Author: Ada, Deep Tide TechFlow

On April 20, Fermi's stock price stood at $5.4.

Six months ago, this figure was about $37. And it had just gone public last October.

A company that was founded less than 12 months ago, with zero revenue, zero tenants, and no noteworthy products, managed to raise $785 million on the Nasdaq, reaching a market value of up to $12.5 billion at one point.

However, both the CEO and CFO resigned on the same day, construction sites halted, insiders sold $68 million worth of stock, short-selling institutions filed fraud accusations, and a collective securities lawsuit has been initiated.

This is the first major collapse of the AI power narrative.

The "World's First" in the Texas Wilderness

The story of Fermi began in early 2025.

Former U.S. Energy Secretary Rick Perry and private equity mogul Toby Neugebauer co-founded this company. The core bet was called “Project Matador”: to build the world's largest AI data center campus on 5,800 acres of land outside Amarillo, Texas, initially powered by natural gas, with plans to add four nuclear reactors in the future.

The plan was to provide 11GW of power capacity and construct approximately 18 million square feet of data center facilities. The label of "world's largest" was frequently mentioned.

The power hunger of AI is real, nuclear energy is green, and Trump signed an executive order to expand U.S. nuclear capacity from 100GW to 400GW. All trends aligned.

The market believed it too. On October 1 last year, Fermi went public at $21 per share, jumping to $25 at the opening, fully oversubscribed. The next day it peaked at $37, a 76% increase over the offering price. Within a few days, this company with no revenue surpassed a market value of $10 billion.

At that time, everyone was buying AI power concept stocks. No need for customers, no need for revenue, just a PPT and a vision invisible to others.

Real Crisis

The first crack appeared last December.

The only anchor tenant of Fermi terminated its contract, widely believed to be Amazon. This tenant had promised to prepay up to $150 million for construction funding, but actually paid nothing.

The short-selling institution Fuzzy Panda uncovered the underlying reason. Fermi had promised to raise $5 to $5.5 billion to ensure project execution, but that money never materialized. The tenant couldn't wait any longer and left.

According to Fermi's lease terms with Texas Tech University, without signed tenants, Fermi was not even allowed to commence construction. This created a deadlock: no tenants meant no financing, no financing meant no construction, no construction meant no tenants.

The construction site halted, and workers said on social media, “We have all been laid off.”

Then came the recent shocking news: CEO Neugebauer and CFO Miles Everson both resigned at the same time. The company packaged it as a “Fermi 2.0” strategic transformation. But the stock price fell another 22%. Since last year’s IPO, investors who purchased FRMI shares have seen losses of up to 78%.

Insiders had already started to run. As soon as the lock-up period ended on March 30, co-founder Rick Perry's son Griffin Perry immediately sold 11 million shares, cashing out $56.3 million. The COO, CFO, and Chief Development Officer quickly followed, with insiders collectively selling over $68 million.

Fuzzy Panda revealed that before the lock-up, Griffin Perry attempted to sell 30 million shares in one bulk transaction.

This is not Neugebauer's first company to encounter a crisis.

In 2022, the "anti-woke" bank GloriFi that he founded collapsed after burning through investments from conservative backers like Peter Thiel, Ken Griffin, and Vivek Ramaswamy, and subsequently filed for bankruptcy. The bankruptcy trustee accused Neugebauer in court documents of “securities fraud,” “extreme self-dealing,” and “fraudulent transfers.”

Fuzzy Panda's report also pointed out that several current executives at Fermi had been partners with Neugebauer during the GloriFi period. Chief Site Development Officer Charlie Hamilton was described in bankruptcy documents as Neugebauer's “longtime friend.” CFO Miles Everson was also accused of participating in unfair transactions involving alleged conflicts of interest.

The bankruptcy court ruled that Neugebauer's multiple transactions constituted fraudulent transfers. Although he had just been accused of fraud in his previous company, his next company raised $785 million on the Nasdaq. The IPO prospectus even stated that these lawsuits could distract management, yet investors still bought in. What does this indicate? It indicates that during bubble periods, people disregard risk disclosures and only care if the story is compelling enough.

A Microcosm Underneath the Bubble

Fermi is not an isolated case. It is a microcosm.

According to data from Sightline Climate, as of April 2026, there are approximately 140 large data center projects planned to launch in the U.S. this year, but only a third of them are actually under construction. The rest are either delayed or canceled.

The bottleneck lies in electrical components.

Transformers, switchgear, and batteries are essential components for every data center construction. Before 2020, the delivery time for large power transformers was 24 to 30 months. Now, wait times can extend to five years. This is structurally unacceptable for data centers with deployment cycles of less than 18 months. A delay in any single component can halt the entire project.

A deeper issue is the generational gap. The U.S. power grid was designed without considering the loads required by artificial intelligence. Data centers can be built in three years, but generating power takes longer. Solar or wind energy generation can take three to six years, gas turbine generation takes about six years, and nuclear energy generation takes more than ten years. Wired Magazine pointed out that when data centers are smaller, this mismatch can be overcome. But now, the scale required for artificial intelligence is so large, with individual facilities’ power measured in hundreds of megawatts, that it has become an insurmountable bottleneck.

OpenAI's flagship project Stargate, which is said to have a budget of $500 billion, has shown no substantial construction progress as of April.

Partners are at odds over site ownership and system control rights. The 800MW expansion of the Texas flagship campus has been canceled. Stargate projects in the UK and Norway have been suspended, and three core executives responsible for Stargate have jumped to Meta.

Meanwhile, Alphabet, Amazon, Meta, and Microsoft are expected to invest over $650 billion to expand artificial intelligence capacity by 2026. Among these, Alphabet alone reports a budget of $175 to $185 billion, burning about $500 million a day. However, the infrastructure to support this ambitious goal cannot develop at the required industry pace.

The last time the U.S. saw a similar scale of energy infrastructure boom was in the late 1990s. The internet bubble and deregulation of electricity markets spawned a wave of natural gas power plant construction, with about $100 billion invested. After the bubble burst, many power plants were left idle.

This time, the scale has increased by an entire order of magnitude. U.S. utility companies have outlined $1.4 trillion in planned spending, 27% higher than last year's forecasts. Investments by tech companies in energy-related infrastructure are already double that of the entire U.S. power industry's annual investment.

But from the third to the fourth quarter, new data center transactions have declined by over 40%. Some analysts believe that capital expenditures by supercomputing companies may be cut in half this year.

The money is shrinking, but the story continues to be told. That is where the danger lies.

A report from Built In summarizes: When suppliers heavily invest in startups, and these startups turn around and spend the money on their own products, real demand and artificially created illusions become intertwined. When a company's customers are also its investors, and revenue growth outpaces actual usage growth, this signals that a bubble is forming.

When the Bubble Bursts

In this food chain, there are three layers of players.

The first layer consists of the real winners. These are companies that already own operational nuclear power plants, like Constellation Energy, which don’t need to build anything new; they just need to transfer existing power plant contracts from the grid to data centers to reap the AI power dividends. Meta signed a 20-year, 1.1GW nuclear power supply contract with Constellation. Microsoft spent $1.6 billion to restart the Three Mile Island Nuclear Power Plant. These are tangible transactions.

The second layer comprises various small modular reactor (SMR) startups, like Oklo, whose stock prices have soared, but not a single reactor has been built. U.S. nuclear power projects are notoriously known for delays and cost overruns; almost all cases in recent decades have failed to meet their original schedules and budgets. But investors don’t care about that.

The third layer is companies like Fermi, which have no nuclear reactors, haven’t begun construction on gas plants, and have no tenants at all. They sit at the bottom of the food chain, selling not electricity, but stories. When the story collapses, there is nothing left.

The collapse of Fermi will not be an isolated event.

When an industry's actual delivery capabilities lag far behind the promises made in PPT presentations, a collapse is only a matter of time.

Of the data center capacity that the U.S. plans to launch in 2027, only 6.3GW is actually under construction, while the announced total is 21.5GW. There exists a gap of 15GW on paper, corresponding to hundreds of billions of dollars in funding and countless promises that will remain unfulfilled.

Who will be the next Fermi? No one knows. But in this race, there is $500 billion looking for electricity, high power transformers waiting for delivery, and a large number of startups that haven’t even solved basic grid access, assuring investors that everything is under control.

The last time the energy infrastructure bubble burst, there were still power plants left. This time, many projects may not even excavate a foundation.

And Fermi's 5,800 acres of land in the Texas wilderness will slowly be buried by time along with all the unfulfilled grand narratives.