You have read every article in this series. You know about the dry wells. The spiked electricity bills. The farmland destroyed. The communities deceived. The wildlife wiped out. The children damaged. The elderly widow who can’t pay her bill.
All of it — every single story — has been built on one assumption: that the AI data center boom is real. That the demand is real. That the $700 billion being spent this year will actually generate returns.
Now three of the most powerful financial institutions in the world — Goldman Sachs, JPMorgan, and the International Monetary Fund — are saying something that nobody in Big Tech wants you to hear:
It might all be a bubble. And when it bursts — just like the dot-com crash of 2000, just like the housing crash of 2008 — the people who will suffer most are not the billionaires who built it.
It will be you.
THE WARNING SIGNS THAT ARE IMPOSSIBLE TO IGNOREMultiple major institutions — including Goldman Sachs, the IMF, and JPMorgan — issued simultaneous warnings about the AI bubble beginning in late 2025. The prevailing expert consensus among financial analysts is that the AI bubble is already showing major signs of strain — with smaller startups and pure-play AI companies at highest risk of collapsing first, while diversified tech giants may weather the downturn but face substantial write-downs and valuation drops. 
Goldman Sachs. JPMorgan. The IMF. All saying the same thing. At the same time. That is not a coincidence. That is a fire alarm.
All the classic signs of a speculative bubble are present and paying out simultaneously in the AI data center boom: steep valuations, overbuilt infrastructure, speculative investment, and artificially engineered growth. The industry has all the hallmarks of bubbles that have burst before — and left devastation in their wake.
Steep valuations. Overbuilt infrastructure. Speculative investment. Artificially engineered growth. That is the exact checklist that described the dot-com bubble in 1999 — one year before it collapsed and wiped out $5 trillion in wealth.
THE COMPANY AT THE CENTER OF THE BUBBLE — AND ITS TERRIFYING NUMBERS
CoreWeave — the AI data center company that went public in 2026 as the hottest IPO of the year — tells the story of the bubble perfectly. Despite explosive revenue growth from $16 million in 2022 to $1.9 billion in 2024, CoreWeave reported negative core earnings and was burning through cash at a rate that could only sustain operations for approximately 9 months as of its IPO filing. It carries $24.5 billion in total debt with $7.5 billion in interest payments due through the end of 2026. And most terrifyingly — 62% of CoreWeave’s revenue comes from a single customer: Microsoft. If Microsoft reduces its AI spending by even a fraction, CoreWeave collapses. 
$24.5 billion in debt. $7.5 billion in interest due this year. 62% of revenue from one company. Nine months of cash runway. This is the company that Wall Street valued at $23 billion at IPO. This is the company that represents the financial model of the AI data center boom.
According to Bloomberg, half of all proposed data centers in 2026 will face delays or cancellation — not because of community opposition, but because of financial reality. Components are not manufactured in the U.S. so contractors need to buy from Canada, Mexico, South Korea, and China — the very countries Trump’s tariffs have made more expensive. Interest rates on construction loans are eating up remaining cash. Corporations are scrambling for bond sales and bridge loans as timelines lag. Heavy users of AI tools are burning $500 to $2,000 per month each — costs that are proving unsustainable for most businesses.
$500 to $2,000 per month. Per user. For AI tools. Most businesses cannot sustain those costs. The moment businesses start cutting their AI subscriptions — the revenue that justifies $700 billion in data center construction disappears.
WHAT HAPPENS WHEN THE BUBBLE BURSTS
Here is the scenario that financial analysts are now modeling — the scenario that Big Tech desperately hopes never makes headlines.
When the AI bubble bursts — and experts increasingly say it is a question of when, not if — many AI-focused data centers will face underutilization or become stranded assets. Some facilities will be powered down to cut costs. New or half-completed construction projects will be halted, leaving empty warehouses and unfinished infrastructure scattered across American farmland, wetlands, and communities. Operators of highly leveraged data centers will face financial stress, leading to auctions of hardware at deeply discounted prices.
Empty buildings. On land that used to be farms. Next to wells that used to provide water. In communities that gave away hundreds of millions in tax breaks. Surrounded by families whose electricity bills went up to power buildings that now sit empty.
That is what a data center bust looks like on the ground. And America has been promised it won’t happen. Just like they were promised in 1999. And in 2007.
The second major risk that could trigger the collapse is technological obsolescence arriving faster than anyone expected. The AI industry is transitioning from the compute-intensive model “training” phase — which requires enormous amounts of power and the biggest data centers — to a less demanding “inference” phase, where trained AI models simply execute their existing knowledge. Inference requires dramatically less power and completely different infrastructure. Data centers optimized for training workloads — the ones being built right now, on your farmland, with your water — could find themselves overbuilt and underutilized within just a few years.
The buildings being constructed right now — the ones draining your aquifer, hiking your electricity bill, destroying your community — may be obsolete before they are even finished. Because the technology they were designed to power is already evolving past them.
MICROSOFT IS ALREADY PULLING BACK — AND NOBODY IS REPORTING IT
Here is the most alarming data point in this entire story. The one that should make every American who has been paying higher electricity bills stop cold.
Microsoft — the single largest customer of CoreWeave, representing 62% of its revenue, and one of the companies that has committed hundreds of billions to data center construction — has begun canceling or pausing data center leases in multiple countries. Microsoft has also taken the extraordinary step of banning the use of AI by its own engineers — the very engineers building AI products — citing concerns about cost and productivity. Heavy users inside Microsoft were burning $500 to $2,000 per month on AI tools. The company decided that was unsustainable.
Microsoft. The company spending $80 billion this year on AI data centers. Is banning its own engineers from using AI. Because it costs too much.
Let that settle for a moment.
The company building the most data centers in human history has concluded that using the AI those data centers power costs more than it is worth. For its own employees.
And yet — your electricity bill is higher. Your water is being drained. Your farmland is being bulldozed. For buildings that Microsoft’s own engineers are being told not to use.
THE DOTCOM COMPARISON THAT KEEPS EXPERTS UP AT NIGHT
The late 2025 and early 2026 debate among the world’s most powerful financial minds has been described as “Doomsters vs Boosters” — with BlackRock CEO Larry Fink and Nvidia CEO Jensen Huang arguing the fundamentals are strong, while Goldman Sachs, the IMF, and a growing chorus of independent analysts point to the unmistakable signs of speculative excess. Google CEO Sundar Pichai acknowledged “there would be overshoots in terms of valuations” — but insisted demand is real. The critical question is: how big is the overshoot?
In 1999, everyone said dotcom demand was real. It was. But the overshoot — the overbuilding, the overvaluing, the overconfidence — was catastrophic. When it corrected, $5 trillion in wealth evaporated. Millions lost jobs. Pension funds were destroyed. Communities that had been promised tech prosperity were left with empty office parks.
The data center industry as a whole is unlikely to fully collapse — because general demand for cloud computing, video streaming, gaming, and enterprise IT remains strong. But AI-specific data centers — the ones built for the most power-hungry, most expensive, most rapidly evolving workloads — face the highest risk. And those are the ones being built right now. On your farmland. With your water. With your tax breaks. With your electricity.
The safest data centers will survive. The AI-specific mega-campuses — the ones that required the most sacrifice from American communities — face the highest risk of becoming tomorrow’s empty warehouses.
THE BOTTOM LINE
Every community in this series gave something.
Patsy Bode gave her home of 20 years in Indiana. The families of Festus gave their political trust — and then took it back. The 81-year-old widow in Virginia gave money she didn’t have. The farmers of Arizona gave their crops. The residents of New Carlisle gave their well water. The people of Utah gave their fight — and won.
All of them were told: this sacrifice is necessary. This is the future. This will create jobs and growth and American competitiveness.
Goldman Sachs, JPMorgan, and the IMF are now warning: the financial model behind all of those sacrifices may be built on sand. The bubble may be inflating. The bust may be coming. And when it does — the empty buildings and the stranded assets will be left behind in communities that gave everything.
While the billionaires who built it will have already cashed out.
We have seen this movie before. In 2000. In 2008. The names were different. The buildings were different. The promises were the same.
Share this with everyone who has been paying higher electricity bills for data centers. They deserve to know the whole story — including the ending that Wall Street is quietly preparing for.
Follow for more data center updates
Source: Built In — “Where AI Data Centers Are Headed After 2025’s Boom” (January 14, 2026) + Goldman Sachs / IMF warnings compiled by Global Research (May 2026)
See less
No comments:
Post a Comment