You’ve probably heard the AI bubble  is popping. That it’s all hype… a gimmick… and the money is drying up. That’s not what’s happening. What’s actually going on is a lot more serious…  and almost nobody is talking about it.

The AI revolution didn’t hit a  financial wall. It hit a physical one.

Right now, roughly 50% of American data center  projects in major hubs are delayed or quietly canceled. Not because companies ran out of cash. Because they ran out of power.

Now, Big Tech isn’t just making  software anymore. They've become some of the most aggressive investors  in one industry you wouldn’t expect… Nuclear power. This is the Limitless AI Lie.

Chapter 1 - The Ghost Clusters We are currently in the midst of a global artificial intelligence arms  race and it’s getting very expensive.

In 2025, The Wall Street Journal reported that  just in the United States, the Big Four – Amazon, Google, Meta, Microsoft – were collectively  spending close to $400 billion annually, with most of that money going on  infrastructure to run giant AI data centers.

These companies are taking on heavy debt, but as  The Journal pointed out, the “current revenue” from AI was, and still is, “relatively tiny.” Even so, the reports of unprecedented spending just kept coming. There’s a reason. And it’s why OpenAI, Oracle, and SoftBank announced they would  spend just under half a trillion dollars on the Stargate initiative. They’re rolling out  five new data centers in Texas, New Mexico, Ohio and the Midwest. CEO Sam Altman promised  that the investment would lead to future breakthroughs in technology, adding that they  are already making, quote, “historic progress.” Big Tech and every other global enterprise  with a focus on AI are expected to spend $7 trillion on building or updating  data centers in the next 5 years.

That’s the plan, but reality  shows a very different picture.

Data centers aren’t easy to run. Of the almost 11,000 active ones in the world in 2026, the US had the lion's share  at around 4,000, well ahead of the UK (511), Germany (507), China (368), and France (344). But just to run the ones we already have is extraordinarily energy-intensive. Cooling them  can take vast amounts of water, with the biggest hyperscale sites using millions of gallons a day. Still, heat can be managed.

The problem is powering them. Reports in 2026 stated that 50% of data centers being constructed in the US  are currently either canceled or delayed, with experts saying that number is going to rise. In major hubs like Northern Virginia – the region known as “Data Center Alley” – projects are  already slowing as the grid starts to buckle under demand. The public sentiment  has changed. Polls show just 35% of Virginia voters give their thumbs up to the  power-hungry projects compared to 69% in 2023.

The delays aren’t about a lack of money  or ambition. Silicon Valley has plenty of both. They’re about supply chains,  engineering limits, and public/political resistance…but most of all, power. The future Sam Altman talks about?

It’s sitting in a queue… waiting for electricity. Right now, America is facing a bottleneck caused by unprecedented demand. So many  of the data center projects are in a traffic jam, waiting to connect. Before anything can plug in, utilities and grid operators must run studies to see whether  lines, substations and transformers can cope with the extra load. They have to assess reliability.  They have to look at the environmental impact.

This can cost millions of dollars  and lead to years of waiting around.

It doesn’t just affect Big Tech. It affects you.

These massive data centers burn through enormous  amounts of electricity. When their energy bills spike, that cost doesn’t just disappear. It  gets passed on. To consumers. To communities.

And when people start seeing higher energy  prices, they push back. That can be lethal to the ambitions of anyone wanting to build  a data center. And even if the projects get past all those hurdles, the physical  system might still not be ready to go.

According to Lawrence Berkeley National  Laboratory, these queues add up to about 2,600 gigawatts of proposed generation  and storage capacity. That’s about twice the size of the current U.S. power system. It’s a backlog as big as the grid itself.

The amount of power these new data centers  will require to keep running is astronomical.

Some of the new ones in the US will use 50 to  100 times more power than it takes to power the Empire State Building… just to run and cool the  servers. Reports say the biggest ones currently under construction will collectively demand as  much electricity as an entire country like Italy.

Big tech wants to spend trillions, but  the “power crunch” is getting in the way.

It’s a problem they can’t code around. Chapter 2 - The Physics of the Bubble On average, a single query to ChatGPT uses  around 10 times more energy than a search on Google. To put another way, one complex  prompt can use about as much electricity as leaving an LED lightbulb on20 minutes. The exact number is debated. And not every query is the same. But the scale is. Because  ChatGPT isn’t handling thousands of requests… It’s handling billions. 2.5 billion queries per day, it all adds up. The US alone experienced about  330 million ChatGPT queries per day in 2025 That’s just the beginning. Recent reports tell us about 1.1 billion people worldwide use AI but that  number is growing fast. 56% of Americans use it in their day to day lives. About 60% of American  teachers have reported experimenting with it.

88% of global firms are now onboard, too. The experts say there are still about 5 billion people globally who don’t use it,  but the adoption curve is rising fast. In the next 5 or 10 years, we can expect many  of those 5 billion will be regular users.

So as more and more people  start experimenting with it, the world will need to generate a lot more power. Goldman Sachs Research analysts have estimated that global data center power demand will grow  160% by 2030. That’s about the equivalent of adding a second 'Germany' to the global energy  grid in just 6 years. Data centers already consume about 1-2% of all the world’s electricity.  If the experts are right, it will be more like 3-4% by the time the decade is over. Where will all this new power come from?

In 2023 alone, new applications for  interconnections surged by about 30%. The data showed that close to 95% of requested capacity  came from solar, wind and battery storage.

But here’s the problem. Renewables aren’t constant.

The wind slows down. The sun disappears. Even in  the best solar regions in the U.S., you only get about 3,500 to 4,000 hours of usable sunlight a  year. That's out of almost 8,800 hours in total.

Less than half the time. In other words, solar and wind won’t be enough to power the AI revolution. Companies are now fighting for whatever power they can get, and whoever controls  the power, controls the future of AI.

Chapter 3 - The Great AI Cleansing The delays in data centers look like a failure of infrastructure. But they may  also be creating something else: scarcity.

And scarcity, for the companies with  all the money, isn’t always a problem.

It can actually be a very real advantage. The biggest tech firms are already making deals with energy companies. Meta,  Google, Amazon, Microsoft, have inked deals with utilities firms all over the USA. They’ll get the power. The smaller players won’t.

Venture-backed startups might be able to raise  money and promise major market disruption, but they won’t be able to compete  for energy. They’ll have to rely on Big Tech for renting and compute power. That could mean that every GPU shortage, every power delay, every price rise will hit  them first. And the state-of-the-art models might become prohibitively expensive or limited  in access. The biggest tech companies will decide who gets access first, who gets the most  computing power, and who gets the best versions.

Meanwhile the market itself is changing. Investors  are losing interest in generic chatbots that write poems and hallucinate history. Capital is  moving toward specialized AI that can write code, optimize logistics, detect fraud, perform medical  analysis, and above all, increase profit margins.

But not everyone will receive the same treatment,  and some might be left waiting. Countries have anticompetition laws, but the gatekeepers of  technology as they consolidate their power over AI, might move faster than the law can keep up. A  specialized startup might not ever truly be able to compete, not if it's denied access. In the gold rush, fortunes weren’t made by the miners, but by those who  controlled the tools and the land.

This time, it isn’t just the tools. It’s the electricity that powers them.

For years now, people have been worrying about the  rise of the machines, rogue AI ending the world, but that’s all science fiction. The  real danger is a handful of monopoly powers who control all the key infrastructure. But to guarantee their monopoly, Big Tech knows it can’t just rely on public infrastructure. So it’s planning to go off-grid.

Chapter 4 - The Nuclear Pivot For decades, tech companies rented power from the grid like everyone else. Now, that’s changing. These days some of the largest firms with seemingly bottomless reserves  of money are getting into the business of procuring their own energy assets. They’re not just software companies anymore.

In 2024, Microsoft made a deal to reactivate Three  Mile Island nuclear power plant, a site forever associated with a 1979 meltdown. One reactor was  shut down permanently. The other ran until 2019, when it was closed due to operating losses. Now, decades later, it’s being revived.

Not for history, but for demand. Under a 20-year power agreement, Microsoft is securing 835 megawatts of  electricity for Pennsylvania’s grid.

The deal wasn’t about powering Pennsylvanians’  personal computers or domestic appliances, but achieving lift-off on its planned data centers. It’s enough electricity to power about 800,000 American homes, and all of  it just for Microsoft's AI.

But will it happen? Well, firstly, a nuclear plant has never in American history  been fully shut down and sprung back to life.

Experts are saying that for Microsoft’s plan  to come to fruition, there will be numerous regulatory hurdles to cross. The permits are  all still pending, and they might not ever come, even if the plant owner, Constellation  Energy, is trying to speed things up.

Nuclear power tends to make people uneasy. Even if  the plant did get a second lease of life, it’s not clear it could be integrated back into the grid. Mark Zuckerberg’s Meta also hit a problem when trying to build a data center next to what was  reported as an “existing nuclear facility.” The media didn’t name the facility but it did  report that Zuckerberg’s plan of operating a nuclear-powered AI center crashed to the ground  because of…bees. A rare species populated the area where Meta wanted to build. It goes to show how Big Tech is thinking outside of the box. Amazon has tried, too.

In 2024, the company signed a $650 million  deal with Talen Energy in Pennsylvania where it plans to open a number of data centers close  to the Susquehanna nuclear power plant. The deal, which is set to end in 2042, could  be delivering almost double the power needed for the city of Pittsburgh. That’s if it ever gets off the ground.

In 2024, the Federal Energy Regulatory  Commission blocked Amazon from taking power directly from a nuclear plant to  feed one of its hyperscale data centers.

The reason? Diverting that much power raised the risk of blackouts for  the communities already relying on that same nuclear supply.And even if the lights stayed  on, there was another consequence… higher bills.

The regulators are standing firm.  Amazon later asked for a re-hearing and the result stayed the same. Google is also looking to nuclear energy to power its future AI ambitions. It’s one of a few big tech firms currently investing in next-generation nuclear energy  in the form of small modular reactors (SMRs).

Smaller means less cost to build. It also means  cooler, so ideally, less chance of a meltdown.

The company hopes to have the data center up and  running by 2030. But long before any green lights start flashing, Google will need approval from  the U.S. Nuclear Regulatory Commission (NRC).

And that won’t come easy. As one analyst wrote, the “path to commercial viability is still riddled  with uncertainty.” SMR designs haven’t been proven at scale and regulatory hurdles have set a high  bar. With SMRs, there’s no easy path forward.

But there’s political will, and that’s important. The current administration has given the thumbs up and has said it wants these small  reactors to be functioning by 2030.

None of the new commercial SMRs are yet operating  in the US, but China and Russia already have working examples. In terms of the global AI  race, this isn’t good news for Uncle Sam.

But even if the tech firms get what they  want, the SMRs won’t solve all the power demands of AI by themselves. They’ll rearrange the  playing field but they won’t be a magic bullet.

They’ll rearrange something else, too. How we, the public, view the very nature of what a tech company is. They don’t just make products to help you be productive at work or search for a good place  to eat dinner anymore. They’re actively trying to become a player in critical US infrastructure. And once they have that kind of power, they’re not going to let go of it. Chapter 5 - The Sovereign Cloud Tech giants are not just buying up the  existing spaces on the Monopoly board; they’re creating new ones for themselves. But not every country in the world will have nuclear power. They won’t all have abundant resources, or a grid capable of supporting large-scale AI  infrastructure. So, if nations want to remain technologically competitive, many might find  themselves relying on cloud platforms and compute capacity controlled by foreign firms. What does that mean?

It means power. Political power.

That dependence nations might have on private  companies could shift global power dynamics in profound ways. Influence would no longer just rest  with governments, militaries, or central banks, but with corporations controlling energy  contracts and digital systems that modern economies will need to keep functioning. Businesses started in someone’s basement or garage in Silicon Valley are becoming  as powerful as sovereign states . Private infrastructure powers with the ability to shape  markets, policy, and holding the keys to access the tools needed to be a modern economy. CEOs from Microsoft, Tesla, Amazon, Google, Meta and others are now regular features at the  White House. They are consulted on AI policy, cybersecurity, competition, censorship,  national security and economic strategy. In earlier periods of history, hidden power and its  invisible hands was often said to be intelligence agencies, media tycoons, shadowy financiers, or  military-industrial complex defense contractors.

Today, Big Tech can be added to the list. As Carnegie Europe said in 2025, billionaire tech titans quote, “now  rival nation-states in influence, shaping the rules of the global digital order.” Their technologies, and increasingly AI, dictate how we communicate, how we trade with each  other. As the recent “all lawful use” argument between Anthropic and the US government shows,  they also play a major role in how we wage war.

Political realists often write that the global  balance of power largely relies on the size of an economy and its military. Well, we can now add  the size of data centers to that equation, too.

In the years ahead, real power may belong  not only to those who govern nations, but to those who own the electricity and computing  power behind artificial intelligence. Some of the biggest firms are not just richer than many  small countries. In the future, they might well have more influence on how the world is run. For some critics, that’s a major concern.

Chapter 6 - The Real Revolution The real power of AI isn’t what it might seem at first sight. Products that make deepfake videos of a pop star or produce videos of  the U.S. declaring war on Great Britain.

No, the power is in all the processes you  don’t see. The grease that will oil the wheels and cogs of the future global economy.  And it happens mostly behind the scenes receiving very little credit for its genius. That is, agentic AI systems for complex tasks used by B2B companies. The focus now is on  Multi-Agent Orchestration, where teams of AI agents working together can handle anything  from workflows to financial risk management.

These systems don’t just answer questions, they  collaborate with staff and optimize procedures at every step of the way. Current reports are  saying the most effective ones can improve efficiency by a massive 30% to 50%. When something breaks, they flag it, and that’s when the human overseers step in. The goal is to save time, save money, and on a scale that is changing entire industries. Think about global logistics.

Some of the biggest logistics firms can lose huge  amounts of money through delays and empty trucks.

If AI cuts waste by just 2%, that still could  be worth billions. The proof? Recent reports say the market for AI in logistics reached $12  billion in 2026, up from $8.2 billion in 2024.

The reason was automation. Not replacing people in warehouses with clever robots, but adding an AI layer  to operations that streamlined many of the processes. Industry reports say in just two years  AI has cut down on operational costs by 20-30%.

Drug discovery and biotech are also expected to  experience an AI windfall. The cost of bringing a successful drug to market can run into billions  and take years. But if AI can help scientists better understand how diseases work, design better  drug compounds and run smarter clinical trials, it could completely change the very  big business of sickness and health.

The potential is huge. But potential right now is the operative word.

Countless analyst reports tell us AI will make  drug companies billions over the next 10 years or so, but it’s not nailed on. If the predictions  are correct, the pharmaceutical industry stands to increase operating profits by more than 10%. The first AI boom entertained us. We’ve had a lot of fun. But all the hype behind it and  overvalued companies will probably soon die out.

What’s happened is a cleansing. When the gimmicks  have been washed away, the serious money will be invested in specialized B2B applications  in just about all the major industries.

Industries that will rely on a handful of  companies with all the biggest data centers.

This brings us to the terrifying  evolution of the tech industry.

Chapter 7 - The Private Empire The largest technology firms are no longer just software vendors. They run global cloud networks that make the world’s economy tick over. They maintain private cyber-defense divisions, armies of the future. They issue cloud  credits that function like economic incentives inside their own ecosystems.  It’s a kind of currency in their digital kingdoms that businesses will need to function. And as you’ve seen, they’re moving into power itself, securing nuclear contracts, gas generation  and dedicated electricity supplies for data centers. They are effectively ending the nation  state’s monopoly on critical infrastructure.

Some of those data centers might one day  stand empty, graffitied and broken. A monument to investor overconfidence abandoned  before a single server was even switched on.

But Big Tech will by then have already  consolidated its power because it secured all the energy deals. The speculative frenzy will burn itself out. Weaker players will  collapse. But then what remains is infrastructure concentrated in fewer hands. The AI boom won’t be remembered as a passing mania people laugh at. It will be remembered  as the moment a handful of companies secured control over compute, energy, and the digital  infrastructure everything else depends on.

Not platforms. Not products. Infrastructure.

Private empires, positioned underneath  the next era of technology, charging rent on everything built above them. But empires don’t just run on power.

They run on cash. What happens when the money  required to sustain it starts to run out? Find out in ‘$115 Billion Burn Rate. The AI Bubble  Just POPPED’. Or click on this video instead.