Democrats won control of both the House and the Senate on 3 November. Will a blue tsunami reshape markets? Mostly not, because the US system sharply limits what Congress can change on its own. But the election has exposed one real risk, and it sits in the states rather than in Washington. The AI data-centre buildout may arrive later than investors assume.
The swing is large. Democrats are projected to hold at least 250 House seats and at least 54 Senate seats, against Republican majorities of 219–214 and 53–47 going into the vote. Yet the president’s veto and the Senate’s 60-vote filibuster will limit how much of the Democratic agenda becomes law, leaving hearings and investigations as the new majority’s main tool. Federal AI policy runs into the same wall. Proposals such as reversing the immediate tax write-off for new investment or taxing computing power are unlikely to become law while the president opposes them. On fiscal policy, the main lever is the debt limit, where the new majority could attach its own priorities and drag out a resolution.
The governors’ races matter more. Texas accounts for roughly 20% of North American data-centre development. Its incoming governor has called for a moratorium on new data-centre construction until new legislation passes, and the state legislature sits from January to May 2027. The pledge targets new permits rather than sites already under construction, which points to delays from 2028. Analysis of project pipelines suggests little risk to 2027 capacity, while 2028 projects in Texas and Ohio are more exposed, with around 12 GW of pipeline through 2030 in question. The stakes are large, with Amazon, Alphabet and Meta alone planning more than $900bn of capital spending in 2027. State-level rules are unlikely to derail the buildout, but they could make it slower and more expensive.
The stock market has already drawn its own conclusions. A brief relief rally after the election turned into a sharp sell-off, and the Nasdaq 100 had its steepest one-day fall since March 2020 on concerns that Democratic gains could hold back data-centre expansion. The large cloud providers fell far less than chipmakers, because spending less would bring forward their return to positive free cash flow. That reading may be too neat. The hyperscalers’ AI capital spending is expected to more than triple from 2025 levels by 2028. The largest operators are also much better placed than smaller, speculative developers to handle the new politics, and tougher rules are more likely to change where data centres are built than to stop them.
For the market, the election matters mainly through the timing of profits. A data centre delayed by a year pushes back revenue for the chipmakers, power suppliers and builders that serve it, and profits that arrive later are worth less today. The question to ask is how much of a portfolio’s expected growth depends on new US capacity coming online on schedule in 2028. Companies with power-secured sites already running are better protected, and may even gain pricing power if new supply is held back.