Federal regulators are rubber-stamping a disaster. The recent regulatory push to greenlight Donald Trump’s plan to fast-track electricity to energy-hungry AI data centers is being cheered by Big Tech and market spectators as a victory for innovation. They are dead wrong.
The media narrative is predictably lazy: AI is the future, data centers need juice, and the Federal Energy Regulatory Commission (FERC) is finally cutting red tape to keep America ahead in the global tech race. This perspective completely misreads how electrical grids actually operate. Fast-tracking power to data centers does not accelerate the future; it cannibalizes existing infrastructure, distorts energy markets, and shifts massive financial liabilities onto everyday ratepayers.
I have spent years analyzing energy grid infrastructure and corporate power purchasing agreements. I have watched tech giants claim they are driving a green energy revolution while secretly hoarding baseload fossil-fuel power to keep their uptime at 99.999%. The current regulatory rush is not a modernization strategy. It is an act of desperation disguised as progress.
The Co-Location Illusion
The core of the administration’s plan relies heavily on "co-location"—building massive data centers directly adjacent to existing power plants, particularly nuclear facilities. The logic sounds clean on paper: bypass the bogged-down regional transmission grid, plug the data center directly into the generator, and avoid years of interconnection delays.
It is a phantom solution.
When a 1,200-megawatt nuclear plant diverts its output directly to an adjacent cloud data center, that electricity does not magically appear from a new source. It is subtracted from the public grid. The regional transmission organization (RTO) suddenly loses a massive chunk of reliable, emissions-free baseload power that it relied on to balance the system during peak demand.
Consider the PJM Interconnection, which services the Mid-Atlantic. If tech giants successfully ring-fence existing nuclear plants in Pennsylvania, the grid operators must replace that lost capacity to keep the lights on in Baltimore, Washington D.C., and Chicago. How do they do it? They fire up old, inefficient natural gas peaker plants or coal units.
The immediate result is a paradox: fast-tracking AI power under the guise of efficiency actually increases grid instability and drives up carbon intensity. You cannot optimize a network by letting its largest users cut the line and hoard the highest-quality assets.
The Interconnection Queue Deception
Regulators claim that fast-tracking data centers solves the gridlock in America's interconnection queues. This is a fundamental misunderstanding of the bottleneck. The queue is not clogged because regulators are slow; it is clogged because speculative energy developers flooded the system with unviable projects, and the physical capacity to upgrade lines does not exist.
Imagine a scenario where a highway is completely gridlocked. The government's solution is to create an express lane, but only for ultra-luxury SUVs. The traffic for everyone else does not disappear; it compresses into the remaining lanes, while the luxury vehicles accelerate wear and tear on the road surface without paying their fair share of maintenance.
Data centers are the luxury SUVs of the energy world. They operate with a flat, continuous demand profile (a 100% load factor). Unlike a residential neighborhood that uses less power at 3:00 AM, a data center sucks maximum power 24 hours a day, 365 days a year.
When FERC accelerates these setups, they allow tech companies to skip the deep grid impact studies required to understand how this constant, unyielding draw affects regional transformers and substations. We are skipping the structural engineering phase of a skyscraper because the tenant is in a hurry to move in.
The Ratepayer Bailout
Let’s talk about the money, because this is where the corporate gaslighting becomes egregious. Big Tech argues that their massive capital expenditures will fund grid upgrades that benefit everyone. The opposite is true.
Under standard utility frameworks, the cost of building new high-voltage transmission lines and upgrading substations is socialized across the entire ratepayer base. When an electric utility builds a multi-billion-dollar line to service a new industrial zone dominated by tech facilities, everyday homeowners and small businesses see their monthly bills climb.
I have reviewed utility rate cases where tech infrastructure drove 70% of the projected regional load growth, yet residential consumers bore the brunt of the capital recovery costs. Big Tech negotiates special economic development tariffs that cap their exposure to these costs. They get the subsidized infrastructure; the public gets the bill.
If a data center company wants immediate access to gigawatts of power, they should be forced to pay 100% of the merchant transmission costs required to bring new, incremental generation online from scratch. No exceptions. No fast tracks. If the AI business model cannot survive paying the true, unsubsidized cost of its primary raw material—electricity—then the AI business model is fundamentally broken.
The False Promise of Virtual Power Plants
A common counter-argument from Silicon Valley insiders is that data centers will eventually act as "virtual power plants" (VPPs). They claim that by deploying massive on-site battery storage and backup diesel generators, data centers can feed power back into the grid during emergencies.
This is a tech-utopian fantasy.
Data center operators are fiercely risk-averse. Their entire compensation structure is tied to keeping the servers humming. The moment a grid operator asks a data center to discharge its batteries or curtail its workload to save a failing local grid, the corporate lawyers will point to their uptime guarantees.
Furthermore, backup diesel generators are permitted under strict environmental limits that restrict their operation to a few dozen hours per year, typically reserved for actual emergencies to prevent facility meltdowns. Turning them into market-facing generation assets to bail out a stressed grid is logistically and environmentally impossible under current clean air laws.
Dismantling the Capital Expenditure Myth
The financial markets are currently valuing tech companies based on how many graphics processing units (GPUs) they can secure and deploy. This has triggered a capital expenditure arms race.
Companies are burning through billions of dollars to build out facilities before they even have a clear path to monetization for their enterprise AI tools. They are building capacity for demand that is highly speculative.
By forcing the energy sector to distort its long-term planning cycles to accommodate this frantic, short-term build-out, regulators are exposing the energy grid to immense asset stranding risk. If the AI monetization wave crests and crashes over the next five years, utilities will be left with billions of dollars in specialized infrastructure built for customers that no longer need it. The tech companies will write off the losses, file for restructuring, or pivot to the next trend. The physical wires and the debt used to build them will remain anchored to the local utility, paid for by consumers for the next thirty years.
The Hard Realities of Physics
You cannot write software to bypass Ohm's law. You cannot patch a physical transmission line with an over-the-air update.
The federal push to accelerate power delivery to data centers ignores the physical reality of our manufacturing supply chains. Right now, the lead time for a high-voltage step-up transformer—the critical piece of hardware needed to connect a large facility to the high-voltage grid—is anywhere from three to four years. The global supply of electrical steel is severely constrained. There is an acute shortage of skilled lineworkers and substation engineers.
Fast-tracking regulatory approval does absolutely nothing to manufacture steel faster. It does nothing to train thousands of specialized technicians overnight. It merely creates a regulatory bottleneck where approved permits pile up on desks while actual projects stall in the mud, driving up procurement costs through bidding wars that tech giants will always win against local municipal utilities.
Stop asking how we can fast-track power to AI data centers. The real question we should be asking is why we are allowing a single, highly speculative industry to hijack public infrastructure planning, distort energy markets, and dictate national energy policy at the expense of everyone else.
If tech monopolies want to build the future of intelligence, they need to build their own independent power grids from the ground up, pay the full market price for their footprint, and stop treating the public grid as a venture capital subsidy. Turn off the fast track. Force them to wait in line like everyone else.
