Energy Is the Only Moat: Why the AI Giants Are Paying Billions for Power
Google and Anthropic just agreed to pay SpaceX roughly $2.2 billion a month for compute. The real story is older than AI: whoever controls the power controls the future. Why energy, not algorithms, is the moat.
Energy Is the Only Moat: Why the AI Giants Are Paying Billions for Power
Google and Anthropic just agreed to pay SpaceX roughly $2.2 billion a month for compute. Strip away the headlines and the real story is simpler and older than AI: whoever controls the power controls the future. Here is why energy, not algorithms, is the moat.
This week the AI arms race revealed what it actually runs on, and it is not genius. It is electricity.
In the span of a few days, two of the most important companies in technology signed away staggering sums for raw compute capacity. Google agreed to pay SpaceX about $920 million per month from October 2026 through June 2029 for access to roughly 110,000 NVIDIA GPUs, a contract worth around $30 billion. Anthropic, fresh off a confidential filing to go public near a trillion-dollar valuation, committed to paying SpaceX about $1.25 billion per month through May 2029 for exclusive access to the Colossus 1 data center in Memphis, more than 220,000 GPUs running on 300 megawatts of power. Together the two deals are worth more than $70 billion and put roughly $2.2 billion a month into SpaceX's pocket for compute capacity.
Read those numbers again and notice what is being bought. Not models. Not research talent. Not data. The single most valuable thing in artificial intelligence in 2026 is the ability to plug in hundreds of thousands of chips and keep them fed with reliable, affordable power. We called this when SpaceX filed to go public, in our breakdown of the IPO and its orbital ambitions. The compute deals confirm it.
The Number Nobody Can Engineer Around
The constraint is physical, and it is getting worse fast. Data centers already consume around 415 terawatt hours of electricity a year, about 1.5% of the global total, according to the International Energy Agency. That figure is projected to roughly double to around 945 terawatt hours by 2030. Some estimates put data center consumption near 1,050 terawatt hours as soon as this year, which would make data centers, if they were a country, the fifth-largest electricity consumer on earth, sitting between Japan and Russia.
The growth curve is the part that should focus the mind. Data center electricity demand grew 17% in 2025, and the AI-specific slice grew 50% in the same year. The IEA's satellite tracking shows that purpose-built AI factories have more than tripled their capacity in the past 18 months. In the United States, data center energy demand is forecast to climb 130% by 2030.
You cannot software your way out of this. A more efficient model still needs to be trained, and training is measured in megawatt-hours. The chips keep getting hungrier, the clusters keep getting bigger, and the grid, built over a century for a slower world, cannot be expanded on the timeline the AI race demands. That is why the IEA expects 15 to 27 gigawatts of onsite natural gas to be powering data centers by 2030. When you cannot wait for the utility, you build your own power plant.
Who Already Knows How to Do This
Here is the part that should interest anyone watching the mining industry. The problem the AI giants are now scrambling to solve, securing large amounts of cheap, reliable electricity and converting it into computation, is the exact problem Bitcoin miners have been solving for over a decade.
Mining taught a generation of operators how to site a facility next to stranded hydro, curtailed solar, or flared gas, how to negotiate interruptible-load deals with grid operators, how to build behind the meter, and how to run hundreds of megawatts of hardware profitably on margins measured in cents per kilowatt-hour. That skill set was treated as a curiosity when the only output was hashes. Now that the output can be AI inference, it is suddenly one of the most valuable competencies in technology. We wrote about this structural shift as it was beginning, in Bitcoin miners are becoming AI infrastructure companies, and about the underlying physics in AI is burning through power, and Bitcoin miners already know how to handle it.
The SpaceX deals are the same logic at the largest possible scale. Colossus 1 is not special because of its chips, which anyone can buy. It is special because it is 300 megawatts of secured, operational power capacity in one place, ready to run. That is the asset Anthropic paid $1.25 billion a month to monopolize. The GPUs are a commodity. The power behind them is the moat.
Why This Is an Ultra Labs Worldview, Not a Hot Take
Our entire operation is built on the conviction that energy is the base layer of everything that matters in this industry. It is why we run our machines on low-cost local hydroelectric power rather than treating electricity as an afterthought, because lower energy cost is not a footnote, it is the whole game. It is why the convergence of AI and crypto, which we mapped in when AI meets crypto, runs straight through the power bill.
The orbital data centers in SpaceX's longer-term vision are the ultimate expression of this same instinct: if the grid is the bottleneck, escape the grid entirely and harvest uninterrupted solar in space. Whether that future arrives on schedule or not, the principle it is chasing is already governing the present. The companies winning the AI race are the ones who treated power as the prize.
For the better part of twenty years, the smartest thing you could own in technology was intellectual property. The lesson of this week is that the smartest thing to own now is a substation. Algorithms are abundant and increasingly open. Talent is mobile. Chips can be bought by anyone with a purchase order. Cheap, reliable, scalable power is the one input that is genuinely scarce, genuinely hard to build, and genuinely defensible. It is the only moat that matters, and the people who have spent a decade quietly mastering it are about to find the rest of the world has come looking for what they already know.
Ultra Labs is a US Bitcoin mining and crypto infrastructure company powered by renewable energy and built on Cardano. This article is for informational purposes only and is not financial advice. Always do your own research.