It was an expected result, but a remarkable one nonetheless, and it sent shockwaves through financial markets. In the US midterm elections, which determine the composition of Congress, the Democrats dealt a decisive defeat to the Republican Party, taking control of both the House of Representatives and the Senate. Donald Trump’s party suffered particularly symbolic defeats in traditionally conservative states such as Kansas, where the Democrats had not won a Senate seat since 1932, and Texas.
It is hardly unusual for the president’s party to lose support in midterm elections, and the outcome had been widely anticipated. Losing control of both chambers, however, is a much rarer occurrence, having happened only six times in the past 150 years. Previous examples include the elections held in the closing days of the First World War, in 1918, and shortly after the Second, in 1946.
If the scale of the defeat was almost unprecedented, the market reaction was equally striking. After initially rising in the aftermath of the vote, the Nasdaq 100 fell 5.1% on Thursday, its worst session since April 2025. The semiconductor index plunged 12%, its second-largest decline on record. Losses continued into Friday’s trading session.
The sell-off was driven by a correction in technology stocks linked to Artificial Intelligence (AI), a sector that has supported corporate profit margins and equity markets in recent years. Investor concerns centred on the risk that newly elected governors and local communities could slow the development of AI infrastructure.
The backlash against data centres
The election has opened up two areas of risk for the Artificial Intelligence industry. The first is political. A Democratic majority in Congress theoretically represents a victory for a political platform that, in some areas, promises greater scrutiny of Big Tech and AI development.
For now, however, this shift remains largely theoretical, and we do not believe the new Congress will be able to introduce measures in the short term that would fundamentally alter support for the development of this technology. Nevertheless, the political shift reflects growing opposition to the unchecked expansion of Artificial Intelligence, a risk factor that markets cannot afford to ignore.
The more tangible and immediate risk, however, comes from local communities. On 3 November, voters also decided on 41 local ballot measures across ten states, from Ohio, which alone accounted for 21, to Illinois, Maine and Kansas. At the centre of these votes was the construction of data centres, the infrastructure essential to AI development. In 40 of the 41 cases, voters chose to impose restrictions on data centres. The only exception was Harwood, North Dakota, where voters rejected additional oversight of new industrial facilities.
In most cases, support for restrictions exceeded 60%, sometimes reaching more than 80%. The ballot measures covered a range of proposals, from outright construction bans to restrictions on facilities exceeding certain power thresholds, as well as requirements for public votes on larger projects.
In one Kansas community, for example, a ballot measure introduced a ten-year ban on facilities consuming more than 100,000 gallons of water per day, equivalent to approximately 380,000 litres.
The election results did not come out of nowhere. Earlier in 2026, six other local votes had produced similar outcomes, in some cases with 90% support. In July, New York State had introduced the first statewide moratorium.
What does this mean in practice?
The ballot measures concern specific circumstances in individual municipalities or counties and, taken in isolation, represent a relatively limited risk. The broader political implications, however, could be considerably more significant.
Elections for 36 governorships were also held on 3 November, and in key states such as Texas and Georgia, candidates promising pauses or stricter conditions on data centre development emerged victorious.
Governors, with the support of state legislatures, can slow down permitting processes, change the rules governing connections to the electricity grid and review tax incentives, which were available in 38 states as of April. They will now have to decide how far to go in restricting new developments.
While an outright construction ban appeared, at least before the election, to be a less likely and less predictable outcome, the expectation is that governors will at least introduce measures to slow the growth of data centres.
These could include requiring data centre operators to cover the costs of grid infrastructure rather than passing them on to households through higher electricity bills, or removing tax incentives. Other options include limits on water consumption, requirements for on-site backup power generation, labour-related conditions and bans on non-disclosure agreements during permitting processes.
Many of these measures are reasonable responses to genuine concerns about household energy bills and water resources. However, they also extend project timelines, increase costs and, in some cases, could prevent new data centres from being developed altogether.
The impact will be particularly relevant for projects expected to become operational in 2028, when 96 gigawatts of new capacity are currently anticipated.
Based on the positions taken by newly elected governors, J.P. Morgan estimates that 37.5 gigawatts could face a pause, while a further 15 gigawatts could be subject to additional conditions. Together, these represent more than half of the total. Swiss bank UBS takes a less pessimistic view, estimating that around 30% of new installations planned for 2028 could be at risk.
During the trading session on 5 November, the hardest-hit companies were chip manufacturers and so-called neocloud providers: emerging specialist operators that rent out computing capacity and have taken on substantial debt to build the necessary infrastructure.
Volatility could persist, but we do not believe the market decline in itself signals a dramatic reversal of US support for AI investment.
Data centre operators may introduce mitigating measures or relocate projects to other states. This could increase costs and squeeze margins without bringing the long-term investment trend to a halt.
For now, this appears to be the most likely scenario, although the underlying social tensions remain unresolved.
The vote seems to reflect more than local opposition to new developments, often referred to as NIMBYism (‘not in my back yard’). Alongside concrete concerns about energy bills and water resources, it may also point to a broader unease about the pace and impact of artificial intelligence.
A lame-duck president: what are the implications for inflation and fiscal policy?
Looking further ahead, another potential medium-term consequence concerns the impact of the new political landscape on US government policy over the next two years.
Without control of Congress, Trump will struggle to secure funding for new initiatives, introduce further tax cuts or pass major reforms. The Democrats, meanwhile, will not be able to legislate unilaterally, as the president retains his veto power.
This could lead to two years of political confrontation or, alternatively, a more collaborative approach, as Trump himself has expressed hope for.
For financial markets, political gridlock is not necessarily bad news. It may be even less concerning if it produces a balance that limits legislative excesses on both sides.
On inflation, a Democratic Congress will make it more difficult to extend existing tariffs or introduce new ones, although the president retains certain independent powers.
Oil prices, meanwhile, will depend primarily on the United States’ international stance.
Following the defeat, Trump could decide to break the deadlock in Iran in an attempt to respond to voters’ concerns, and this possibility influenced crude oil prices immediately after the election.
However, the crisis in the Middle East remains complex, and a swift resolution appears unlikely.
Business as usual
Traditionally, midterm elections attract relatively little attention from financial markets and, despite the volatility seen in recent days, we believe this will ultimately prove to be the case again over the medium term.
That does not mean equity valuations linked to artificial intelligence investment are without risk.
Political events can trigger sharp swings in the short term, as recent days have shown. Over longer horizons, however, history suggests that markets tend to be driven more by economic cycles, macroeconomic trends and corporate earnings than by politics.
We believe the medium- to long-term outlook remains positive in this respect, supported in part by the AI revolution, which should settle into a more sustainable pace of development.
The country is unlikely to abandon the race to develop a technology widely regarded as crucial to its economic and geopolitical influence. Nevertheless, that journey will inevitably encounter setbacks.
For this reason, we continue to believe that a diversified approach to investing remains the most appropriate strategy for most investors.
*As with all investing, financial instruments involve inherent risks, including loss of capital, market fluctuations and liquidity risk. Past performance is no guarantee of future results. It is important to consider your risk tolerance and investment objectives before proceeding.
