The AI boom just turned Texas into a test case for whether we can build the future without choking on it.

The Summary

The Signal

Nokia's earnings beat tells you AI infrastructure spending is real money, not PowerPoint promises. The networking equipment orders flowing to Nokia are the physical manifestation of every "we're investing in AI" earnings call over the past year. But the equipment is only half the story.

Texas has become ground zero for AI's physical footprint, and the electricity required to run these facilities is creating an environmental accounting problem nobody wants to solve. The state's relatively loose regulatory environment and available land made it attractive for data center development. Now those same factors mean there's less oversight on how developers meet their massive power requirements.

"Tech companies and data center developers are devising new ways to get enough electricity, including building out their own power sources."

Here's the pattern emerging:

  • Utilities can't expand capacity fast enough to meet AI infrastructure timelines
  • Developers are building captive power generation to maintain construction schedules
  • Gas turbines get built because they're faster to deploy than renewable projects
  • Air quality becomes someone else's problem

The fundamental tension is temporal. Data centers need power now. Renewable projects take years to permit and build. Gas turbines take months. When you're racing to capture AI market share and your competitor is signing deals for capacity you don't have yet, you choose the fast option.

This isn't unique to Texas, but Texas is where the collision is most visible. The state hosts major facilities from virtually every hyperscaler, and the current wave of planned projects could push pollution levels back two decades. That's not a rounding error. That's a measurable public health regression in service of training frontier models.

What makes this different from previous infrastructure booms:

  • The timeline is compressed (everyone building simultaneously, not sequentially)
  • The power density per facility is unprecedented
  • The economic incentives favor speed over sustainability at every decision point

Nokia's strong quarter reflects real capital deployment into this infrastructure layer. The money is flowing. The racks are filling. The models are training. And somewhere in the background, the air quality gains of the past twenty years are quietly unwinding.

The Implication

If you're building AI applications, your compute costs include an externality you're not paying for yet. Texas today, but this pattern will repeat wherever power is cheap and regulation is light. Watch for two things: regulatory pushback that actually slows deployment timelines, and the first major hyperscaler to credibly commit to only powering new capacity with renewables, even if it means waiting.

The companies profiting from AI infrastructure buildout like Nokia have no incentive to slow down. The question is whether local governments will force the issue before the air quality data becomes politically impossible to ignore.

Sources

Bloomberg Tech | The Verge AI