The U.S. midterms crystalized the politics of data centers; the next few months will show what that means for capex, permits and power contracts. Several newly elected governors campaigned on restricting development. We don’t think that changes why the buildout is happening, but it does change how, where and how fast it proceeds.
Four Dates That Matter
- Late 2026: Permitting reform. The one major energy bill with bipartisan support and a realistic chance of passing before the new Congress takes office on Jan. 3.
- December 10: Texas audit results. We expect regulators to separate serious data center projects from speculative ones, which could speed up the serious ones.
- January and February: Fourth quarter earnings and 2027 capex guidance. We expect guidance to keep rising, but how companies discuss funding may matter as much.
- January onward: New governors and legislatures. Campaign promises must become laws and appointments. Watch Texas, Ohio and Georgia.
Hyperscalers: Spending Stays High, Funding Moves to the Center
We don’t expect the elections to change the pace of AI innovation or the race among big cloud providers and model developers such as OpenAI and Anthropic. Some projects will slip on permitting or local opposition, but the largest companies are locking in land, buildings and power now and deciding on chips later, leaving room to slow spending if demand weakens.
If demand holds, we expect capex estimates to keep rising and free cash flow to stay under pressure for one to two years.
Hyperscaler Capex Is Expected to Keep Rising (US$ B)
Financing is now central. One large cloud provider has raised roughly $115 billion of long-term debt this year and may consider selling equity, and another is raising about $98 billion of equity. We expect more of this ahead of fourth-quarter earnings.
The cushion is real: operating cash flow is strong and backlogs are healthy. The bigger risk is slowing cloud or other revenue growth, not a lack of funding.
Power and Utilities: Where Politics Meets Cash Flow
For utilities, what matters is how campaign promises become laws, regulatory decisions or permitting delays. We are watching who pays for new generation and transmission, whether electricity stays affordable, who is appointed to utility commissions, and how much proposed demand becomes signed contracts.
Several factors limit the downside. In Texas, the legislature is Republican and the Senate must confirm new utility commissioners, and the outgoing governor had already paused approvals, so scrutiny is bipartisan. Affordability and reliability both require more generation, so we think the likeliest outcome is a middle ground in which data centers pay their own way, including by supplying some of their own power.
Results will vary by state. Utilities in Florida, Louisiana, Indiana, Missouri and Iowa look better placed to win new load than those in less receptive states such as Ohio. Where projects are signed, long-term take-or-pay contracts should keep earnings steady.
Midstream Energy: Gas Demand Isn’t a Partisan Issue
The case for gas pipelines rests on the need for reliable power, with utilities building gas plants again. The Federal Energy Regulatory Commission, which approves interstate pipelines and LNG terminals, retains a 3–2 Republican majority, limiting near-term risk. Senate confirmation of future commissioners is the longer-term lever. During the last Democratic Congress (2021–2022), energy was the best-performing S&P 500 sector in both years.
Semiconductors
Chipmakers carry more delay risk than the companies building data centers. Near-term demand is largely committed against existing infrastructure, but the industry is planning capacity for demand years out. If projects slip in states with large planned builds, particularly Texas and Ohio, oversupply becomes a risk, with potential effects on demand, pricing and profitability in 2028 and beyond.
We think some of this is already in the price, with valuations compressed and stock performance weak. We still expect the long-term build to continue, with more data centers and more AI usage over time. But semiconductors are a cyclical industry, and periods of digestion are possible even as the direction of travel is up.
What Would Change Our View
Our baseline is a high but more selective buildout. We would turn more cautious if we saw:
- Hyperscalers signal a slowdown. Lower 2027 capex guidance, stalling cloud growth or order digestion at chipmakers.
- Pauses become law. Moratoriums that survive legislative and legal challenges, especially in Texas.
- Cost-sharing turns punitive. New fees and taxes that make projects uneconomic.
- Funding strains appear. Wider credit spreads on AI debt or a weak reception for equity raises.
Selectivity Over Direction
We expect the AI buildout to continue, with more differences between states, companies and projects. That should favor companies with strong balance sheets, signed contracts and a clear path through permitting. The next few quarters are less a verdict on the AI trade than a test of who is best placed within it.
