Analysts say data-center limits may favor efficient chips over building equipment

Efficient AI chips and data-center cooling equipment may face different futures after the U.S. midterms. In Morningstar’s November 11 analysis, strategists argued that construction constraints could push developers to buy more computing output per facility rather than add floor space. That could favor leading chipmakers while leaving equipment suppliers more exposed to a smaller construction footprint.

More output from each building

Denny Fish, portfolio manager and lead technology equity analyst at Janus Henderson, sees an incentive to squeeze more useful computing from each site. If building becomes harder, developers could prioritize efficient chips that maximize token output—the text units AI systems produce. He identified Nvidia as a potential beneficiary of that shift.

Bank of America analyst Vivek Arya described a related spending choice. Developers could cut future investment in land, power and building shells, including grid connections, while protecting purchases of next-generation chips. His argument concerns how construction budgets might be allocated, not a confirmed change in customer orders.

Fish drew a different connection for electrification equipment and liquid cooling: their sales depend on how many square feet of data-center space gets built. A smaller physical footprint could therefore weigh more heavily on those suppliers than on companies selling the chips developers use to increase output within an existing facility.

That is a relative advantage, not protection from a downturn. Fish cautioned that a material slowdown would affect all chipmakers with data-center exposure and could reduce earnings expectations for 2028 and 2029. His forecast separates which suppliers might hold up better from whether the sector would escape damage.

The votes imposed different kinds of limits

The political trigger was concrete, but not uniform. Ballotpedia’s November 6 tally found that voters favored limits on data-center development in 40 of 41 local ballot contests on November 3. Six earlier contests also favored limits, bringing the year’s total to 46 of 47. Those are counts of ballot outcomes, not the share of Americans opposing data centers.

The results included five full bans, 12 partial bans and three six-month moratoriums. Other outcomes involved zoning, voter approval, financial guarantees or nonbinding expressions of opposition. That mix matters for construction forecasts: a temporary pause, a site-specific restriction and an advisory vote do not impose the same obstacle on a developer.

Markets reacted sharply. The Morningstar US Artificial Intelligence Index fell more than 9% on November 5, its worst single-day drop since March 2020. The supplier forecasts assess the consequences of that post-election environment; they do not describe a newly enacted national restriction.

An earlier peak in spending growth

Dan Ives’s firm now expects Big Tech’s capital-spending growth rate to peak in 2027, rather than 2028 or 2029. The Motley Fool reported the revision, citing CNBC’s coverage of his November 4 note. The forecast concerns the pace of spending growth, not a prediction that total investment will stop rising in 2027.

Existing capacity could also become more valuable. JPMorgan analysts, quoted by Morningstar, argued that tighter computing supply could support pricing power and operating cash flow at the largest cloud providers. By contrast, debt-funded specialist AI clouds such as CoreWeave and Nebius face greater exposure to financing costs and construction hurdles, strategists said.

Location remains another option. Fish said projects could move to states more receptive to development. Barclays credit strategist Dominique Toublan offered a similarly qualified outlook in a November 4 note cited by The Motley Fool through CNBC: local, bipartisan opposition would make construction pricier rather than stop it. His assessment favored companies with the financial strength to secure scarce power.