A single 300-megawatt AI data center consumes enough electricity to power a small city. The cost per kilowatt-hour in Northern Virginia has risen 40% in two years. The grid does not lie, it only waits to be read. Now, state legislators in Arizona, Virginia, and Oregon are demanding profit-sharing agreements from Big Tech operators before granting new permits. This is not a political stunt. It is an accounting reality that the industry has refused to acknowledge.
Context: The Energy Appetite That Broke the Grid
The narrative around AI data centers has been one of prosperity. Companies like Amazon, Google, and Microsoft are building facilities at a pace that overwhelms local utility capacity. These centers run 24/7, drawing baseload power that was previously reserved for residential and industrial use. States initially welcomed the tax revenue. But the power demand has become a liability. Utilities are forced to build new transmission lines and natural gas plants, costs that are passed to ratepayers. The revolt is not ideological. It is structural.
Policymakers are now pushing for a share of the profits from these data centers, arguing that the public should benefit from the economic surplus generated by public infrastructure. The proposed models vary: a percentage of gross revenue, a per-megawatt levy, or a requirement to invest in local renewable energy. The core demand is the same: make the data center operators pay for the true cost of their energy consumption.
Core: The Mathematical Certainty of Unsustainable Cost
Based on my audit experience analyzing energy contracts for Bitcoin mining operations, I can state with certainty that the economics of AI data centers are far more fragile than the public relations suggest. Let me walk through the numbers.
A typical 100MW data center costs approximately $600 million to build. The annual electricity cost at $0.08 per kWh is $70 million. Assuming a 70% utilization rate for AI compute, the revenue from renting GPU clusters is roughly $300 million per year. Gross profit before depreciation and taxes is $230 million. Now add a 15% profit-sharing requirement. That reduces net profit by $45 million. The margin collapses from 38% to 23%. The grid does not lie, it only waits to be read.
But the real flaw is not the profit-sharing percentage. It is the assumption that energy costs will remain stable. Every year, utilities file for rate increases to cover grid upgrades. In Virginia, the proposed rate increase for 2025 is 12%. At that rate, the same data center's electricity cost jumps to $78 million, reducing profit before sharing to $222 million. After profit-sharing, net profit drops to $189 million. The return on investment falls below 10%. This is not a sustainable business model.
The ledger does not lie, it only waits to be read. The data centers are essentially arbitraging cheap energy by lobbying for special rates and deferred infrastructure costs. The profit-sharing demand is simply a mechanism to internalize the externality. But the market has not priced this risk. The current valuation of AI infrastructure companies assumes no regulatory friction. That assumption is mathematically invalid.
Contrarian: What the Bulls Got Right
To be fair, the bulls have a point. AI data centers generate significant economic activity. They create high-paying jobs, spur innovation, and attract ancillary businesses. The energy demand also forces utilities to modernize the grid, which benefits all ratepayers in the long run. The profit-sharing model could be designed as a subsidy for renewable energy deployment, creating a virtuous cycle.
But the blind spot is centralization. These data centers are owned by three or four companies. They are clustered in a few regions. If one state introduces a profit-sharing requirement, the operators can simply move to another jurisdiction. This creates a race to the bottom, where states compete to offer the lowest energy costs and regulatory leniency. The result is a fragmented system where no one pays the true cost of energy. The grid becomes a common pool resource that is depleted without accountability.
The ledger does not lie, it only waits to be read. The profit-sharing demands are a sign that the free ride is ending. The question is not whether the data centers will pay more, but whether the operators will adapt by decentralizing their energy sourcing. Blockchain-based energy credits, peer-to-peer grid trading, and on-site renewable generation with battery storage are the logical next steps. The operators who invest in these solutions now will survive the regulatory shift. Those who fight it will face margin compression and eventual exit.
Takeaway: The Accountability Call
The grid does not lie, it only waits to be read. The profit-sharing demands are a corrective mechanism for a market failure. The core insight is that energy is not a free variable. It is a constraint that must be accounted for. The AI data center boom will continue, but the cost structure will change. The winners will be those who build decentralized energy infrastructure, not those who lobby for loopholes. The ledger is being written, and the state legislators are the accountants. The question is whether the industry will read the numbers before the margin disappears.