Key Takeaways

  • Tech equities dropped 5% to 10% in after-hours trading as midterm election returns signaled broad Democratic victories.
  • Data center permitting freezes in states like Texas threaten to block 10 to 15 gigawatts of expected power capacity by 2027.
  • Because AI infrastructure drove roughly 75% of recent US GDP growth, cutting the buildout pace in half creates an estimated 50-basis-point drag on economic growth.
  • California voters rejected Proposition 40, demonstrating that even during Democratic sweeps, voters draw a hard line against asset seizure taxes.
  • Core AI labs will stay insulated from hardware bottlenecks because end-user software demand and private software margins remain intact.

Politics Just Hit the Power Grid

On election night, Brad Gerstner watched tech stocks drop immediately. As Gerstner observed, “AI stocks have had a 10% reversal in the last hour” (play from 1:28:45) as early returns suggested broad Democratic gains. The trigger was not an interest rate hike or a missed earnings print. It was political shock. As returns pointed toward a Democratic sweep in Congress and key governorships, public equity markets began pricing in regulatory friction for compute infrastructure.

Gerstner pointed directly to state-level permits. In Texas, a political shift could lead to an immediate moratorium on new data center interconnections. As Gerstner warned, “the big fear is that in a blue tsunami particularly in the state of Texas you have hosa who will issue a moratorium on new data centers.” (play from 1:29:23) The problem is not money. Hyperscalers and venture funds have the cash to build. The problem is physical power hookups. Gerstner calculates that local roadblocks could cause “at least 10 to 15 gigawatts of the expected build not come online” (play from 1:33:14) through 2027.

Infrastructure Bottlenecks Will Not Kill Application Software

The macroeconomic stakes are massive. “About 75% of GDP growth, I think came from AI infrastructure,” (play from 1:39:16) Gerstner noted. If state moratoriums cut that construction pace in half, he warned of an immediate “50 basis point headwind to GDP growth next year.” (play from 1:39:22) That drag hits hardware suppliers, utilities, and chip makers first.

Yet Gerstner makes a sharp distinction between steel in the ground and software in production. A power bottleneck limits cluster size, but it does not extinguish customer demand for models or products. Leading research labs remain profitable cash machines with clear paths to public offerings. When hardware access tightens, efficiency becomes the winning strategy. Model makers find ways to run inference on leaner compute budgets, and application builders spend less time chasing giant parameter counts and more time driving revenue from workflows that work today.

At the same time, policy fear has clear limits. Gerstner highlighted the defeat of California's Proposition 40, an asset seizure tax proposal. Even amid broad progressive wins, voters flatly rejected taxing unrealized capital. Capital markets can absorb regulatory gridlock over physical infrastructure; they revolt against direct wealth expropriation.

What to Do With This

Audit your compute dependency by the end of the week. If your software product relies on clusters expanding by 4x next year to drop your unit costs, assume those megawatts will not arrive on schedule. Shift your engineering roadmap toward quantization, smaller fine-tuned models, and multi-cloud inference providers that own their power generation.