---

---

Hectic Power: The Data Center Problem We Aren’t Really Talking About

Section One

In 1992 Snap had a number one record and a generation was 20 years old. The grid was already middle aged and quietly failing. Nobody told us. Nobody pulled the trigger. Here we are, Gen X, now middle aged, and we need to do something NOW.

The reason this is suddenly urgent has a name. Data centers. Specifically the hundreds of hyperscale facilities being proposed, permitted, and built across the country to power the artificial intelligence boom. They are arriving fast, they are arriving large, and they are landing on top of an electrical grid that was never built to carry them.

Politicians on both sides have found their preferred performance. Opposition plays well on social media; water fears, rate spikes, corporate greed, regardless of whether the underlying diagnosis is accurate. Support plays equally well in economic development press releases; jobs, investment, American competitiveness, regardless of whether any of those promises are structured to actually materialize. One side is fighting a villain that is partly a symptom. The other side is celebrating a solution that is partly a subsidy.

Nobody is talking about the actual problem. The American power grid needs a complete overhaul and the data center boom is the first forcing function in fifty years powerful enough to pay for it.

The American electrical grid was largely built in the 1950s and 1960s and never comprehensively modernized. Not once. Seven decades of patches, extensions, and jury-rigging have produced a 1970s architecture now being asked to carry loads it was never designed to handle simultaneously: hyperscale AI inference running twenty-four hours a day, cloud infrastructure that never sleeps, residential air conditioning expanding geographically as heat pushes northward, electric vehicles adding the draw of a small commercial operation to every residential block, and distributed solar feeding back in a direction the original one-way system never anticipated.

Any one of those loads would stress a fifty-year-old system. All five at once is a different category of problem entirely. Think of it this way. The grid is a duct-taped power strip that is already running the space heater. Data centers just walked in carrying the Christmas tree. And that is before we talk about the EV charger in the garage.

Data centers did not create this situation. They exposed it. And if you want to understand why your above-ground power line looks exactly the same as it did in 1992, same poles, same wire, same moderate wind event away from a neighborhood outage, the answer is not the data center down the road. It is fifty years of deferred decisions finally arriving all at once.

The anti-data-center sentiment is understandable. The pro-data-center enthusiasm is understandable. Both are aimed at the wrong target. The fight is about the Christmas tree. Nobody is looking at the duct-taped power strip it just got plugged into. We will come back to that power strip later because the stakes get considerably higher before this argument is done.

---

Right Concerns. Wrong Argument. Worst Outcome.

Section Two

Every faction in this debate is correct about what they care about. That is what makes the argument so frustrating and the opportunity so easy to miss.

The permitting advocates are right. A ten-year approval timeline for transmission infrastructure is indefensible when China is breaking ground on the equivalent in eighteen months. Every year of delay is a year of competitive advantage handed to a government that does not hold public comment periods.

The water activists are right. Hyperscale data center cooling draws on municipal water systems built with public bond financing, maintained by ratepayers, and never designed for this volume. The concern is not a myth manufactured by NIMBY politics. It is a real problem exacerbated by inefficient cooling practices that persist because nobody has built incentives to innovate past them. Liquid cooling, immersion cooling, and closed-loop systems exist and work. They are not universally deployed because the current framework does not require it.

The rate impact crowd is right. PJM, the transmission network serving a quarter of the entire country, 65 million ratepayers across thirteen states, currently operates roughly 182 gigawatts of generating capacity. Data centers are projected to add demand equivalent to a third of that total over the next decade. A decade ago PJM’s annual load growth was 0.3 percent. It is now 3.6 percent, twelve times higher, driven almost entirely by data centers. That demand drove 63 percent of the capacity cost increase in the 2025–2026 auction, adding $9.3 billion to the regional bill in a single year. Capacity prices jumped nearly tenfold between 2024 and 2026. The average PJM family is looking at $70 more per month by 2028.

Now multiply that across the four other major grid regions covering the rest of the country. PJM is the canary. The same demand curve is hitting MISO across the Midwest, SPP across the Great Plains, WECC across the West, and ERCOT across Texas. The data center buildout is a national load event landing on a patchwork of regional grids with no coordinated response and no structured obligation framework. Grandma is paying more in Pennsylvania. She is also paying more in Ohio, Illinois, Virginia, and Texas. She just doesn’t know exactly why and neither does the politician telling her who to blame.

The clean energy advocates are right. Two-hundred-year-old combustion technology is not just an environmental liability. It is an efficiency and competitiveness liability. Coal plants are slow to ramp. Natural gas carries permanent fuel cost volatility baked into every kilowatt-hour forever. AI inference demand is spiky and unpredictable. The mismatch between legacy generation and AI load profiles is architectural. Aging power systems chasing a fast intelligent load they were never designed to serve.

The objection you hear most often, what happens when the sun doesn’t shine and the wind doesn’t blow, was a reasonable concern twenty years ago. In 2026 it is a half-truth that has outlived its accuracy. Grid-scale battery storage, pumped hydro, and advanced geothermal don’t care what the weather is doing. Small modular nuclear reactors run twenty-four hours a day regardless of season. The honest answer to the intermittency objection is that we are not proposing to run AI data centers directly off solar panels. We are proposing to build a grid architecture where storage smooths variable generation into reliable baseload that can actually serve AI load profiles.

High tech infrastructure deserves high tech power. Advanced nuclear, enhanced geothermal, and storage-backed renewables match the demand curve better and make American AI structurally cheaper to run than Chinese AI on coal.

The economic development advocates are right that data centers represent real capital investment. But they are overselling the long-term jobs story and underselling the platform potential. A hyperscale facility employs dozens of permanent workers not thousands. The construction jobs are real and finite. The innovation multiplier, the supply chain activity, the startup ecosystem, the technical workforce development requires intentional policy to materialize. It does not happen automatically because a server farm exists in your county. If communities are going to accept the land use, the water draw, and the rate impacts, they deserve a structured return that goes beyond a ribbon cutting and a handful of permanent positions.

The community opposition groups are right that something is wrong. But without a coherent national framework giving them honest accounting of costs and benefits, they are defending themselves with whatever tools are available, including half-truths that are close enough to real concerns to be genuinely dangerous. In Utah a data center footprint the size of Manhattan moved through approval processes that bypassed normal public scrutiny. In Plymouth Township, Pennsylvania a local facility tried to railroad planning approval and drew organized bipartisan opposition built on a combustible mix of legitimate grievance and social media misinformation. Some of that opposition was coordinated on a social media platform owned by the country currently outbuilding us eight to one on power infrastructure.

That is what happens when legitimate concerns meet social media algorithms and geopolitical blind spots in the absence of a coherent national framework. Every faction has a legitimate grievance. Every faction is negotiating over the furniture while the foundation cracks underneath them.

Here is what none of them are talking about.

The hyperscalers are not waiting for this argument to resolve. Faced with a grid that cannot deliver what they need they are building private power generation and designing new facilities to operate independently of public infrastructure entirely. In early 2025 virtually all data center power flowed through the public grid. By early 2026 nearly a third of planned new capacity is designed to go off-grid. Microsoft, Google, and Amazon have effectively concluded that the American grid is not a reliable partner for the AI economy they need to build.

They are building their own.

And they are still using the land. Still drawing the water. Still driving the roads. Still connecting to transmission lines ratepayers have funded for seventy years. They have simply decided that the grid, the one shared infrastructure investment that could have been the exchange mechanism, is not worth their participation.

We know where this leads. The companies that control AI infrastructure will control the power that runs it. Not as a side business. As a vertical monopoly. Google Power. Amazon Energy. Meta Electric. X Grid. Private generation networks answerable to shareholders not utility commissions, not state regulators, not local communities, and not the ratepayers who built the infrastructure they are currently using for free. No universal service obligation. No rate case. No public comment period. You will buy your electricity from the same company that sells you cloud services, serves your ads, and hosts your social media. Your only recourse when the price goes up will be to use less of all of it. And what happens after they declare bankruptcy when their grid starts a forest fire.

The communities fighting data centers today over water and rates are worried about the wrong company at the wrong stage. The data center down the road is not the threat. Google Power is the threat. And the window to prevent it is open right now and closing faster than anyone in this argument seems to recognize.

The debate we are having is about symptoms. The debate we need is about the foundation. And while we are having the wrong one the duct-taped power strip is running out of outlets and the replacement being shipped by Amazon does not even have an underwriters lab seal of approval.

---

The Infrastructure Investment Pattern

Section Three

America has been here before. Three times last century the country faced a moment where shared infrastructure was the only answer, private capital alone would not build what everyone needed, and the argument that broke the political logjam was not economics or equity or environment. It was survival. And every time the investment paid off in ways nobody fully predicted when they made it.

Rural electrification. The interstate. The internet. Three different decades. Three different threats. Same pattern every time. Build the shared foundation. Structure the private access. Let broad prosperity follow. All ships rose.

The fourth moment is here. And we are spending it arguing about the furniture.

Rural Electrification

In the 1930s a third of the country had no reliable electricity. Private utilities had done the math and made a rational decision. Rural America was too expensive to connect and too poor to generate adequate returns. The Great Depression had collapsed farm incomes. The Dust Bowl had turned the agricultural heartland to powder. The country that fed the world was running on kerosene lamps and hand pumps while urban America had moved into the modern age.

FDR’s Rural Electrification Administration did not ask private utilities to serve rural communities out of generosity. It built the infrastructure, structured the access, and within a decade transformed American agriculture, rural productivity, and the quality of life for millions of people the market had written off as uneconomical. Farms electrified. Productivity soared. Rural communities that private capital had abandoned became viable participants in the American economy. All ships rose because the foundation got built. The utilities that had refused to connect rural America did not disappear. They adapted, competed, and in many cases thrived on top of the infrastructure the public built. That is the pattern working exactly as designed.

The Interstate

In 1956 Eisenhower signed the Federal Aid Highway Act. The formal name tells you everything about how it was sold. The National Interstate and Defense Highways Act. Defense was not a footnote. It was the argument. The US needed roads wide enough to move troops and equipment, to evacuate cities, to disperse populations in the event of nuclear attack. The Soviet threat made every faction say yes simultaneously. Republicans heard national security. Democrats heard jobs and economic development. Business heard markets and logistics. Rural America heard connectivity and access. Everyone got something real and the spine was survival.

The interstate did not just move tanks. It moved goods, people, and ideas at a scale that restructured the entire American economy. The trucking industry. Suburban expansion. The supply chain revolution. Fifty years of GDP growth. All of it rode on top of a defense investment nobody initially justified on economic grounds.

The Internet

ARPANET was not built to connect college campuses. It was built to survive a nuclear strike. The defining architectural principle was route around damage, find another path, never depend on a single node. It was a direct response to Soviet missile capability. The Pentagon funded it. The defense establishment built it. And then the commercial internet, the world wide web, the entire digital economy, Amazon, Google, Facebook, and every company that followed, all of it was a co-benefit of a defense investment made because survival demanded it.

The Pattern

Interstate. Rural electrification. Internet. Same logic every time. Existential threat. Shared public infrastructure built on a survival argument. Private sector innovation structured around access to that foundation. Broad prosperity as the co-benefit.

None of these programs were sold primarily on their economic benefits. None required the private sector to be charitable. All of them structured private access to public infrastructure and generated innovation and wealth at a scale nobody predicted when the original investment was made.

And all of them had a forcing function that cut across every political division and made every faction say yes simultaneously. The Soviet threat. The Depression and the Dust Bowl. National survival. When the argument is existential the political logjam breaks.

The Electron Gap

China’s industrial electricity prices run 30 to 60 percent lower than US markets. Their grid is younger, more modern, and being purpose-built around AI infrastructure demand with the full force of centralized state planning behind it.

Morgan Stanley forecasts that US data centers could face a 44-gigawatt electricity shortfall within three years. The hyperscalers already know this. That is why they are going off-grid. That is why Microsoft committed $15 billion to data center development in the UAE. That is why the geographic map of AI infrastructure investment is being redrawn around power availability rather than population centers or fiber density.

The US has the best AI chips. China has the energy to run them at scale. And DeepSeek demonstrated that the chip gap is closable faster than anyone predicted. The export controls on advanced semiconductors that America bet on as its insurance policy are being picked open from the inside while we argue about water rights in Plymouth Township.

This is bigger than an energy race. The country that powers the AI economy shapes it. The values, the rules, the surveillance architecture, the censorship defaults, the version of history that gets encoded into systems running on that infrastructure. All of it follows the power. A world running its AI on Chinese infrastructure is not just a world where America lost an economic competition. It is a world where the question of whether a tank runs over a protester in a public square gets answered by the infrastructure layer before any human makes a decision. That is a question about whose values get built into the foundation that everything else runs on. And right now we are losing that argument one gigawatt at a time while arguing about water rights in Plymouth Township.

The Intergrid

The interstate connected physical America. Rural electrification connected rural America to the modern economy. The internet connected digital America. The Intergrid connects the energy layer that everything else runs on, designed from scratch for the era we are actually in rather than the era we left fifty years ago.

Not a patched grid. Not an upgraded grid. A reimagined grid built on the same architectural principle that made the internet survive a nuclear strike. Multi-path. Regionally resilient. No single points of failure. Power routes around damage the way packets route around a downed node. Modern generation tied structurally into new transmission corridors. Coordinated capital flowing into a shared platform. Workforce development built in as a first-class component from the start.

The Intergrid is the fourth entry in a pattern America has executed three times when survival demanded it. It is not a green program. It is not a government wish list. It is the same deal America made three times last century. Build the shared foundation. Structure the private access. Let broad prosperity follow.

And if we build it right the next layer, interplanetary infrastructure, orbital compute, the edge network that does not yet exist, builds on something solid. If we get it wrong we will power our data centers with satellites, call the debris field progress, and wonder why the lights keep going out.

This moment is identical. We have the pattern and the power.

---

New Threat. The Same Pattern.

Section Four

The Threat

The United States has the best artificial intelligence chips in the world. China has the energy to run them at scale. China added 429 gigawatts of new power capacity in 2024. The United States added 51. That eight to one gap is the defining infrastructure deficit of this era and it is compounding faster than the political debate has caught up with.

Data centers are projected to require between 80 and 100 gigawatts of new US power capacity by 2030. To put that number in terms every American can feel regardless of where they live: 100 gigawatts is roughly the combined total electricity output of the following states.

Nearly 30 states worth of generating capacity, added to the grid in less than five years, on top of a system already straining before AI arrived, while permitting timelines run seven to ten years and new transmission corridors take a decade to approve.

Goldman Sachs projects US data center demand alone will more than double in the next two years, from 31 gigawatts today to 66 gigawatts by 2027. The Department of Energy estimates the AI sector needs 50 gigawatts of new capacity by 2028 just to maintain current competitive position. We are on track to deliver less than half of that on time. PJM, one grid region covering a quarter of the country, projects 32 gigawatts of new peak demand through 2030 with data centers driving 94 percent of that growth. Multiply that across the four other major grid regions and the national supply gap becomes a national competitiveness crisis.

China has a head start on grid architecture and an eight to one advantage on new generation capacity. Their transmission infrastructure is newer, their planning is centralized, and their time horizon is measured in decades not quarterly earnings cycles. But their generation mix is still heavily reliant on 150-year-old combustion technology and that is where the distinction matters most and creates a once in a generation opportunity, pun intended.

Generation choices made today lock in operating costs and competitive position for thirty years. The Trump administration is pursuing rollback of emission standards that currently limit new coal construction, and that fight will play out in courts and agencies for years. But the timeline argument does not depend on the regulatory outcome. Even a coal plant approved today, financed today, and permitted overnight does not produce power before 2036. The AI infrastructure window is five years not fifteen. 150-year-old combustion technology cannot win a race it cannot finish in time. The physics do not negotiate with the politics.

Modern generation wins this race on cost, efficiency, and construction speed. The permitting timeline is the shared bottleneck across all generation types and it is a framework problem not a technology problem.

On cost Lazard’s 2025 analysis is unambiguous. Utility scale solar generates power at $29 to $62 per megawatt-hour. New coal runs $68 to $166 per megawatt-hour on a new build basis. Natural gas falls in between but carries permanent fuel cost volatility baked into every kilowatt-hour for the life of the plant. Modern generation’s marginal cost after construction is effectively zero. No fuel to buy. No commodity price to absorb. No thirty-year contract locking in yesterday’s costs.

On construction speed the numbers are decisive. Once a site is permitted and financed solar construction runs eight to twelve months from groundbreaking to power. Wind runs six to twelve months. Battery storage can go from contract to operation in under eighteen months. These are not slow technologies. They are fast technologies trapped in a slow framework. The three to seven year total project timeline is permitting and interconnection studies, not concrete and steel. Fix the framework and modern generation deploys faster than any 150-year-old combustion alternative at a fraction of the construction cost.

Coal is slow on both dimensions. Permitting runs three to five years. Construction runs seven to nine more. Total timeline ten to fourteen years minimum. The AI infrastructure window is five years not fifteen.

SMRs are targeting first commercial deployments around 2030 to 2035. Regulatory approval is the long pole. Construction itself runs three to five years once licensed. Google has already contracted with Kairos Power targeting this decade. The baseload future is nuclear. The near-term future is solar, wind, and storage deployed fast through a framework that removes the permitting bottleneck.

150-year-old combustion technology converts roughly a third of its fuel into usable electricity and loses the rest as heat. Modern generation skips the heat cycle entirely. And unlike centralized fossil plants feeding power through long transmission corridors with line losses at every mile, modern generation can be sited adjacent to the load it serves. A data center campus with co-located solar and battery storage draws power from meters away not hundreds of miles.

China is building on 150-year-old combustion technology now and betting on a transition later. America can build modern generation for modern infrastructure, skip the transition entirely, and arrive at the destination China is trying to reach before they get there at lower operating cost, higher efficiency, and without thirty years of fuel price volatility baked into every AI inference our companies run.

The Applied Pattern

Private capital is not the villain in this story. It is behaving exactly as private capital always has when public infrastructure does not exist to shape the incentive structure.

The interstate did not get built by asking the trucking companies, the shipping industry, and the logistics operators who would profit most from it to voluntarily contribute. Left to their own devices they would have built private toll roads serving their own routes and called it a market solution. The public would have gotten the gaps.

The internet did not get built by asking AOL, CompuServe, and Prodigy to voluntarily open their walled gardens because it was good for the public. Left to their own devices they would have built competing private networks serving their own subscribers and called it innovation. The open internet emerged because publicly funded infrastructure established open standards that made the walled garden model obsolete.

The hyperscalers are behaving rationally right now. Faced with a grid that cannot deliver what they need they are building private power generation and designing new facilities to operate independently of public infrastructure. They are getting the land. They are getting the water. They are getting the roads. They are getting seventy years of ratepayer-funded transmission infrastructure. They are getting the public internet backbone their entire business model runs on. And they are deciding that the grid, the one shared infrastructure investment that could be the exchange mechanism, is not worth their participation.

They are building the private toll road. The public is getting the gaps. The ratepayers are getting the bill. And if nobody intervenes the endgame is Google Power and Amazon Energy, private generation networks answerable to shareholders not utility commissions, not state regulators, not local communities. No universal service obligation. No rate case. No public comment period. Your electricity bill and your cloud bill arriving from the same company with the same leverage over both.

We have seen this before. It is the pre-interstate highway system replaying itself in the energy layer. And we know exactly how it ends without a framework to change the incentive structure because we have ended it three times before.

The numbers make the extraction visible. Nearly 900,000 technology workers have been laid off since 2020. Over the same period the largest technology companies added roughly fifteen trillion dollars in combined market capitalization. The wealth did not disappear when the jobs did. It moved from salaries and middle class spending power into equity valuations held by the top fraction of one percent of shareholders.

The dot-com era made a million millionaires. Stock options, early employees, engineers across hundreds of companies. Wealth concentrated at the top but distributed meaningfully into a broad technical class. The AI era is a different story. Eight companies now account for 36 percent of the entire S&P 500, a concentration one senior S&P analyst described as having no historical comparisons. The gains are flowing into founder shares, institutional holdings, and the portfolios of people who were already wealthy enough to own significant positions. The AI economy is producing two dozen billionaires and one or two trillionaires while the CEOs of the largest AI companies run a betting pool on who builds the first billion dollar company with a single human employee.

That bet has already been won. Matthew Gallagher launched a telehealth startup in September 2024 with $20,000 and zero employees. First full year revenue was $401 million with a 16.2 percent net profit margin. His only hire was his brother. The company is tracking toward $1.8 billion in 2026. He built nothing new. He rented existing infrastructure, public internet backbone, cloud platforms running on public power grids, municipal water systems, optimized the customer-facing layer above it and kept every dollar of margin.

That is a description of a system with no exchange mechanism. And it is about to scale in ways that make the individual story look modest.

China is not watching from the sidelines. Chinese local governments are already building government-subsidized infrastructure for solo founders to replicate this model at national scale. Suzhou announced plans for 1,000 AI-powered solo enterprises by 2028. Shanghai is handing out free computing power to solo founders. The one-person billion dollar company is not an American innovation story anymore. It is becoming a Chinese industrial policy. The same model America stumbled into by accident China is building deliberately with state capital and national coordination.

The wealth concentration is accelerating. The job displacement is accelerating. The ratepayer transfer is accelerating. The private secession from shared infrastructure is accelerating. The external competitive threat is accelerating. Two fronts. One window. And the window is not staying open.

Making AI pay for access to American commons is not a radical proposition. It is the interstate model and the internet model applied to 2026. You want access to what we built together, you contribute to keeping it built. The trucking companies drove on the interstate and generated enormous wealth doing so. They did not build it. They paid into the system that maintained it. That is the exchange. That is all this is.

The AOL walled garden did not survive open standards. The private turnpike did not survive the interstate. Private power generation will face the same pressure once an Intergrid framework establishes open interconnection standards, structured access obligations, and a fast lane for those who contribute to the platform everyone runs on.

The regulated utility model adds one real friction point worth naming. Utilities are guaranteed return monopolies with limited incentive to accelerate modernization unless regulators force it. The Intergrid framework has to flow through those same utility commissions. The answer is rate structures and interconnection requirements that make modernization the path of least resistance rather than a disruption to a comfortable business model. A requirement with teeth not a wish.

The deal is straightforward. You want the fast lane, eighteen months not ten years, you sign the grid investment obligation proportional to your load draw at interconnection approval. A binding condition of doing business on shared American infrastructure. Something every faction in this debate, left, right, and center, can read and find something real in.

---

The Intergrid Vision

Section Five

The interstate was not a wider dirt road. It was a different category of infrastructure entirely. New engineering standards. New right of way. New interchange architecture. Built for the traffic loads of the future not the traffic loads of 1935. The upgrade was not incremental. It was generational.

The Intergrid is a different category of infrastructure built for the load profiles of the next fifty years rather than the load profiles of 1965. Not a patch. Not an upgrade. A reimagining built on an architectural model we already proved works. We called it the internet.

The internet survived because it was designed with no center. No single node whose destruction takes everything down. Traffic finds another path. Damage gets routed around. The system degrades gracefully rather than collapsing catastrophically. That was not an accident. It was a design made in response to Soviet nuclear capability targeting centralized communications infrastructure. The Pentagon built survivability into the architecture from the first line of code.

The American grid was designed for the opposite. Centralized generation flowing one direction through hub and spoke distribution to end users with no alternative path when something breaks. One transformer failure cascades into a regional blackout affecting millions. A hurricane hitting the Gulf Coast or an ice storm hitting Texas does not just inconvenience people. It kills people and costs billions because the power has nowhere else to go.

The Intergrid applies the internet’s core architectural principle to energy infrastructure. Multi-path. Regionally resilient. No single points of failure. Power routes around damage the way packets route around a downed node. A hurricane hits the Gulf Coast and the Southeast draws from the Mountain West. A heat dome parks over Texas and demand spikes route to available generation in the Midwest. The system adapts rather than fails.

The technology exists. High voltage direct current transmission can move power efficiently over long distances with minimal loss. Grid-scale storage using iron-air batteries, pumped hydro, and next generation chemistries buffers the mismatch between generation and demand. Advanced power flow controls and software-defined grid management squeeze significantly more capacity from existing corridors without physical construction. The pieces are available. The architecture is proven. What is missing is the framework that builds them into a coherent system rather than a collection of disconnected regional projects.

Generation: Modern by Design

The Intergrid does not require an environmental ideology. It requires an engineering and economics argument that every faction can engage with honestly.

150-year-old combustion technology is the wrong answer for a twenty-first century infrastructure problem on every dimension. It is slow to ramp, expensive to operate, and thermodynamically wasteful as it converts roughly a third of its fuel into usable electricity and loses the rest as heat. Modern generation skips the heat cycle entirely.

The fast lane requires modern low-carbon generation because modern generation is cheaper to build, faster to come online once permitted, and carries near-zero marginal cost after construction. No fuel to buy. No commodity price volatility baked into every kilowatt-hour for the life of the facility. American AI running on near-zero marginal cost power is structurally cheaper to operate than Chinese AI running on coal. That arithmetic compounds over decades.

Solar construction runs eight to twelve months once permitted. Wind runs six to twelve months. Battery storage deploys in under eighteen months. Enhanced geothermal targets four to seven years total. Small modular nuclear reactors are targeting first commercial deployments around 2030 to 2035, providing the long-term baseload backbone the grid needs once the near-term capacity gap is addressed by solar, wind, and storage. The bottleneck for all of them is permitting not construction. That is a framework problem. The Intergrid fast lane is designed to solve it.

Cooling and Water: Efficiency by Obligation

The water concern raised by communities opposing data centers is legitimate and the Intergrid framework addresses it structurally rather than rhetorically.

Current hyperscale cooling draws heavily on municipal water systems because the economics have never required anything better. Evaporative cooling is cheap when water is free. Liquid cooling, immersion cooling, and closed-loop systems exist, work, and use dramatically less water. A modern immersion-cooled facility can operate with near-zero water consumption compared to a traditional air-cooled facility drawing millions of gallons annually. These technologies are not deployed universally because the current framework does not require them to be.

The Intergrid obligation framework changes that. Efficient cooling architecture becomes a condition of the fast lane the same way low-carbon generation does. You want eighteen months not ten years, you build with modern cooling systems that do not drain the municipal water table. That transforms the water concern from a community opposition talking point into a solvable engineering requirement with a market incentive attached.

Advanced cooling also enables higher compute density in smaller physical footprints. More compute per square foot means less land per gigawatt of AI capacity. Less land means easier siting. Easier siting means faster permitting. The efficiency gains compound across every dimension of the opposition argument simultaneously.

Decentralization: The Internet Model Applied to Compute

The internet did not survive because it had one great data center. It survived because it had thousands of interconnected nodes with no single point of failure. Traffic found the best path. Redundancy was architectural not optional.

The current AI infrastructure model is moving in the opposite direction. Hyperscale campuses concentrating enormous compute density in single locations, dependent on single grid connections, vulnerable to single points of failure whether weather, cyberattack, or simple equipment failure. That is the hub and spoke grid model applied to compute. It has the same vulnerability profile.

The Intergrid enables a different model. Distributed data center infrastructure sited closer to modern generation sources, connected through the multi-path Intergrid backbone, with redundancy built into the architecture the same way it is built into the internet. A hurricane hits a Gulf Coast data center campus and compute routes to Mountain West capacity. A cyberattack targets a Mid-Atlantic facility and traffic shifts to Midwest nodes. The system adapts rather than fails.

This also directly answers the community opposition argument. Distributed data centers across more locations with smaller individual footprints means lower water draw per site, lower grid impact per site, more distributed job creation, and more manageable community impact than a single hyperscale campus landing on one township’s planning commission agenda with three months notice. The Plymouth Township problem is partly a scale and concentration problem. Distributed architecture addresses it at the design level rather than the political level.

Smaller regional facilities also enable lower latency for AI inference closer to population centers, which matters increasingly as AI moves from training workloads in remote locations to inference workloads serving real-time applications in every city. The architecture that is more resilient is also the architecture that performs better. Redundancy and performance point in the same direction.

Capital: One Ledger

The current approach to grid investment runs three separate programs on three separate timelines with three separate accountabilities. Federal infrastructure money flows through one channel. Utility capital expenditure flows through rate cases at state commissions. Private data center investment flows wherever land and power are cheapest.

None of them coordinate. All of them underperform as a result.

The Intergrid framework replaces three programs with one coordinated regional ledger. Federal investment anchors the transmission backbone. Utility capital expenditure is directed by interconnection requirements that make modernization the path of least resistance. Data center obligation proportional to load draw, required at interconnection approval, funds the distribution upgrades that connect the backbone to the communities that need it.

One accounting system. One timeline. One regional plan. Federal, state, and private capital each following their own incentive structure inside a framework designed to produce a coherent collective result.

Workforce: A Career Not a Construction Job

Grid modernization at Intergrid scale is a twenty-year construction and maintenance program. New transmission corridors. Substation upgrades. Distribution modernization. Cybersecurity hardening. Software-defined grid management deployment. Advanced cooling system installation and maintenance. Ongoing operation of a system dramatically more complex than what exists today.

Twenty years of work is a career pipeline not a construction job.

The Intergrid workforce program is built into the framework as a first-class component from the start. Regional labor market analysis identifies the specific gaps. Community college and apprenticeship partnerships build the pipeline. Prevailing wage requirements ensure the jobs pay enough to rebuild the middle class that the AI economy is simultaneously hollowing out. Every congressional district gets something real, skilled trades careers that last a generation, not a promise of trickle-down innovation from a data center that employs forty people.

Rural Republican districts get infrastructure investment and career pathways. Post-industrial Democratic districts get the same. The workforce program is the political mechanism that makes the coalition hold across every region and every party affiliation.

Resilience: Built for What Is Actually Coming

The Intergrid is dramatically harder to break than what exists today and that argument belongs in the national security conversation not just the energy conversation.

Extreme weather events are increasing in frequency and severity regardless of where anyone stands on the underlying causes. The economic cost of grid failure is now measured in billions per event. The AI infrastructure running on top of it cannot tolerate the outage profiles the current system produces. Resilience is not an environmental argument. It is a business continuity and national security argument.

The cybersecurity case is the one nobody is making loudly enough. The current grid is a patchwork of aging control systems with inconsistent security standards and documented vulnerabilities that state actors have already demonstrated the capability to exploit. A reimagined Intergrid with modern architecture, software-defined controls, and multi-path redundancy is dramatically harder to attack and dramatically faster to recover from attack than what exists today. Every node that goes down leaves the rest of the system routing around it. That is not a grid feature. That is the internet’s core survival mechanism applied to energy infrastructure.

This is a defense argument. A homeland security argument. An argument that cuts across every ideological faction because nobody wants the lights going out because a foreign actor found the one transformer that takes down a region.

The farmer in rural Pennsylvania gets reliable power. The family in North Philadelphia stops losing power in heat waves. The small manufacturer gets grid capacity to electrify their floor. The hospital maintains operations through the next hurricane. The military base maintains power independence from a grid that adversaries have already mapped and studied for years.

All boats rise. Not because of charity. Because the architecture of the Intergrid makes rising the default rather than the exception.

And if we build it right the next layer builds on something solid. Interplanetary infrastructure, orbital compute, the edge network that does not yet exist, all of it runs on a foundation designed for the demands of this century rather than the last one. If we get it wrong we will power our data centers with satellites, call the debris field progress, and wonder why the lights keep going out.

---

The Ask.

Section Six

Every great American infrastructure moment followed the same arc. History presented an existential threat. The moment was met with investment. And what got built did not just outlast the threat. It helped defeat it.

The Depression and the Dust Bowl broke rural America. FDR met the moment with rural electrification and the economic multiplier helped end the crisis that justified the investment. The Soviet threat made American vulnerability undeniable. Eisenhower met the moment with the interstate and the prosperity and military mobility it generated helped win the Cold War. The nuclear threat demanded a communications infrastructure that could survive a first strike. The Pentagon met the moment with ARPANET and the open architecture that emerged made centralized authoritarian control of knowledge increasingly untenable everywhere it spread.

Three times last century. Same arc every time. Existential threat. Investment. Defeat of the threat and prosperity on top of it.

The moments do not wait.

AI, China, the data center boom. The Intergrid moment is now.

The electron gap and the DeepSeek moment are creating the urgency. The wealth concentration is creating the political pressure. The workforce displacement is creating the coalition. And whether American AI leads with freedom and equality or gets overrun by an infrastructure layer built on different values entirely is creating the existential threat.

Every element that broke the logjam on the interstate, rural electrification, and the internet is present simultaneously in 2026 in a way it has not been before and may not be again.

The question is not whether the Intergrid gets built. The private sector is already building pieces of it for themselves. The question is whether it gets built as shared public infrastructure that lifts every boat or as a collection of private toll roads that serve the yachts and leave everyone else on the dirt road with an above-ground power line and a higher electric bill.

That choice is being made right now. In utility commission rate cases. In interconnection approval queues. In congressional appropriations committees. In local planning commission hearings in townships nobody has ever heard of. The absence of a framework is itself a decision. And the decision the market makes without a framework is extraction.

Here is the framework.

These seven asks are a framework. Each one is buildable within existing regulatory and legislative structures. None require a new agency, a new tax, or a new ideology. They require a decision that the Intergrid is America’s fourth public infrastructure moment and that the moment is now.

Look at what is converging simultaneously. China outbuilding us eight to one on power capacity and purpose-building the infrastructure layer that encodes whose values run the AI economy. State actors who have already mapped the vulnerabilities of a grid held together with seventy-year-old control systems and duct tape. A handful of corporations building private power networks on public commons, eliminating hundreds of thousands of jobs in the same quarter they add trillions in market capitalization, and running betting pools on who builds the first billion dollar company with a single human employee. Sixty-five million ratepayers absorbing the cost of all of it on their monthly electric bill while politicians on both sides perform for their preferred audience on a platform one of which is owned by the country outbuilding us.

These are not four separate problems. They are one problem with four faces. The missing infrastructure layer that leaves America vulnerable to all of them simultaneously is the same infrastructure layer that addresses all of them if built correctly. The Intergrid is not just an energy policy. It is the answer to the Venn diagram.

The window is open. Every forcing function is aligned. Every political coalition is available. History does not grade on a curve. It records what happened and who was standing there when the moment was right there in front of them.

---

Energy Disclosure

This article was researched and drafted through an extended AI-assisted session involving hundreds of exchanges, multiple full web search passes, and iterative drafting across six sections. Standard single-article AI energy estimates of 0.5 to 0.8 kWh do not apply here. A more honest estimate for a session of this length and complexity runs between 5 and 15 kWh, equivalent to running a 100-watt bulb for 50 to 150 hours. Anthropic does not publish per-session energy figures, which is itself an argument for the transparency standards this article advocates. The irony is noted.

---

References

Grid Infrastructure and Power Demand

1. IEEFA. “Projected Data Center Growth Spurs PJM Capacity Prices by Factor of 10.” July 30, 2025. ieefa.org

2. Introl. “PJM Rate Shock: Data Centers vs Ratepayers.” February 28, 2026. introl.com

3. NRDC. “Rising Demand from Data Centers Driving Reliability, Cost Concerns.” November 5, 2025. nrdc.org

4. Utility Dive. “Data Centers Were 40% of PJM Capacity Costs in Last Auction.” January 7, 2026. utilitydive.com

5. Canary Media. “PJM Capacity Costs Hit Record as Grid Falls Short on Supply.” December 18, 2025. canarymedia.com

6. CoBank. “Patchwork or Partnership? 2026 Will Prove a Pivotal Year for Grid Governance.” December 10, 2025. cobank.com

7. PJM Inside Lines. “PJM’s Updated 20-Year Forecast.” January 14, 2026. insidelines.pjm.com

AI Data Center Power Demand Projections

8. Goldman Sachs. “US Data Center Power Demand Projected to Double by 2027.” May 2026. goldmansachs.com

9. Brookings Institution. “Global Energy Demands Within the AI Regulatory Landscape.” April 21, 2026. brookings.edu

10. RAND Corporation. “AI’s Power Requirements Under Exponential Growth.” January 28, 2025. rand.org

11. Visual Capitalist. “Chart: The Shortage of US Data Center Capacity 2023 to 2028.” December 16, 2025. visualcapitalist.com

12. Deloitte. “Can US Infrastructure Keep Up With the AI Economy?” December 24, 2025. deloitte.com

The China Competitive Gap

13. Searchlight Institute. “Seizing the Data Center Buildout for Grid Modernization.” March 9, 2026. searchlightinstitute.org

14. Energy Grid Modernization 2026. “Grid Modernization and National Competitiveness.” January 16, 2026. teamsilverline.com

15. The Hill. “In the AI Era, Infrastructure Is Destiny.” October 24, 2024. thehill.com

Wealth Concentration and Workforce Displacement

16. PYMNTS. “The One-Person Billion-Dollar Company Is Here.” April 3, 2026. pymnts.com

17. Future Digest. “The One-Person Billion-Dollar Company Is No Longer a Prediction.” April 8, 2026. futuredigestnews.substack.com

18. Crunchbase News. “Q1 2026 Shatters Venture Funding Records.” April 1, 2026. crunchbase.com

19. BuildMVPFast. “One-Person Unicorn: AI Agents, Solo Founder, Billion Dollar 2026.” May 2026. buildmvpfast.com

Generation Economics and Build Timelines

20. Lazard. “Levelized Cost of Energy Analysis 2025.” Referenced via SolarInfoPath. April 21, 2026. solarinfopath.com

21. SEIA. “Solar Market Insight Report 2025 Year in Review.” March 12, 2026. seia.org

22. EIA. “Annual Energy Outlook 2026: Electricity Market Module.” April 3, 2026. eia.gov

23. Carbon Credits. “What Is SMR? The Ultimate Guide to Small Modular Reactors.” February 18, 2025. carboncredits.com

24. Enerdata. “Small Modular Reactors: Advancing Nuclear Power Generation.” June 20, 2024. enerdata.net

Hyperscaler Off-Grid Trend

25. Data Center Frontier. “How the White House’s Executive Order and AI Action Plan Might Reshape US Data Center Development.” 2025. datacenterfrontier.com

26. Bipartisan Policy Center. “Strategic Federal Actions Aim to Strengthen AI and Energy Infrastructure.” January 23, 2026. bipartisanpolicy.org

27. Data Center Knowledge. “2026 Predictions: AI Sparks Data Center Power Revolution.” April 17, 2026. datacenterknowledge.com

Historical Context

28. General historical record. Federal Aid Highway Act of 1956 and National Defense Highways. Rural Electrification Act of 1936. ARPANET funding 1969. Standard historical record confirmed across multiple academic and government sources.

---

All URLs confirmed active as of June 2026. DarkAIDefense.com maintains editorial independence and received no compensation from any organization referenced in this article.