Democratic victory puts AI trade under pressure
The Democratic victory in the US midterm elections has triggered a sharp reassessment of the AI trade. Not because demand for AI is suddenly disappearing, but because investors are questioning whether the massive expansion of data centres can proceed as quickly and at the costs previously expected.
The market reaction was exceptional. The Nasdaq Composite fell 5.5% on Thursday, the S&P 500 Index declined 4.4%, and the US semiconductor index dropped more than 12%. AI and semiconductor companies were hit particularly hard. The sell-off was also evident in European and Asian markets. Despite yesterday’s decline and the losses over the past few days, most equity markets are still showing a strong positive return for the year as a whole.
From chip shortages to power and permitting constraints
A Democrat-controlled Congress increases the likelihood of stricter conditions for new data centres. Democratic policymakers are primarily concerned about the impact of data centre expansion on electricity prices, power-grid capacity, water consumption, and costs for local communities. Measures taken by individual states and local governments to restrict the development of new data centres could even prove more important than federal regulation.
This changes the case for the build-out of AI. Further growth requires not only sufficient chips and capital, but also power, land, grid connections, and political approval. Initial indications suggest that around 30% of the expected expansion of US data-centre capacity in 2028 could be affected by potential construction restrictions. This does not necessarily mean that all of this capacity will disappear, as projects could also be delayed, scaled back, or phased in over a longer period.
Why is the market reacting so strongly?
AI has been one of the strongest and most popular investment themes of recent years. Expectations for semiconductor manufacturers, data-centre companies, and other AI infrastructure providers had already priced in a very rapid expansion of AI capacity. The election result directly challenges that assumption.
Moreover, concerns already existed about elevated expectations and whether the hundreds of billions of dollars being invested in AI will generate sufficient returns. The election result should therefore be seen more as a catalyst for the correction than its sole cause.
This also explains the significant differences within the AI sector. Hyperscalers such as Microsoft, Amazon, Alphabet, and Meta have strong balance sheets and extensive existing infrastructure. Increasing scarcity of data-centre capacity could even provide them with a competitive advantage. The initial stock market reaction was therefore positive for these companies, although yesterday these hyperscalers were also caught up in the sell-off, albeit to a lesser extent. The risks are greater for companies that still need to finance, build, and bring substantial data centre capacity online. This applies, for example, to companies such as CoreWeave and Nebius. Semiconductor companies such as Nvidia and Broadcom could also become more vulnerable if the construction of new capacity is delayed.
Where are the biggest risks?
We see three particularly vulnerable areas.
First, highly leveraged data centre companies. They could face delayed projects and higher financing costs at the same time. At CoreWeave, for example, this stress is already visible in the credit market.
Second, semiconductors. Structural demand related to AI is likely to remain strong, but current earnings expectations already incorporate a significant amount of future capacity. If data centres are built later than expected, demand for GPUs, memory, and networking equipment could also be pushed further out.
Finally, power and infrastructure. Companies such as GE Vernova and Bloom Energy have benefited significantly from expectations of explosive growth in electricity demand from data centres. Project delays could push some of that growth into later years. There is, however, an important nuance. If new infrastructure becomes more difficult to develop, existing and already permitted capacity could actually become more valuable. Companies that already have strong positions in this area could therefore benefit.
What does this mean for investors?
We would not write off the AI theme, but we would approach it more selectively. The election result reinforces the distinction between companies that already possess the necessary infrastructure, access to power, and strong balance sheets, and those that still need to finance and build their future capacity.
For investors, it will therefore be important to monitor closely over the coming months whether stricter regulations actually prevent the construction of data centres or mainly delay it. Much will also depend on the investment plans of major hyperscalers such as Microsoft, Amazon, Alphabet, and Meta. As long as they maintain their AI investment plans, underlying demand for infrastructure and chips should remain strong.
Credit markets also warrant close attention. Further increases in financing costs for data centre companies could put planned expansions under pressure. Ultimately, however, the key issue is whether the enormous investments in AI increasingly translate into revenue, productivity gains, and earnings growth.
The debate is therefore shifting from the promise of AI to the returns on AI. The key question is no longer simply how strong demand for AI will become, but rather which companies have the balance sheets, access to power, and existing infrastructure needed to serve that demand profitably. That distinction could determine the next winners and losers within the AI theme.
Joost Olde Riekerink
Equity research & advisory expert