Bitcoin mining and electricity
Bitcoin mining consumes a large amount of electricity. Houlihan Lokey calculated when bitcoins are trading for $50,000 a coin, the electricity is effectively converted into bitcoins for $400 a megawatt hour.
Lately, bitcoin mining companies have been approaching wind, and to a lesser degree solar, companies with ideas for how to help them maintain revenue during periods when their power plants are curtailed, or ordered to reduce output, due to congestion on the electricity grid.
A panel of experts talked about these and other subjects at our 31st energy finance conference in South Carolina in mid-June. The following is an edited transcript.
The panelists are John Belizaire, CEO of Soluna Computing, Dave Perrill, CEO and co-founder of Compute North, PJ Lee, managing partner of Everstream Energy Capital Management and co-founder of Compute North, and Cole Muller, head of Cumulus Growth, a digital infrastructure platform that is owned by Talen Energy. The moderator is Noah Pollak with Norton Rose Fulbright in Washington.
MR. POLLAK: John Belizaire, many people may not understand how bitcoin is mined. Can you explain the process briefly?
MR. BELIZAIRE: Think of bitcoin mining as a security system. Bitcoin is a digital commodity that is global in use. You want to make sure that transactions in bitcoin are recorded on a blockchain so that they are clear and there is no funny business happening. This is done on a blockchain, typically a digital ledger.
Imagine each of you is sending money to each other, and there is no central party to manage the process. We are all keeping a list of the transactions. John sends Noah five bitcoins. Noah sends Dave three bitcoins. Everyone is writing down the transactions in the common ledger. The information is translated into a code in the ledger that lets everyone spot when the ledger has been tampered with.
The way we do this is through use of a magic machine. You put the ledger sheet into one side of the machine and out the other side comes a single number that encapsulates everything on the sheet. If you change one digit in the ledger, the number coming out of the machine will change, so there is no way to change the ledger without everyone noticing it.
The machine only works in one direction. If I give you the single number, there is no way for you to recreate what is on the ledger page. Take that number and stamp it basically on the top of the ledger page.
Every time a new page is added, it is stamped with a number. You end up with a stack of ledger pages. They are chained together. If you put a new sheet and the number from the previous sheet into the machine, you get a new number that you can stamp on the new ledger page. This essentially connects all of the ledger sheets together starting with the first page.
Each ledger sheet is a "block." The series of ledger pages is a "blockchain." The folder housing all of the ledger pages is the entire history of every transaction since the start of the bitcoin network.
The chaining of those special numbers is done by a process called bitcoin mining.
To turn each ledger page into a single number requires a tremendous amount of computing power. The magic machine is a cryptographic algorithm that was developed by the US National Security Agency called "SHA-256," and its purpose is to record and protect large volumes of information in an efficient way.
How do I encourage privately-owned machines to perform that work?
I keep changing the thing they have to do and make it very specific. Not only do they have to add a piece of information to the page to generate the number on the other side, but the number has to look a certain way. In order to do that, you technically have to keep making changes to the ledger sheet and putting the number through the machine until you get to the right number. Usually, it is a certain number of zeros in front of the special number and, over time as I do this, I find it is a searching algorithm.
All of the privately-owned computer servers are doing this number crunching over a long period of time. The amount of computing power required is massive. The computer servers require energy to run. Thus, bitcoin mining has a direct correlation to energy. Think of it as energy required to protect the network.
Bitcoin mining is a security system on the network. The relationship to energy introduces some opportunities. It can be a powerful catalyst for renewable energy because the computing by any single server or group of servers can be paused and then started again. This flexibility can be used to help integrate more renewables onto the grid.
MR. POLLAK: That's a great segue to the next question. Dave Perrill, talk about what your business has to do with bitcoin mining and how it overlaps with what the renewable energy developers and financiers in this room do.
MR. PERRILL: I have worked for 25 years with data centers and managed services and I co-founded Compute North in 2017 when my co-founder, PJ Lee, and I got really interested in bitcoin.
We believe bitcoin mining is leading the way to what we call TIER 0™ computing. Data centers are classified from tier one to tier four based on the amount of redundancy and, therefore, reliability. Tier four is the big boys: Amazon, Microsoft, Oracle, Google. They are mission critical and, therefore, they require built-in redundancies that make their data centers very expensive.
Bitcoin is at the other end of the spectrum. Our applications are compute intensive and, therefore, energy intensive, but they are not mission critical.
Another data application is machine learning. Examples are use of data to control self-driving cars, do genome sequencing or climate modeling. Anything that is compute intensive but can be interrupted, we think will move to a type of model where how energy is procured is important. Compute North is focused on how to marry the energy load to the energy generation.
MR. POLLAK: Cole Muller, describe how your business differs.
MR. MULLER: Talen Energy Corp owns about 13,000 megawatts of generating assets across the US. I run our growth platform called Cumulus Growth that is focused on digital assets that are being built adjacent to our power assets.
Our focus is on offering low-cost, reliable and zero-carbon power to run digital assets. Our first effort has been to connect our 1,000-MW nuclear facility to data centers. There are two 500-MW generating units. We have been focused on building a digital infrastructure campus next to each 500-MW unit, which really has two businesses. One is focused on the hyperscale data centers belonging to people like Facebook and Google as customers. The other business is the tier zero bitcoin mining facilities, as Dave called them.
Our thesis is that there are a couple value chains in the bitcoin mining market. The one that is of most interest to us is vertically integrating all of the pieces of that value chain. Owning everything from the power generation to the computer servers and getting to a "hash rate" that ensures a profit from the bitcoin mining. For us, it is about getting behind the meter, cutting out the middleman if you will, getting lower energy costs and protecting the business across the entire vertically-integrated value chain.
MR. POLLAK: PJ Lee, exactly what is a data center?
MR. LEE: A data center is a facility that hosts different types of server equipment. In many cases, the data is mission critical. For example, it streams Netflix or it is hosting email, or it is a website or it handles e-commerce transactions.
Data centers in a tier-four environment are built and engineered with redundancy in mind: redundant backup power, redundant fiber connectivity. They need to be up 99.999% of the time.
Increasingly new markets are emerging that are computationally heavy, but do not require the redundancy. This is what Dave called TIER 0™. Examples are bitcoin mining, artificial intelligence, machine learning, graphics and image rendering.
As this industry of digital infrastructure continues to evolve, one of the fascinating things that we are helping to accelerate is a hybrid model.
A data center may keep certain workloads on-premise that are important to keep in its own environment, but also rely on third-party providers for certain applications or accessing certain tools. That is already a hybrid model.
TIER 0™ is the third piece of the hybrid model where certain processor-heavy workloads could be outsourced in a non-mission critical fashion and instead of 99.999% of uptime, maybe uptime is merely 90% or 95%.
That is a perfect complement to the intermittency of renewables or to the need for grid stability or frequency regulation for the power grid. At the same time, in this configuration, each of these three different applications — the hyperscale or the TIER 0™ through software — could do something called "workload orchestration," where they could move different processor applications from one location to another location or to multiple locations.
As data centers continue to evolve with the strong desire to lead with ESG as their selling point to the customer base, there is a huge opportunity for renewable energy generators.
MR. POLLAK: John Belizaire, expand on what PJ just said. You do bitcoin mining. Where do you get your electricity?
MR. BELIZAIRE: We look for power plant owners that are having a hard time monetizing all of their energy.
We look for areas where the grid is highly congested and where lots of renewable power plants are coming online and their batteries are insufficient to address transmission concerns.
We partner with those power plant owners and essentially deliver to them a solution to their curtailment challenges. It is a purpose-built data center.
The facility we build is designed to convert the electrons that would otherwise be wasted into a powerful form of computing that is flexible. We do a curtailment assessment. We look for projects with curtailment challenges. We help the owners understand the effects of curtailment on their businesses. Most of them already understand it, but they do not understand how they might introduce a data center on site to address the concern.
Finally, we have a structure that can coexist with the existing project finance structure of the project and ensure that production tax credits are unlocked that otherwise would be lost.
The project owner gets a turnkey solution to its problem. We can do it with wind, solar and hydro facilities all over the world.
MR. POLLAK: Cole Muller, can you dive a little deeper into the problem you are trying to solve and the solution that is being offered?
MR. MULLER: We have 1,000 megawatts of digital infrastructure next to our zero-carbon nuclear facility. My job is to look at the rest of our fleet and figure out what is the next wave of digital computing and what opportunities it offers for another "energy and compute park." We take generating assets, decarbonize them to become baseload lower-carbon sources of generation, pair them with renewables that we are building adjacent to our existing facilities and then ultimately marry the generation to digital infrastructure and computing power.
We think this helps both with the energy transition and with enabling the grid to keep baseload resources on line for when those assets are needed.
For example, we own a bunch of coal plants in PJM. We make our money in the winter when the baseload power is needed, but two years ago, almost none of our coal plants ran during the winter. Folks started saying we no longer need coal. We can move now off fossil fuels rather than wait until 2030 or later to do so. Fast forward 18 to 24 months later and that is clearly not the case.
Our view is it will be a bumpy road as we go through the energy transition. Bitcoin mining and other digital computing applications will help us stairstep down and keep reliable generation on standby for times when it is needed by the grid. At the same time, we can incentivize renewables, battery storage and other kinds of clean energy to come on line more quickly.
MR. POLLAK: Dave Perrill, when you are looking to build a new project, what is your site selection process? What is your development process?
Mr. PERRILL: It comes back to energy first. The latest site that we are building now has a 300-MW load. Load factors are very high: 95+%.
We can site nearly anywhere, which helps to tackle grid congestion and stranded assets.
Last but not least, our operations are interruptible. We have the ability to shut down, and that downtime can be a relatively flexible number, roughly around 5%, but can be greater or less depending on the opportunities.
Beyond that, we start to look at what is the energy mix and how the economics work. We want to find a win-win model that works for all parties involved. We handle the development, design and construction of the data center. Our customers are the bitcoin miners. The electricity generator is our counterparty in a transaction to procure electricity.
MR. LEE: We can connect at the power plant busbar and negotiate a bilateral PPA or we can be grid connected and work with the utility to find a substation where we can interconnect.
We have a different business model than the vertically-integrated miners that typically build a large facility that is intended to remain in place for decades. Our approach is more modular where we mass produce outfitted containers in a contract manufacturing facility. Each 40-foot container is approximately a 2-MW load.
We think in terms of a portfolio where we start with a few different sites with one counterparty and then move some of those compute centers to other locations. The congestion points change with expansion of transmission capacity, so we are always trying to think two or three years ahead to where the puck is going.
MR. POLLAK: John Belizaire, what is your site selection process and how long is your build time for a project?
MR. BELIZAIRE: We try to target nodes where we know there is lots of congestion. We have an energy team that uses AI machine learning to analyze billions of bits of data with lots of different factors and then zooms in on places where we can invest.
We identify potential partners. Sometimes when partners approach us for a curtailment assessment, we tell them on which of their projects we would like to focus because we already identified where the challenges might be. Then we look at the location of the facility, the current leasing structure on the site, the substation and the different counterparties that are going to need to be involved. We have different structures for areas served by coops versus investor-owned utilities. We have strategic relationships with some of the larger transmission service providers that gives us some insight into particular locations.
We look at what has happened in the last four years at the node. Then we look forward in order to understand what will happen in that particular part of the grid. That helps us get a sense of the size of the facility we should build.
We also use a modular design for our data centers.
We started as a developer. We lived this pain ourselves in North Africa, building very large wind farms there and designing a vertically integrated solution. We used that experience to determine what the optimal size for a data facility is in a location. We will either sublease the site or partner with the power plant owner to connect us with the land owner so that we can negotiate a lease.
It usually takes six months to have a facility up once all of the main agreements are in place.
MR. POLLAK: Cole Muller, I want to bring this back to bitcoin and crypto as a general matter. All of your businesses to some degree either are dependent on, or derivative of, bitcoin, bitcoin mining, bitcoin revenues, bitcoin prices.
There has been a pretty significant drop in bitcoin prices so far this year from something like $70,000 per bitcoin to less than $20,000. This has led to headlines such as "The music has stopped for crypto."
How has this affected your future goals?
MR. MULLER: Rewind six months ago with bitcoin nearing $70,000 a coin and there were a ton of projects. Everybody was putting down deposits for miners and looking to expand, expand, expand. Now it is the exact opposite. Historically, bitcoin pricing has not stayed in one spot for very long. There is a lot of volatility. The key is to build a sustainable business that can last through the cycle.
It starts with electricity costs. Site selection is super important. Your business model also has to be able to withstand the inevitable down cycles, whether it is a 50% or 80% drop in bitcoin prices.
The trick is to match your revenue — the bitcoin or hash price that you are generating — to your power price and some of your other infrastructure costs. Power generators manage a spark spread. Ours is called a hash spread. You need to lock in a margin.
It is a fascinating space right now. There is a ton of opportunity. The natural inclination in a bitcoin down cycle is to pull back. It is cheaper just to buy the bitcoin right now, but I think developing sites and having an eye for opportunities is really how you are going to win.
MR. POLLAK: PJ Lee, is he right that it is cheaper to buy the bitcoin and, if so, why press ahead with your business model?
MR. LEE: We started five years ago when there was a lot more volatility than what we are experiencing today. Bitcoin prices fell from $20,000 to $3,000. We decided that just being a speculator on a commodity is not where we want to be. Entering into longer-term contracts was a way to reduce our cost of capital, introduce project finance leverage and have more stability.
We switched to a co-location service provider. We are effectively a landlord. We collect a co-location fee, or rent, from our customers every month and, if they pay in bitcoin, we convert it that day to cash or they pay in cash. We have never been focused on the crypto price. We are always focused on stable, contracted, recurring cash flow.
Anyone who is long bitcoin and short power is in a tough place today. That is a spread that you cannot really hedge long term.
There will be consolidation and some dislocation. Mining bitcoin is still profitable, even with depressed prices. The biggest things to get your head around as an investor or lender are whether this industry is here to stay and whether other industries will find other uses for the same type of infrastructure.
Our bet five years ago was "this industry is real." It has a credit profile that is a bit avant-garde, let's say, but bitcoin miners literally print money for a living. They have pretty decent profitability. The value of their equipment in any cycle is pretty high if you need to foreclose.
MR. BELIZAIRE: Let me add to that. There are few things the audience should keep in mind.
I did a professorial introduction to bitcoin mining earlier. There is another concept that we at Soluna call the "golden triangle." The bitcoin network is dynamically attuned to adjust the amount of work required to secure the network as the price begins to move up or down. As the price of the currency starts to drop, the amount of energy required to solve for the next increment of bitcoins also decreases.
Your LCOE, or levelized cost of energy, is also very important. Those who have very high power costs will tend to leave the network, and that leaves more of the pie for existing participants. That means the lowest cost energy allows you to stay in the network.
I have been a technologist for 20+ years. I am not an energy guy. I came to the industry about five years ago. This is a technology that looks very much like the internet did about 25 years ago. It has an incredible amount of potential.
What the technology does is very difficult to do, so the fact that it has continued to perform over the last 14 years is really fascinating.
That's why it fundamentally has value, and it will for the next decade, because of the fact that the technology can be leveraged to catalyze renewables and to do a host of other things. We remain very bullish because we are diversifying and doing lots of different things that essentially make this whole sector a new infrastructure sector, much like batteries or transmission are for the modern grid.
MR. POLLAK: Let's see how many audience questions we can fit in.
MR. HOULE: Tom Houle, CEO of Accelergen Energy. I have a two-part question. How are you addressing the avant-garde credit quality in your transactions? Are there alternative uses for the mining equipment or servers, assuming they don't become obsolete?
MR. LEE: Tackling your first question, our business model, as an intermediary or a developer of digital infrastructure, is to have a widely diversified customer base. We have more than 150 customers whose demand for capacity is quite significant. There was a lot of "spend to buy and deploy machines, get your place in line, get those machines quickly." I don't think this industry fully appreciated the complexity of developing large-scale infrastructure. It's hard permitting, negotiating contracts and arranging for interconnection.
There is a massive demand for capacity, such that if one counterparty defaults, there is a long line of other counterparties willing to take that seat.
Maybe that changes over time, but that is the situation right now. That diversity of customer base and the fact that the liquidation value of the equipment is orders of magnitude larger than any financing helps with the credit issue.
There are ways to enhance the credit by inserting lenders that provide equipment finance between the investor and the mining company. It takes a joint lien with the investor over the asset. In case of a default, the investor is looking at the counterparty risk of the financial institution, which has a much better balance sheet than the miner.
As far as whether the equipment will retain its value, the equipment is ASIC — application-specific integrated circuits — which are the workhorse of the bitcoin mining world, and GPUs, also known as graphic processing units.
The GPUs are fungible. There are plenty of other uses for GPUs. The ASIC today is really just designed for one thing: the proof-of-work algorithm for mining bitcoin.
As for the future, that is a whole other panel as to where this industry goes. We think of the SHA-256 algorithm as a sort of physical layer, similar to the internet 25 to 30 years ago. On top of that physical layer, more software application layers can be built. For example, the Lightning Network is a layer two protocol layered on top of the internet that facilitates faster transactions.
It is really early days for where things can go in an internet 3.0 context and what this infrastructure can be used for. It is not even the end of the first inning yet.