AI Is Not Hiking Your Electricity Bill. Your State's Energy Policy Is. And the AI Industry Has a Lot to Learn From Bitcoin Miners.
Lynne Kiesling at Knowledge Problem wrote one of the most important energy pieces this year. The thesis: blaming AI data centers for rising electricity bills is politically convenient but economically wrong. Retail electricity rates are set by state public utility commissions, not by the existence of large industrial customers. The infrastructure costs driving your bill higher were baked in long before a single AI server came online. Decades of deferred grid maintenance, transmission upgrades, wildfire hardening, and the structural costs of transitioning generation portfolios are what you're actually paying for. AI is a convenient scapegoat.
Kiesling makes the counterintuitive but correct point that large new industrial customers can actually lower average rates by spreading fixed costs across a bigger base. The question isn't whether data centers use electricity. Of course they do. The question is whether the regulatory framework allocates costs fairly and whether new generation gets built to meet new demand. That's a state policy question, not a technology question.
Here's the angle I keep coming back to: the AI industry has a lot to learn from what Bitcoin miners went through over the last decade. Miners learned the hard way that you can't just show up in a community, consume a massive amount of power, and expect a warm welcome. You have to be strategic about site selection. You have to build relationships with local communities, governments, and utilities. You have to position and help articulate and educate local communities, governments, and utilities that your operation will be a net positive for the community through tax revenue, grid stability, and potentially demand response. The miners who figured this out, particularly in ERCOT, are thriving. The ones who didn't are not as well positioned.
There is an important nuance, though. Bitcoin miners have the advantage of being relatively more location-agnostic. They can set up shop in rural areas with stranded energy and zero neighbors. AI data centers don't have that luxury. Latency matters for inference. Proximity to metropolitan fiber networks matters. So the community development challenge is actually harder for AI companies because they need to be closer to where people live. That makes the narrative even more important.
And on that front, the AI industry's messaging could not be worse. Leading with "AI is going to replace your job and we need all the electricity in your neighborhood to train the models that will replace you" is maybe the worst possible pitch to a community. Bitcoin miners learned to lead with energy abundance. The framing should be: increasing energy production and usage is tightly correlated with human flourishing and higher per capita GDP throughout history. Not only should we all want more of it, but if we look to our geopolitical counterparts across the world, particularly China, it is abundantly clear that we are far behind them in terms of a power generation landscape that is fit for our modern times as we accelerate into the digital age. These companies should be leading the charge for new power generation and telling communities, "We're going to build so much energy capacity that we can train all our models, run all the inference, and provide your city with abundant, cheap energy as a byproduct. We're going to bring tax revenue. We're going to bring jobs."
Here's the even bolder idea: we should be building dedicated power generation facilities, natural gas plants, nuclear plants, even coal plants, that don't have grid interconnects at all, built at massive scale specifically to run data centers. Both mining data centers and GPU data centers. These facilities become proving grounds for experimenting with next-generation power technology models. This is something I discussed on stage yesterday at the Texas Energy and Mining Summit with Gideon Powell from Cholla Energy, a company that's been building at the intersection of hydrocarbon exploration, Bitcoin mining, and data center infrastructure in the Permian Basin. The concept is straightforward: instead of fighting over existing grid capacity, just build new generation from scratch with compute as the anchor tenant.
This is not a new idea for us. We've been beating this drum at TFTC for years. Back in February 2024, Jamie McAvity from Cormint explained on the TFTC podcast that Bitcoin miners send the strongest possible signal to the generation community to invest and commit to new power generation resources. In March 2024, Pierre Rochard and I discussed on TFTC how regulators need to get out of the way and let the free market, driven by Bitcoin miners and AI compute, build more energy infrastructure so we can achieve energy abundance at scale. I've been saying for a long time that miners are already buying energy generation companies outright, building behind the meter, and vertically integrating the entire stack from transmission to transformers to data centers. This is already happening. But we need to create a real sense of urgency around power generation expansion. The question is: what can we do to incentivize more capital allocators, local governments, state governments, and the federal government to accelerate the construction of new power plants? Bitcoin mining is incredibly useful in these endeavors because you can spin up the generation and have a guaranteed buyer of that power from day one. It de-risks the entire project. Similarly with AI data centers, there is clear demand for compute and inference for the foreseeable future, so you can underwrite that demand on the generation side as well.
What's interesting now is watching the former ESG zealots in Silicon Valley come around to the same position. For years, these companies signaled their virtue by opposing fossil fuel development while their own data centers consumed enormous amounts of power from the grid. Now that AI demands more power than the existing grid can supply, suddenly they're interested in nuclear, natural gas, and building new generation. We're happy to have them on board the power generation expansion team. Seriously. But we need to work together to communicate this clearly to the public because there are a lot of misconceptions out there about what data centers actually do to electricity prices and community resources.
Honestly, nobody knows exactly what the other side of AI proliferation looks like. There will be some collateral damage and volatility as society adjusts. But the right posture is humility paired with conviction: "We should be leaning into it, not retreating from it. We'll figure it out together." That's what the best Bitcoin mining operations already figured out. The AI data center industry is about five years behind on community relations. They should be studying what worked in Bitcoin mining and adapting it.
|
