What’s Next for the AI Trade After the Midterms?

Key Takeaways

• A tougher regulatory environment will likely slow data center construction, delaying some AI hardware revenue, but strategists say the buildout is still a huge opportunity for investors.
• Money managers say firms with existing infrastructure, scale, and durable competitive advantages could even benefit if a slowdown exacerbates supply/demand imbalances.
• Analysts say a slowdown poses a greater risk to highly leveraged neoclouds and infrastructure suppliers whose sales are more tied to data center square footage than computing capacity.

The already-wobbly artificial intelligence trade took another hit from the midterm elections, as investors reconsidered the outlook for the data center construction boom that has lifted many stocks. Despite the fresh headwind of uncertainty about what the Democratic Party midterm sweep will mean for the data center buildout, strategists say AI opportunities remain, but investors will need to be more selective.

A slowdown could offer the hyperscalers—Microsoft MSFT, Alphabet GOOG/GOOGL, Amazon AMZN, Meta Platforms META, and Oracle ORCL—greater pricing power, ease pressure on their cash flows, and enhance their structural advantage over neocloud competitors. The divide between haves and have-nots in the semiconductor industry could expand as data center developers seek to cram as much computing power as possible into each facility. Suppliers of ancillary services like power generation and cooling could face headwinds from a smaller data center footprint.

“Right now, the path of AI is unfortunately unknowable,” said Sameer Samana, head of global equities and real assets at the Wells Fargo Investment Institute, on a call with clients Friday. “The only thing I do know is it’ll be a more moderate pace than it used to be.”

AI has driven stock market gains for the better part of four years. Since the beginning of 2023, shortly after the release of ChatGPT, AI stocks have returned nearly twice as much as the broader market. In that time, the rally endured its fair share of setbacks, but last week’s was arguably the most punishing yet. Two days after the election, the Morningstar US Artificial Intelligence Index dropped more than 9%—its worst day since the onset of the covid-19 pandemic in March 2020. The index’s year-to-date return dropped into single digits last week for the first time since April.

The rout underscored the volatility of a highly concentrated and uncertain market, leaving investors wondering whether the panic was a hiccup, like prior AI selloffs, or the beginning of a new reality.

Last week’s slump was sparked by Democrats who campaigned against data centers getting elected to statewide offices in states with high concentrations of data centers, like Texas, Ohio, and Georgia. They ran on imposing moratoriums on new data center development and implementing more stringent regulations around electricity and water use. Investors fear opposition to data centers will slow their construction, softening demand for semiconductors, networking equipment, and other AI-enabling hardware.

Data Centers Aren’t Dead

Spending on data centers has energized the stock market and broader economy for years. The hyperscalers and SpaceX SPCX are expected to spend more than $900 billion on capital expenditures this year, nearly twice 2025’s total, which was itself nearly double 2024, according to JPMorgan. Their investments in AI-enabling infrastructure have helped companies outfit data centers with semiconductors, servers, networking equipment, and more. As a result, the fates of some of America’s most valuable companies, like Nvidia NVDA and Micron MU, are now inextricably linked to the data center boom.

Strategists say development may slow, but companies have options. Denny Fish, portfolio manager and lead technology equity analyst at Janus Henderson, says future data centers that might have broken ground in Texas may move to more hospitable states. He notes that on the same day tech stocks sold off, Alphabet announced a $17 billion data center project in Republican-led Wyoming. “This data center capacity will get shifted one way or another,” he says.

Companies also have plenty of incentives to aggressively pursue data center projects. “Every one of these CEOs and founders view this as existential,” says Fish. In addition, “at the federal level, this is non-optional.” The White House still considers American AI leadership a national security imperative.

All these factors give some strategists and money managers confidence that the AI buildout has plenty of runway. “We took advantage of the opportunity when things became dislocated as it relates to valuations,” says Marc Nabi, equity investment director at Capital Group. “Even though there might be near-term headwinds from a regulatory standpoint, the long-term opportunity of market share leaders should persist.”

What the Midterms Mean for the Hyperscalers

One potential winner of slower data center construction: hyperscalers. “This really benefits the companies that have capital and have existing data center footprints,” says Janus Henderson’s Fish. Investors may have thought the same last week. The hyperscalers were mostly spared from Thursday’s carnage, and most of their stocks are up since the midterms.

Demand for computing capacity has exceeded supply for years, which has helped accelerate growth in hyperscalers’ cloud businesses. If moratoriums and harsher regulations slow the addition of new capacity, that supply/demand imbalance may be exacerbated. “Tighter constraints on compute capacity could support greater pricing power, higher incremental margins, and improving operating cash flow for hyperscalers,” wrote JPMorgan analysts in a note late last week.

Slower development could also ease pressure on hyperscalers’ free cash flows, which turned negative this year for the entire group except Microsoft. That cash burn has become a symbol of the data center buildout’s immense costs and uncertain returns, and a major drag on investor sentiment. It has also forced hyperscalers to sell debt and issue equity, increasing their interest rate risk and diluting shareholders.

After last week’s selloff, JPMorgan analysts “see attractive risk-reward” in hyperscaler stocks and “believe the group has room to re-rate higher over the medium term.”

What Is the Outlook for Semiconductors?

Semiconductors were among the hardest-hit stocks during last week’s rout. The group had been in a slump since late June, following a blistering run-up in the second quarter. Thursday’s selloff deepened the pullback. The PHLX Semiconductor Index finished Thursday 37% below its June peak, 5 percentage points more than the average peak-to-trough drawdown in past boom/bust cycles, according to Vivek Arya, managing director and senior analyst at Bank of America.

After Thursday’s selloff, 63% of technology stocks were trading below their 200-day moving average—a key trendline analysts use to gauge a stock’s momentum. It was one of the worst readings on that front in years, close to “historical washout territory,” according to Darrell Cronk, president of the Wells Fargo Investment Institute.

Strategists say some babies have likely been thrown out with the bathwater. “The semiconductor ecosystem in general should remain relatively robust for several years, just given the supply-demand imbalances,” says Fish of Janus Henderson. But he predicts markets are “moving away from the rising tide lifts all boats [trend]” that characterized the chip rally earlier this year.

Fish says a data center slowdown would likely benefit the chipmakers with the most efficient hardware. If constructing data centers becomes more cumbersome, developers have an incentive to maximize their facility’s token output, which is a potential tailwind for technology leaders like Nvidia.

To be sure, all chipmakers with data center exposure would be impacted by a material slowdown. “It probably takes a little air out of the [earnings] estimates for 2028 and 2029,” says Fish.

BofA’s Arya estimates some of that air has already been let out. “Semis already trade at a 22% discount to the broader market,” he wrote last week, “suggesting a decent chunk of [EPS] cuts may already be reflected” in stock prices.

Neoclouds at Risk

Strategists say the outlook is haziest for neoclouds like CoreWeave CRWV and Nebius NBIS, cloud computing upstarts constructing purpose-built AI data centers. Unlike hyperscalers, neoclouds have no existing businesses generating cash to funnel into data center construction. Instead, they have borrowed to build facilities to rent to other companies, expecting future cash flows to cover their debt.

Due to their modest scale, neoclouds are considered especially vulnerable to increased costs and regulatory hurdles. “Their business models are predicated on the idea that they can go out and find land, space, and power,” says Fish, “and that’s going to become tougher.”

Access to capital markets is another concern. Investors may respond to a more uncertain data center landscape by demanding greater compensation for lending to neoclouds, increasing borrowing costs and potentially hampering growth. Investors have already signaled discomfort with neoclouds’ borrowing. CoreWeave’s debt yields spiked last month amid mounting concerns about its leverage.

“Currently, their access to capital is more limited” than that of the hyperscalers, says Capital Group’s Nabi, yet their businesses “are predicated on having access to equity and fixed income markets.”

Can Industrials Remain AI Beneficiaries?

Industrial sector stocks were among the best performers through the first half of the year, as hyperscalers’ surging capital expenditures drove up shares of companies supplying data centers with power generators, heavy machinery, cooling equipment, and other ancillary hardware. Many of those stocks that rallied have, like chip stocks, pulled back since June.

Samana of Wells Fargo sees opportunity in industrials. “Those [companies] have a leg in both the AI camp and the real economy,” he said last week.

Others think a data center slowdown could hurt these stocks’ growth more than those of semiconductor and server suppliers. “In practice, capex cuts could target future land, power, and shell [the data centers’ physical structure and grid connections],” while developers prioritize spending on next-generation chips, wrote BofA’s Arya.

Some suppliers might be hurt more than others if their AI-related revenue is driven by the number and size of data centers. “Think about electrification equipment and liquid cooling,” says Janus Henderson’s Fish. “That’s really dependent on how many square feet of a data center you have.”