Transcription
AI has become one of the biggest conversations on the planet, especially in the United States. But almost nobody talks about the real cost behind that one AI-generated image or conversation. Every model being trained is burning through power the American grid was never designed to handle. And now, demand is exploding. The United States needs more electricity than ever, and it needs it to be reliable, clean, and permanent. Every state in America is scrambling to solve that problem. But one state might have already cracked it. It just cost $35 billion and nearly broke the company that built it.
Today, the newest AI facilities are pushing 1,000 megawatts. Enough to fuel a small city. And that’s just one building. There are now dozens of these AI campuses scattered across the U.S., with even bigger ones already under construction. By 2030, the Department of Energy expects data centers will eat up to 17% of all electricity made in the United States. It’s pretty simple. The grid wasn't built for this. What the AI economy values most is power that never stops flowing. And they’re doing everything they can to get it. In 2024, Microsoft signed a 20-year deal with Constellation Energy to restart Three Mile Island Unit 1. Amazon went shopping for entire power plants. Meta started paying reactor developers directly. Google joined them. These are some of the richest companies on Earth, and even they cannot get the electricity they need fast enough. There’s a lot of money riding on this. Hyperscale tech spending will top $600 billion in 2026. And every dollar of it is useless without a socket that delivers power on demand.
One state had it all figured out. Georgia switched on two nuclear reactors at Plant Vogtle in 2023 and 2024, turning it into the largest nuclear plant in the United States. A single site powerful enough to run millions of homes, nonstop. Carbon-free and built to run for 60 to 80 years without a break. That's the kind of electricity system the AI economy is now desperately searching for. And right now, Georgia is the only place in the country that’s actually doing it. But getting there was never really about engineering. It was about everything that made new nuclear power basically impossible for decades.
To understand why Georgia is one of the only places in America still capable of pulling this off, you have to go back to the moment the country forgot how to build nuclear power. Three Mile Island didn’t kill American nuclear power; it buried it in paperwork. After the meltdown in 1979, every new reactor was trapped in a permit process so heavy it could take years just to break ground. Then Chernobyl made it political. Public support collapsed, utilities backed out, and nuclear power stalled. Natural gas became the escape hatch. The shale boom flooded the U.S. with cheap fuel, and gas plants could be built in under 2 years instead of a decade. Nuclear wasn’t just slow; it was the worst financial bet in energy. The factories that once made reactor pressure vessels shut down. Engineers who could seal containment domes left the industry. America didn’t just stop building reactors; it lost the skills to build them. For a generation, the U.S. didn’t complete a single new reactor. Every state that tried to bring nuclear back hit the same problem. South Carolina tried. Florida flirted with the idea. The TVA studied site after site. Each one ran into costs that doubled, then tripled. America had to relearn a skill it had spent thirty years forgetting.
Georgia inherited all of those problems when it broke ground at Vogtle in 2009. So why did America bet its nuclear future on Georgia? Because someone had to pay the cost of relearning how to build a reactor from scratch, under modern rules, with a workforce that barely existed anymore. Georgia’s Vogtle Units 3 and 4 don’t just produce huge amounts of electricity. They represent an entirely different generation of nuclear reactors. The older reactors built in the 1970s and 80s depended heavily on pumps and backup systems to keep the core cool during an emergency. The new AP1000 reactor was designed around a far more terrifying question: What happens if everything fails? A total blackout. The kind of situation where operators have minutes to get systems back online. That exact nightmare is what caused Fukushima. When the plant lost outside power, cooling systems failed, and within days, 3 reactors melted down. The AP1000 was built so that scenario theoretically can’t happen. A massive water tank sits above the reactor holding hundreds of thousands of gallons. If everything else dies, the system opens automatically and gravity takes over. Water floods the core with no electricity or pumps. As heat rises, steam hits the inside of the steel containment shell, cools, condenses, and falls back down again. It creates a self-sustaining cooling cycle powered entirely by physics. Nothing has to switch on because the system was designed to work without anything switching on in the first place.
That one feature cut out thousands of valves, pumps, and miles of cable that older reactors needed. The plant has about half the parts of a comparable older reactor. It takes up less space. It can be built in pieces in a factory and bolted together on site. On paper, all of that was supposed to make the AP1000 cheaper and faster to build than anything before it. In real life, Unit 3 has been running above 96% of its maximum output, meaning it’s producing power almost constantly. For a brand-new reactor, that’s impressive. Most new plants spend years dealing with breakdowns, shutdowns, and unexpected problems. Vogtle came online and… kept running. For utilities, that consistency is everything. If a reactor can deliver power day and night without stopping, companies can build entire long-term contracts around it with confidence.
In April 2026, Westinghouse, the company behind the AP1000 reactor design, asked the Nuclear Regulatory Commission to make Vogtle Unit 4 the template for America’s nuclear future. Every new AP1000 license, blueprint, and construction plan would trace back to what was built in Georgia. And that matters. There are only 6 AP1000 reactors operating on Earth today. The entire Western world only has the two reactors in Georgia. What happens there over the next decade could decide whether America builds a new generation of nuclear plants… or never tries again.
But getting to this point wasn’t easy. The original budget was about $14 billion. The final number, when every spreadsheet was settled, hit $36.8 billion. The project ran 7 years late and roughly $21 billion over budget. By any normal logic, the project should have been killed at least three times. The first near-death came in March 2017, when Westinghouse filed for bankruptcy halfway through construction. It had signed fixed-price contracts that nobody could actually deliver on, because the work was harder than anyone admitted. The company collapsed under the weight of its own promises. Its parent company, Toshiba, took a $9 billion writedown and quit the American nuclear business. Georgia Power and Southern Company didn't have a contractor anymore. They had a half-built nuclear plant, thousands of workers on site, and no one to manage them. The state could have shut the project down. Instead, Southern Nuclear took over project management and brought in Bechtel, the same firm that built the original Vogtle Units 1 and 2 in the 1980s, to run day-to-day construction.
The cost overruns weren't only about the bankruptcy. The promised factory-built parts, where chunks of reactor would be made off site and shipped in, fell apart almost immediately. Workers had to rebuild parts on site, killing the whole money-saving idea behind the design. On top of that, the project burned through what little American nuclear workforce was left. The plant was forced to train almost 13,000 technicians as it went. That training number tells the real story. America didn't just build a power plant in Georgia. It rebuilt a workforce. Burke County had to set up trade programs for a generation of workers who had never even seen a reactor under construction. Local welders had to be upgraded to nuclear-grade certifications. Engineering firms shipped staff in from Tennessee, South Carolina, and Florida just to keep up. More than 30% of Vogtle hires were military veterans, many of them coming straight from the Navy's nuclear program, which had kept building submarine and aircraft carrier reactors. Every one of those workers now carries that training into the next reactor America tries to build, wherever it ends up.
Before Vogtle's expansion broke ground, Burke County was one of the poorest counties in Georgia. Unemployment ran above the state average for most of the 2000s. That all changed. By the peak of construction, Vogtle was the largest private employer in the county by a distance. Burke County's property tax digest grew significantly as construction infrastructure, worker housing, and supporting businesses followed the project in. The 13,000 workers trained at Vogtle built careers that would not have existed in Burke County without it.
Then came the bills. Georgia Power passed costs onto its customers through a mechanism called "construction work in progress," which is exactly what it sounds like. Customers pay for the plant before it produces a single watt. Monthly residential bills rose when Unit 3 came online in 2023, and climbed again as Unit 4 followed. Over time the typical bill went up around 10%. The Georgia Public Service Commission, the elected body that regulates utilities, approved almost $7.5 billion in construction costs to be passed directly to customers, with shareholders absorbing the rest. A natural gas plant of the same size would have cost roughly one-tenth as much. Solar would have been cheaper still. Based on those numbers, the smart decision was always to drop the project. The Public Service Commission also found that Georgia Power repeatedly gave regulators inaccurate cost estimates over the years. Numbers that made the project look more viable than it was. The honest figures would have buried it. The criticism wasn't theoretical. Regulators argued the project was being kept alive by projections that no longer matched reality.
So was Vogtle a disaster of bad management? Or was it the unavoidable price of relearning how to build something the country had thrown away? It's both. The bad management is on the record. But so is the larger reality that America was trying to kickstart an industry it had effectively abandoned. Any restart was always going to absorb costs the original estimates failed to capture. The country needed a finished reactor. Somebody had to pay for the lesson. Georgia signed the check.
That's why America is going all in on Georgia. Every other state looked at the bill and said no. Georgia said yes, complaints and all. Once the reactors started spinning, the conversation moved from "what did this cost" to "what does this make possible." Plant Vogtle is the most expensive power project ever finished in the Western hemisphere. But the project did something no spreadsheet could fully measure. It proved the United States could still finish a large nuclear reactor at all. $36.8 billion bought two reactors. But it also rebuilt supply chains, trained labor, and established a working template future projects can now reference instead of inventing from scratch. Large scale nuclear construction has struggled across most of the world. Georgia became one of the few places that actually pushed a project all the way through to completion.
The best way to see what Georgia actually did is to look at the state that started the same project and gave up. About 200 miles (322 km) east of Plant Vogtle, on the banks of the Broad River near Jenkinsville, South Carolina, sits a concrete pit. It’s what’s left of V.C. Summer Units 2 and 3. South Carolina poured around $9 billion into it before pulling the plug in July 2017. The utilities behind it, SCANA and state-owned Santee Cooper, cancelled the project just months after Westinghouse collapsed. The reactors were never finished. The site was abandoned. The contractor was the same. The reactor design was the same. The bankruptcy hit both projects on the same day. So what was different about Georgia?
Georgia Power serves about 2.7 million customers. SCANA's South Carolina Electric & Gas served roughly 700,000. Spread the same multi-billion-dollar bill across four times as many ratepayers, and the per-household pain becomes politically survivable instead of fatal. The second difference was how the rules were written. Georgia Power is allowed to recover honest building costs from customers under a system where the Public Service Commission has near-total say. South Carolina had a similar setup on paper, but its rules around how much could be charged to customers were tighter. When Westinghouse fell apart, South Carolina's regulators didn't have the political cover to keep pushing forward.
The third difference was who owned the plant. Vogtle is co-owned by 4 entities. Georgia Power holds the largest share at 45.7%, while the rest is split among Oglethorpe Power, MEAG Power, and the city of Dalton. That structure spread the financial burden across rural co-ops, municipal utilities, and one major investor-owned company. So no single player carried the full risk. V.C. Summer was essentially a two-utility project, and one of them was state-owned. That made every cost overrun a political fight.
Then there’s the difference that mattered most. Georgia had already built two reactors at Plant Vogtle in the 1980s. The site was permitted, licensed, and tied into the grid. It was a brownfield expansion. V.C. Summer was much closer to a greenfield build, working off a site with only one older reactor for company.
But you also have to look at a harder to measure factor… the political nerve at the moment of crisis. When Westinghouse collapsed in March 2017, both states were forced into emergency hearings within weeks. South Carolina's politicians, looking at fewer customers and louder complaints, chose to cancel the project. Georgia's leadership went the other way. They decided that abandoning a half-finished reactor after spending billions would be even worse than absorbing the additional cost required to finish it. That decision is the moment Georgia’s path split from South Carolina’s. That gap is why no other state has been able to copy what Georgia did.
The bet on Georgia worked because Georgia was, structurally, the only state where it could have worked. Burke County had one advantage no other proposed site could match. The permits, the water rights, and the nuclear expertise from the 1980s build were already in place. When Georgia expanded Vogtle, the water question was never reopened. The licensing fight never happened. Every other state would have spent years in court before breaking ground.
Georgia’s link between utility and government also runs deeper than in many larger, more politically divided states. The Public Service Commission is elected statewide and historically produced commissioners more willing to back long-term utility expansion. On top of that, the legislature consistently reinforced the project through tax rules, financing laws, and policies that allowed costs to be recovered while construction was still ongoing. States with more fragmented regulatory systems would have faced years of additional legal and political resistance before construction ever reached this stage.
Compare that to California or New York. Both have nuclear histories - Diablo Canyon in California and Indian Point in New York - but today both are defined more by pressure to shut plants down than to build new ones. Neither state could have hosted Vogtle. Both punish utilities for proposing exactly the kind of project Georgia executed. Illinois is more complicated. It runs the largest nuclear fleet in the country with 11 reactors. But every one was built before construction costs exploded in the 1980s. Its deregulated market makes new nuclear power almost impossible. Utilities can’t guarantee the long-term pricing needed to recover massive upfront costs. That’s the trap: deregulation rewards what already exists, but blocks what comes next. Which is why nuclear expansion has clustered in regulated Southeast states instead of the big nuclear states up north. Only Georgia had the conditions to drag one across the finish line.
And it’s only going to get bigger. In December 2025, Georgia Power filed a $16 billion plan to build enough new electricity generation by 2031 to power several million homes at once. The Public Service Commission spent nearly 25 hours in public hearings just on that proposal. The plan was driven by a single category of customer that did not exist at a meaningful scale five years ago.
In January 2025, Amazon committed another $11 billion to data center buildouts in Georgia. Not future intent, but already deployed capital. Earlier that same year, it had snapped up huge parcels of land in the state for more campus construction. And it’s not alone. Microsoft, Meta, and Google all maintain major footprints in Georgia, with Atlanta-area facilities that consume more electricity than mid-sized industrial plants. The region has turned into a hyperscale cluster. Meta’s facility in Newton County makes the scale visible. One campus, already under expansion, pulls roughly 500,000 gallons of water a day - around a tenth of the county’s daily usage - just to support its computing load. And even that is no longer enough. Newton County is now reviewing multiple new proposals, including data centers asking for water and power at even larger scales. Every one of these projects is competing for the same thing: electricity the U.S. grid is struggling to guarantee.
The Public Service Commission’s December 2025 ruling is what made the expansion real. Georgia Power committed to building a massive wave of new electricity capacity specifically for incoming industrial and data center demand. Crucially, it agreed to absorb much of the financial risk itself if some projects didn’t materialize. That commitment changes the risk calculation for every company deciding where to build next. Utilities don’t gamble billions unless they already know demand is coming. Which means the AI and data center boom heading toward Georgia is likely much larger than what’s been publicly announced so far.
Atlanta already has some of the best fiber connectivity in the country thanks to its telecom history. Land outside the city is still relatively affordable. The climate helps reduce cooling costs. But the biggest advantage is power. Georgia can offer large amounts of reliable, always-on electricity backed by the biggest nuclear plant in the country. Georgia is now writing big-load tariffs, rules that govern how data center customers buy power, in ways that let hyperscalers line up dedicated clean generation alongside grid power. That setup effectively allows companies like Microsoft or Amazon to plug straight into the nuclear power Georgia just finished building.
Here's where the timing gets strange. Vogtle Unit 4 entered commercial operation in April 2024. AI build spending hit a wall of growth that exact same year. Two completely separate timelines. A 15-year construction project and the runaway growth of AI computing, met in the same window. Georgia happened to finish the only new American reactors of the decade right when the country's largest companies discovered they couldn't get reliable electricity anywhere else.
Other states figured out too late what was happening. Virginia, the historic capital of American data centers, hit its grid limit before it could add new generation. Texas, with its independent grid, has become a magnet for natural-gas-fueled data centers but can't offer the carbon-free baseload that big tech. Arizona has space and sun but no firm baseload to back it up overnight. In Northern Virginia, the world’s largest data center hub, PJM Interconnection was forced to pause new project reviews just to clear the backlog. By the time a tech giant finishes the paperwork to connect a new campus to the grid, a competitor in another state has already built and powered up three. Georgia’s advantage is time-limited, but it currently controls one of the only fully operational large-scale clean baseload systems in the United States.
And competition isn't just other states. Every month the country fails to add new firm baseload, hyperscale AI campuses either get built somewhere else, increasingly overseas, or they don't get built at all. Both outcomes hand the AI economy to whoever else can supply the power. Right now that's Saudi Arabia, the United Arab Emirates, and certain provinces of China. In the UAE, partnerships like G42 and Microsoft are fast-tracked because power is allocated upfront through state-linked utilities. Land is already allocated aside for data centers, approvals move through a single coordinated system, and demand for compute is often locked in before construction even begins. That removes the usual delays that slow hyperscale buildouts elsewhere. Every gigawatt Georgia delivers is a gigawatt those countries can't sell to American tech companies looking for a home. The state that lived through the financial agony now controls the one product the entire AI economy is desperate to buy.
And now every utility in the country is watching what happens next at Vogtle. Not what happened during construction, that story is over, but what happens during operation. If the units run above 90% capacity for the next decade with normal maintenance costs, the AP1000 design becomes the proof that restarts large-scale American nuclear building. If reliability drops, fuel costs spike, or major repairs surface, the case against new builds hardens for good.
So far the evidence is strong. According to the Department of Energy's Loan Programs Office, Unit 4's construction phase actually ran 30% cheaper than Unit 3's, even though they were built side by side. It shows that learning was happening in real time. The same workforce that took 15 years to finish the first unit cut meaningful cost and time off the second one. PricewaterhouseCoopers projected that a fleet of 10 new AP1000 reactors could generate nearly $93 billion in U.S. economic output. It would also support tens of thousands of jobs over more than a decade of construction. And the biggest shift wouldn’t just be economic, it would be cost. Once the industry starts building reactors repeatedly, the construction costs could fall by roughly a third compared to Vogtle.
Whether that outcome materializes is the central question for U.S. nuclear expansion. If the next AP1000 lands meaningfully below Vogtle’s cost baseline, nuclear becomes competitive with gas at scale. If it does not, Vogtle becomes a ceiling rather than a stepping stone. There is precedent for both outcomes. Countries that built continuously reduced costs significantly. Countries that paused saw cost inflation persist. The key variable is repetition, not design alone.
And the idea of more nuclear reactors isn't theory any more. Westinghouse has announced plans for multiple AP1000 deployments in the United States with construction targets toward 2030. A new AP1000 project has already been proposed in South Carolina, the same area that previously abandoned the technology. Other states are also pursuing different nuclear pathways: small modular reactors, advanced fast reactors, and restart programs for retired plants.
Georgia's success is often told as a story of one state's political will. The financial reality is more complicated. The project survived the Westinghouse bankruptcy partly because the Department of Energy had issued an $8.3 billion federal loan guarantee in 2014 - the largest nuclear loan guarantee in DOE history. By the time the project closed out, the total federal credit exposure had grown to around $12 billion as the scope expanded. Without it, the financing would have fallen apart regardless of what Georgia's politicians decided. Washington bet that Georgia would finish. Georgia made sure Washington didn't lose.
The question now is whether the next ten AP1000 reactors can be financed the same way or whether the federal loan program scales fast enough to match what the industry is being asked to build. Instead of one-off nuclear builds, the strategy is now standardization and repetition. The federal target is now to roughly quadruple nuclear capacity by mid-century. That scale can’t be reached through bespoke builds. It depends on repeatable deployment of standardized designs. Vogtle is the first completed proof of that.
Georgia spent $36.8 billion and two decades of political capital to answer one question. Can America still build nuclear power? After bankruptcies, lawsuits, regulatory wars, customer revolts, and a workforce that had to be rebuilt from scratch, the answer is yes. Expensively. Barely. And only because one state was structurally set up to absorb damage that would have killed the project anywhere else.
The harder question is what comes next. Whether Vogtle represents a reset point that lowers future costs through repetition, or a structural ceiling that defines how expensive nuclear construction has become. The reactors are running. The skill is rebuilt. The reference plant is licensed. Every major electricity market in the United States is now testing its own version of the same question Georgia already paid to answer. The next phase of industrial power is already being decided in real time. The state that absorbed every cost the project created is now the only state positioned to capture every benefit the AI economy is willing to pay for. It completed the only large-scale working example the United States currently has of how to rebuild nuclear capability at all.
Thanks for joining me today. I hope you enjoyed the video, and if you want to see more videos like this, then please consider subscribing and turn on notifications. It really does make a difference. Thanks again, and I'll see you next time on Map Pack.