The Exchanges

Q Low, zero, or negative, then you can create enough of an incentive to get that wind built, and if that, uh, new source of load, in this case a bitcoin mine, is not soaking up 100% of that wind, the rest of that goes into the grid and still provides value in decarbonizing the grid overall. So I think that's the argument. Um, does that hold any water for you?

A Uh, no. Uh, so I think, let's break this apart into different pieces. Uh, so How are mines buying power today? So they are not signing new PPAs with new wind to as off takers to kind of make these plans happen. They're signing PPAs with retailers and utilities, um, mostly conventional types of deals, such as being put on base interruptible rates. So, you know, it provides demand response, but it's basically whatever the supply mix of the, of the utility or of the retailer, that is the power they're consuming. As we mentioned, sometimes they can sell power back. Um, none of the big mines have indicated any sort of offtakes with new renewables. I think there's some fundamental reasons for that. One is a sort of tenor mismatch, where, you know, most PPAs, you know, the tenor used to be well above 10 years, maybe up to 20 years. Today, 10 years can even be fairly long, but it's still, let's say, eight to 10 years is what, you know, a new project would look for. Uh, you know, Bitcoin mines are, I think, to put it lightly, a very volatile type of, in a very volatile business. And so signing very long-term agreements is maybe challenging, if not impossible, or at least costly to do.

AI assessment note: “Uh, no. Uh, so I think, let's break this apart into different pieces.”

Q the importance of the methodology that you use to figure out what impact, uh, Bitcoin mining load has on electricity, and the details of that methodology drive different outcomes. So let's talk a little bit about that. For example, how do we think about the relationship between both Bitcoin mines and electricity pricing, and also how we think about emissions impact and sort of the different ways to calculate that?

A Okay. We can kind of, I guess, take those in two parts. Um, so on the power price part, it, it, it certainly depends on what kind of market you're in. If you're in a sort of vertical utility market, there's one world. The one that I focused on in the study that, um, that Wood Mackenzie did for this story, this study, um, was in ERCOT, which is an organized power market. So as most people are aware, ERCOT is a, um, central clearing market. And so basically ERCOT is stacking up all the generation Offers, um, and then looking what load is, and then they basically dispatch least cost generation to match whatever demand is, and they need to do that subject to, you know, physical grid constraints and, and just the realities of the grid. So it's not literally a just pure kind of economic exercise. And so, broadly speaking, so the methodology leverages, um, what we have in Wood Mackenzie in-house. Our market intelligence team works with traders and asset owners in Across, uh, markets in the US, in particular in ERCOT, to every morning to, we have detailed grid models, um, that incorporate the topology of the grid, what congestion looks like, um, you know, what are all the assets and how they are dispatched. Um, and so using that and kind of load forecasts, weather forecasts, and a set of proprietary grid sensors for kind of real-time congestion are, you know, running sort of, um, Scena…

AI assessment note: “so on the power price part... the methodology leverages... detailed grid models”

Q gas would have otherwise been flared, um, and you can remove flare gas. So We'll take that one as kind of its own unique case, but let's, let's talk about those first two. So starting with bitcoins are a flexible, a form of flexible load that can participate in demand response programs and thus make the grid more reliable and support more renewable generation. How do you think about that?

A So yeah, I mean, uh, definitely mines are incredibly flexible in maybe an unprecedented way. Um, so there's no, there's absolute truth to miners' claims about that. Uh, most mines can drop up to close to a hundred percent of their load within seconds, if not minutes. Um, and I've seen this, uh, within data. They really do demonstrably drop quickly. So that's true. Um, mines are certainly acting as demand response very aggressively. Um, in ERCOT, they are participating as what are called load resources, which is a kind of class of participation model for large loads. Um, there's two types. There's controllable load resources or CLR, And there's non-controllable, which is NCLR. Most mines are NCLR, which basically means you can, they're treated by the grid as these blocky, um, sort of on-off resources, where, uh, if ERCOT dispatches them, or if the frequency dips below a certain point, they will automatically, uh, need to trip off immediately and kind of reduce their load. CLRs are more rare in that there are fewer mines doing it. It requires much more, uh, Um, granular dispatch. They need to be able to follow, uh, what's called a SCED dispatch. The, essentially what the, the market engine instructing them on a five minute basis. Um, so they are participating in demand response. It is incredibly lucrative. You know, if you were to participate in, um, the responsive reserve servic…

AI assessment note: “mines are incredibly flexible in maybe an unprecedented way. Um, so there's no, there's absolute truth”

Q Which is interesting because, um, If you just, well, I guess high level, most new generation is not coal and natural gas, right? But this is not new generation that is getting built, generally speaking, I guess is what they're finding. It's just ramping up existing generators, which tends to be coal and natural gas. Is that basically the logic?

A That's right. I mean, you have this, you know, the supply stack in any market is, you know, the first to get dispatched, there's wind because it has a PTC, solar, which historically has an ITC, Uh, increasingly we'll have a PTC and can big negatively. Um, but generally speaking, you always, the, the early loads on the system where a small amount of load will always be met by renewables. But as you start increasing upwards, you're always getting into the part of the supply curve served by coal, natural gas, combined cycle, open cycle, you know, oil plants, uh, oil, oil fire generation. And so at the end of the day, by you being there, You are causing those most expensive and those most dirty plants to be dispatched. So that is the effect of, you know, of when you're considering an intervention, that intervention, in this case, adding Bitcoin mining.

AI assessment note: “That's right. I mean, you have this, you know, the supply stack”

Q mines, what have we seen so far? Are they, to your point, basically they're just arbitraging electricity price versus Bitcoin price. Are they enough in the money at all times that basically they're just behaving like, Load for the most part. We'll talk more about their participation in demand response and stuff like that, but just like in a, in a baseline scenario, are they, are they base load load?

A They're, they're, they're close to it, but they're not, it's not 24 seven. Um, what I have seen, uh, surveying a number of mines is they're kind of, they're online roughly in the high seventies in percentage, you know, say 78% of the time, all the way up to the high nineties. So vast majority of the time they are consuming, but it's, it's not quite as regular as that. Like you are kind of flat You know, demand for a long period of time, and then maybe you're off for 10 hours, it is very spiky. So if you kind of follow these load curves, they're kind of on for a while, then they drop off precipitously, maybe off for an hour, or like, or a few minutes, and pop right back on. Um, and so you have this very irregular shape, but they are, the bottom line is that they are on a vast majority of the time, and some miners really almost all the time.

AI assessment note: “They're close to it, but they're not, it's not 24 seven.”

Q become, right? You said we have whatever it was, three point something gigawatts of Bitcoin mines operating in the U.S. today. Do we have a sense of how much is coming? Is it going to become a significant portion of overall load in the U.S. or portions of the U.S.? Has it slowed down since Bitcoin prices came somewhat down? Or, you know, how much should we care about this?

A So, I mean, it's certainly, it's never going to be comparable with the overall load of the, of the U.S., but it will Um, you know, it's, it's hard sometimes to separate the claims, uh, of the size of developments, um, from what will actually happen. I think roughly speaking, it's certainly set to double or triple in the next few years. I don't have a precise number, but it's certainly large enough that it's going to have an outsized impact. You know, the number that, that I gave the, uh, 1.8 billion per year in Texas, it's conservative because, you know, that's, Really only trying to count what is really clearly known to be operational. So I think it is a significant, um, you know, impact. We live, we are trying to achieve something with the energy transition that is a extreme reach goal. And, um, we are trying to reduce emissions in a very highly constrained economy, a very highly constrained grid. And so we need to basically be laser focused on every, um, Way that we spend our resources. And the question is, can we afford to throw, you know, to have many gigawatts of load that is essentially spinning wheels, kind of playing the lottery to win Bitcoin, which is what Bitcoin miners are doing. There's not only that, but it's, it's literally gets down to the, you know, the interconnection, the T and D kind of equipment that are used for miners. It's the, the interconnection that …

AI assessment note: “it's certainly set to double or triple in the next few years.”

Q That's interesting. So it's an interconnection-driven siting decision there. Basically, the reason you would want to turn back on a defunct existing site is predominantly that the grid infrastructure is ready for you to put something on the grid that, I mean, it's doing the opposite, right? You're saying it, it was a generation, now it's load, but at least the infrastructure is there?

A Yeah, that's right, because the infrastructure is a really key part. It's, you know, it also, not only is it costly to set up this kind of T&D infrastructure to support a massive energy consumer like mine, Um, but it also takes time to do it. You need to get permitted. So if you have everything kind of already set up, um, then it's a sort of an easy win for them to do it. And if you have a sort of very dependable generation source where you kind of know the cost, you, you know, you have this arbitrage that the miners have of, you know, knowing roughly what Bitcoin is going to be. And if you know what your supply is, you can say, this is going to pencil really nicely. I want to sit right here, kind of run my own generation and know that I'm going to make money a vast majority of the time.

AI assessment note: “Yeah, that's right, because the infrastructure is a really key part.”

Q load that otherwise raises prices and perhaps increases emissions, but it is super flexible. It is a great participant in demand response. And it can be highly price sensitive, and it's very flexible. Does that, I mean, I guess from the perspective of, uh, the need of the grid for load flexibility, which clearly we have, um, How strong a factor in favor of Bitcoin mining is that to you?

A So I would take that into kind of two parts, and I want to strongly challenge the price sensitivity part. Um, in terms of just the basic part of load flexibility and providing DR and, you know, isn't DR a good thing? Isn't flexibility a good thing? Um, I mean, the analogy I like to give, uh, is that it's like saying that crediting miners for demand response is like saying smoking is good for your health. Smoking a pack a day is good for your health. Because you are able to refrain from smoking on special occasions when asked. So obviously if I'm smoking a pack a day and I'm able in a family reunion not to smoke, that doesn't mean that smoking was good for my health to begin with. It just means in these very brief moments when I'm choosing, uh, based on someone else's request not to smoke, I'm not kind of doing my body harm. And I think that's really the, the way to, to view it. It's, this is an enormous load that has come onto the system and that we're basically paying it Which is creating all these kind of burdens that you just enumerated, and that you're then saying, I'm going to pay it to sometimes go off, and see, I'm willing to come off, so I'm helping people. So it's like, basically, uh, you know, the brief absence of a harm doesn't make something a good. That's how I view it. We could also get into kind of talking about system reliability and how they contribute to that,…

AI assessment note: “the brief absence of a harm doesn't make something a good.”

Q it's turned into CO two if it's flared. Um, so you, you know, it's not as bad as natural gas leaking into the atmosphere, but it's still CO two leaking into the atmosphere. So if instead you could take that gas and then, you know, use it to power a Bitcoin mine, you're avoiding the emission of CO two that would have occurred otherwise. How do you think about that?

A I mean, I, I agree with how you characterized it. Um, I, I think, I don't know enough to really be critical or to kind of, to give it an official kind of thumbs up, but, um, my understanding is that flaring is very kind of deeply imperfect in that you do a lot of methane leakage or more than you'd want, and that if you are properly capturing, uh, this flared gas, this gas, and you can run it through, you know, a sort of a gen set to, and to run, uh, a Bitcoin mine, you end up with more purely You know, um, combusted and therefore kind of CO two that comes out of it. And so therefore it's an improvement. It's, it's a reduction in the emissions of that gas that was, that had to come out because basically you don't have the, the natural gas transmission at these sites to economically to route it and to do anything useful with it. So it's gotta come out and, you know, assuming that the, uh, the well is operational and it seems like an improvement and therefore that it would be a net benefit.

AI assessment note: “I agree with how you characterized it... it seems like an improvement and therefore that it would be a net benefit.”

Q So, 3.9 gigawatts, I mean, let's, let's contextualize that for a second. That's a big number. It's measured in gigawatts, but I mean, as compared to, for example, broader data center world, still pretty small, uh, In the U S do you think of that as being like a meaningful number in the context of overall electricity dynamics in the U S?

A I mean, it is, if you think of kind of the scale of, uh, you know, the size of, let's say the, the competitive markets like ERCOT, um, and PJM, I mean, PJM is a large market, but in ERCOT, um, you know, several gigawatts, which is around what we think around two, um, is, you know, pretty sizable chunk of if ERCOT is on the order of That's something. So it partly has to do with where it's located and can have an, like an outside impact in one region. And, uh, you know, the mines themselves are pretty broad in terms of the range of capacity. They go down to around 10 megawatts, um, and then up to four 50, which is the largest. And, you know, you can compare that, you can compare it to the overall grid or to data centers, which is one way of doing it. You can also compare it to homes. And, um, If you, you know, a reasonable baseline of a home is around one and a half kw. So, you know, they range roughly from 10,000 homes to several 100,000 homes in, in, uh, in load.

AI assessment note: “I mean, it is, if you think of kind of the scale of”

Q That's super interesting. And is there a direct correlation between the price of electricity and their spiky load profile? Like, is it, is it clear from the data that when electricity prices spike, Bitcoin mines turn off and then they turn right back on when prices come down? Or is it more complicated than that?

A So I haven't really, I haven't done that analysis. I haven't sort of aligned it with the price of Bitcoin and, and to, to get that dynamic. Um, I, you know, honestly, I, a little bit Scratch my head looking at some of these load curves, because I, I don't really believe there could be rhyme or reason that price movements don't move intra-hour in this way. That would kind of explain this sort of the, the craziness of it. So I think there's probably a lot of factors going on. Some of them are probably, you know, you don't want to ascribe strategy when it's just, you know, could just be O&M style stuff going on or some other sort of issue that's not really intentional. Um, but, but one thing to highlight is that some of the power purchase agreements That miners have secured require a high load factor, meaning, you know, you need a high average power relative to your max power, which basically requires you to be on most of the time. And so I've seen one contract that requires above 85%. So in some cases, the, um, the energy supplier is requiring them as, as a term in the contract to be mostly on.

AI assessment note: “So I haven't really, I haven't done that analysis.”