We are moving from a model of fossil-fueled generation, meaning grid operators must find ways to modulate demand up and down, rather than just moving supply up and down
INNOVATION CIRCLE
Join us on social networks
Bitcoin mining has long been attacked by policymakers as an expensive, wasteful industry consuming power that would otherwise be going to households or valuable industries. Indeed, the scale of global Bitcoin mining is massive: Coin Metrics estimates that Bitcoin miners today consume 13.5GW of power, equivalent to 15% of the peak generation capacity of the Texas grid. But in recent years miners have adapted to changing grid conditions and found ways to make their presence much more benign — whether this involves exploiting entirely stranded sources of power, like flared gas, by co-locating with under-monetized renewables, or by participating in grid flexibilization initiatives.
Bitcoin miners are pioneers in this respect, in the future we can expect other types of energy-intense industries to follow their lead. In future years, I expect that Bitcoin miners will be looked on by environmentalists and policymakers not with scorn, but with grudging admiration. In time, it will be undeniable that miners helped develop a new type of industrial “smart load” that is more able to accommodate renewables, bring load to the generation source and turn down when necessary.
The green transition is changing the way electrical grids work. We are moving from a model of fossil-fueled generation which ramps up and down to accommodate load to one in which variable, intermittent renewables play a much larger role. This means a few things: grid operators must find ways to modulate demand up and down (to accommodate the unpredictable nature of wind and solar), rather than just moving supply up and down.
This is known as demand response, and environmentalists consider this a vital tool in architecting an energy transition — bringing about a world where energy consumers are able to respond in real time to changing grid conditions. The IEA has called for dramatically more demand response in order to meet Net Zero scenarios. This means that households are increasingly being asked to install smart thermostats which can strategically curtail when electricity is scarce. But no one wants to turn down their AC on a hot day — even better if industrial consumers of power are able to perform this service. And indeed, energy-intensive industries like aluminum smelting, steel plants, cement production, paper pulping and oil refinery do play this role. However, all these industrial processes cannot fully ramp down their usage, and they can’t curtail indefinitely or on very short notice. Bitcoin miners by contrast can ramp down fully on a moment’s notice and stay off indefinitely (since the process of mining Bitcoin is actually trillions of distinct mathematical operations each second — there is no “progress” in SHA-256).
For this reason, miners have found themselves very able to participate in demand response, and have begun building this into their strategies. As it turns out, in some markets, it’s more profitable for a miner to plan to be online 95% or 90% of the time and to turn off strategically when energy is scarce (and expensive). In fact, grids benefit from loads that can do this — that they will actually pay miners to turn themselves off — since that’s cheaper than paying a generation source to come online quickly. In Texas, during the recent summer heat in which the grid was taxed, Bitcoin miners went offline, freeing that power up for other uses.
Advertisement
Stay safe in Web3. Learn more about Web3 Antivirus →
A recent paper from the Energy Systems Integration Group shows exactly how this works. As the power grid experienced congestion during a few hot July days, the price of power spiked, and Bitcoin miners strategically went offline.
There have been concerns that Bitcoin miners might overwhelm the Texas grid, but that didn’t happen. Texas has been able to manage its influx with the large flexible loads task force. In practice, miners achieve a dual benefit by paying for more generation to be added to the grid (and in Texas, the vast majority of new power being brought online is wind, solar and battery storage) while turning down when needed. They are a model citizen as far as renewable grids are concerned.
By being location agnostic, miners are also able to exploit sources of power, especially renewable, that are otherwise underutilized. Ordinarily, fossil-fueled generation is built near population centers. However, wind and solar resources may not be near cities or industrial parks, so expensive transmission must be built to deliver that power to load centers. Miners can mine from anywhere, so they can go straight to the source. We see this with Crusoe Energy, which mines right on remote oil and gas wells, using excess methane. Only a few industries historically have been able to do this. In the past, aluminum smelters would locate themselves right on abundant power sources, like hydro in upstate New York. In some cases, Bitcoin miners have moved into these old facilities.
Newer industrial loads should follow this model, especially as modern grids incorporate more remote wind and solar (and transmission remains a bottleneck). Already, other power-intense industries like green hydrogen production, desalination and fertilizer production are mirroring the development of Bitcoin mining. Other industries will be challenged to follow Bitcoin’s model. Cloud computing, which is growing rapidly because of AI, is a candidate.
Currently, ordinary data centers are less able to interrupt themselves like Bitcoin data centers do, since they do suffer serious costs if they were forced to turn down on short notice. Cloud computing providers provide uptime and reliability guarantees to their clients, so they can’t tolerate an outage at the data center level. AI data centers doing model training probably won’t be able to tolerate downtime at all. But inference (the practice of interrogating an existing model) could potentially be made interruptible.
As AI continues to grow and consume more and more energy, policymakers and the press will ask the same questions of the AI sector. AI data centers should follow the example of Bitcoin miners: they should look to co-locate with renewables, bring their load right to the generation source, and figure out how to bake in unscheduled downtime into their operations. Their future might depend on it.
Leave feedback about this