, high-integrity data center. And so, you know, because they are the cutting edge machines in the network, they are the most efficient, they're dominating the hash rate. So you start to have totally different characteristics based on the vintage of the machine. So you get even more complexity there.

Yeah. No, I mean, there's that. There's failure rates, which, you know, are kind of actually like dependent on the age of the machine and, you know, other things like just what type of environment. you're running it on how humid it is outside, how much actually the supply of energy is is interrupted, because turning these machines off and then back on again almost certainly shortens their useful life. Right. So there's a lot to consider there, but we wanted to make our model as simple as useful, or sorry,

as simple as possible while remaining useful, given that whatever we say is still going to ultimately be be wrong. So. And it's also, you just can't know the future and your depreciation timeline that you set today depends on the rate of advancement in chip technology in the future, which is impossible to know. Yeah. And then you can default to some of these laws, I guess. You guys have this nice chart in the paper. Yeah. And I'll say like roughly everybody I know in the mining industry was surprised by how big the leap was in Bitmain's most recent model in terms of efficiency. So these things are very, very far from certain. Efficiency is generally starting to plateau kind of the same way we've seen Moore's Law

start to break down, just because it used to be easy to make a gain because we were bad at making chips, and now we're, you know, as a group, slightly better at making chips, so it's harder to make big gains. But, yeah, the rate of growth is far from, certain rate of hash rate growth is far from certain, especially as more founders start to come online in the United States and elsewhere. So there's a lot of considerations there, and yeah, we're not fortune tellers. Yeah, and I mean, the hash rate growth is also just a function of political arbitrariness, you know, 50% of the network went offline suddenly. And now it looks like

some of it's coming back online in China, as far as I can tell. But now you have Kazakhstan. shutting down parts of their Bitcoin mining industry of U.S. states looking to ban mining. So it's interesting that so much of this comes down to political risk, which is basically unforecastable, I would say. I've had you guys on for 45 here. I would guess I would just, to both of you ask, given what you have found, and in particular, in terms of the construction of this model, what would you advise? to an analyst that's sort of covering the public Bitcoin mining companies would be the advice you give them. It's a good question.

I think the advice I would give is to, I think we need to look a little bit further beyond just EBITOM multiples or enterprise value to hash rate, not just keying in on, you know, target, call it 2022, you know, hash rate numbers. But spending a little bit more time focusing on the actual operation. So deeper dives into their cost of production, how much are they spending on these machines of quoted hash rate, how much of that has been committed, right? You could throw out a target, call it, whatever, eight X a hash, but if you only have 25% of that committed to date, what does that imply about the future capital raising efforts

of this company? Are they going to have to raise more equity, raise debt? I think just more kind of analysis and kind of conversations about the actual execution, you know, of achieving some of those targets and then just thinking a little bit more about, you know, how some of these shipping or bottlenecks are playing a part into their quoted delivery schedules for their machines. But overall, I would just invite analysts to, you know, spend more time kind of also thinking about metrics that would be specifically relevant to this industry. Cram, any last thoughts from yourself? Yeah, I like what Brandon said. I think it's also just, you know, all of these numbers, all valuation multiples are ultimately kind of made up. So I think these stocks are most useful when you compare them

against other ones. So I think if you can kind of get past the original chasm of people tend to have trouble saying like, yeah, look, all mining stocks are expensive. Well, everything's expensive, right? Just compare it to the next cheapest one. So that's my two cents. Awesome. Well, really appreciate the time. Two of the best and brightest voices in mining, honestly. Thank you for publishing this. This has been helpful for my own purpose. too and thanks for coming on today guys thank you neck thanks for having us