Fermi just raised $682 million and closed at $32.53, its first day up 55%. 

The company is now worth $19.3 billion. Here's what makes that number remarkable: Fermi has zero revenue.

This isn't a software startup promising future profits. 

It's a REIT building power infrastructure for AI data centers. Investors are betting nearly $20 billion that they can deliver.

Fermi's IPO pricing already signaled strong demand. But the 55% pop on day one?

That's not normal enthusiasm. That's investors scrambling for exposure to AI infrastructure before the window closes.

The company plans to build Project Matador on 5,800 acres leased from Texas Tech. The first phase targets 1.1 gigawatts by 2026. Full buildout? 11 gigawatts. 

To put that in context, most large data centers run on 50-100 megawatts. Fermi is planning capacity for over 100 hyperscale facilities.

But here's the thing: their first tenant deal is non-binding. Revenue won't start flowing until 2027 at the earliest. That's a three-year gap between a $19 billion valuation and actual cash coming in the door.

That gap matters more than you might think.

Fermi's co-founder is Rick Perry, former U.S. Energy Secretary. 

The campus carries President Trump's name. These aren't decorative details. In a business where permitting, grid access, and regulatory approval can kill projects before they break ground, political connections are infrastructure.

Texas has become the default home for data center expansion. Cheap land, business-friendly regulation, and existing energy capacity make it the obvious choice. But even in Texas, building gigawatt-scale power infrastructure requires coordination across utilities, state agencies, and local governments. 

Fermi's leadership team knows those conversations. They've had them before. That matters when you're asking investors to fund a project with no revenue and a 2027 timeline.

Fermi isn't in the major digital infrastructure ETFs yet. Funds like VPN and SRVR hold established data center REITs with proven cash flows. But if Fermi executes, that changes.

A successful buildout and tenant acquisition would likely trigger inclusion in these funds. That would push more capital into the sector and lift valuations across existing holdings. ETFs offer diversified exposure without the binary risk of betting on a single zero-revenue company.

Here's what we know: AI models are getting bigger. Training runs are getting longer. Inference workloads are scaling faster than anyone predicted two years ago. All of that requires power. Lots of it. And data centers can't run on optimism.

Investors' enthusiasm for AI infrastructure is real. Renaissance Capital and other market analysts are calling this a "gold rush" moment. But gold rushes are messy. Most prospectors go home broke.

Fermi's valuation assumes they can deliver on time, secure binding tenant agreements, and scale from zero to 1.1 gigawatts in under three years. That's aggressive. 

Construction delays, cost overruns, and tenant negotiations that fall apart are standard risks in real estate development. Now add in the complexity of utility-scale power infrastructure and hyperscale data center requirements.

The first tenant deal being non-binding is a red flag. Not a dealbreaker, but a reminder that commitments mean nothing until contracts are signed and deposits are paid.

If Fermi hits its 2026 milestone, expect more SPACs and IPOs in this space. The market is hungry for AI infrastructure plays, and capital will flow to anyone with credible land, power access, and political support.

If they miss, or if the first tenant deal falls through, this becomes a cautionary tale about valuation getting ahead of fundamentals.

For now, the market is betting on execution. A $19 billion valuation with zero revenue only makes sense if you believe AI infrastructure demand will outpace supply for the next decade. The data suggests that's likely. But "likely" and "guaranteed" are different words.

Investors chasing direct exposure to Fermi are taking execution risk. Those looking at ETFs like VPN or SRVR are getting diversified access to the sector without betting everything on one company's ability to deliver.

The AI infrastructure buildout is real. The question is whether Fermi can turn political capital and ambitious plans into gigawatts and revenue. We'll know by 2027.