I have been pretty quiet lately and I decided that I wanted to follow up on a discussion I had with Rich Miller at Data Center Knowledge last week about an article I was mentioned in that you can see HERE.
Rich and I could talk about Bitcoin and data centers until long after dark on any given day and yesterday was no different. The article he wrote was about the economics of Bitcoin mining as it relates to data centers. I will use this post to do a deeper dive than what he was able to use in the article.
The discussion (and article) was around how active the Bitcoin mining market was, and how few deals had been done and he was seeking an explanation. As someone who gets plugged into things quickly, in this case Bitcoin, and the almost 20 years in the business I guess I would do until a real expert came along. The nuances of the bitcoin market aren’t that different than the data center market of the mid to late 90’s and the subsequent rise of the .com era. Computing changed, and with it the hardware, how we used it, what it did and how we made it useful all morphed.
The Tier III data center design that has dominated the past 10 years for design, leasing, and use were built for a specific application – the shift from client-server and mainframe apps to the TCP/IP world. Facilities were using more and more electricity because the computers at the time – the 486 – were more powerful and needed more electricity. The form factors changed and the computer footprint went from a Volkswagen bus to a pizza box. The other dramatic shift, while fundamentally the same, was in the amount of heat these pizza boxes threw off.
Designs changed, pizza boxes grew more advanced, and the physical footprint shrank while the computing power went up in less space. The heat footprint changed with the power density (how tightly a footprint an electrical draw fits into) and so more heat in smaller spaces meant more cooling and so the cooling plants grew as a result. Still with me? More computing power = more heat to deal with. Incremental growth every few years as new hardware was developed and sold.
Until Bitcoin rigs came along and f*cked shit up.
I say that tongue in cheek because the rigs are changing a lot of things. They are making ASICs sexy again (still an arms race to the bottom), and they are making a 30 year old multi billion dollar industry think outside the box for a solution. The Tier III design has a lot of risk mitigation built into it. Uninterruptible Power Supplies (UPS) with heavy duty batteries placed in redundant arrays and configurations, those are backed up by generators also laid out in redundant fashion, and everything has to be deployed in pairs and in many cases overbuilt, just in case. Why? The technology inside the data center costs a lot and a lot of people depend on their livelihood for the technology to be available. It’ s cheap insurance (well not THAT cheap, but it beats $100,000 per minute downtime) in most cases and serves it’s purpose which is to get as close to zero downtime as possible.
Bitcoin, and cryptocurrency mining in general, is not mission critical today. If the rigs lose power, no biggie. When the power comes back, they start back up, go like hell until they don’t again. The software is self healing, so when power comes back, so does the mining. You lose a few bitcoin, no one dies or loses their job. Not yet anyway. So for the miners, street power, the more reliable the better, is fine.
Miners are also a frugal bunch. I am Scottish so I know frugal. ‘Good enough’ is great, cheap is better. So a few early adopters flocked to the Pacific Northwest where there is cheap hydro power and tried to find a warehouse that was on the same power system as the dam down the street. Simple. Although like many other things in life simple does not mean easy. The utilities spent an inordinate amount of time with miners who knew how to buy a rig and mine currency but didn’t understand utility scale. Utilities think in gigawatts, not 1100 amp power supplies. Big disconnect. So the warehouses are being scooped up and guys who aren’t proficient with utility scale have some figuring out to do on how to take 10MW of power and power a building – that is not a data center – like a data center. It’s simple but not easy.
So why not use any data center? There are thousands around the world, right?
Remember the Tier III build I took us through? Bitcoin miners don’t need the redundancy and all of the expensive equipment to go with it, and the data center operators already paid for and bolted the systems to the concrete so they are sunk costs. It’s like Nissan only making Infinity FX50’s for the people that want to drive Versa’s. The Tier III guys have a bunch of overbuilt facilities and even if they leased the space at less than cost, they still least the space for less than cost and drag the margin down across the board and Wall Street doesn’t like the build/buy high, sell low model. It costs a lot to change a factory over to build a different animal.
So we have a ton of demand, and no matching supply. In other words – Opportunity.
So what is a data center operator to do? Take it in the shorts? Some have. A deal is a deal in this day and age and revenue coming in offsets carry costs of an overbuilt Tier III building. Here is the issue with that – a data center operator goes out and raises money against a plan that has the amount of the investment (millions) and a certain rate of return. If the operator starts selling at rates way lower than the plan, and below the rate of return, then we have a problem. What is a really, really tough spot as an operator (I have been an owner/operator) is that here I have an exploding market that I can’t service because my cost of goods is too high on stuff I already built and can’t un-build.
There is a group out there raising $10M in convertible notes to service this need that the incumbents can’t touch. I think the investors who get in will do quite well. Why? It’s about supply and demand. Right now there is a ton of demand. I know of 151 MW right now, today being shopped for. The demand continues to grow. Current owner operators want the demand but when I tell them that the price point needs to be $50/kw, the silence is deafening. They can’t do deals at $50/kw for facilities that cost $75-2,000 a foot to construct.
So there is opportunity here. How much? Well 151 MW equals $7.5M in monthly rent at $50. $90M a year. That’s a lot of opportunity…