(Nov 5): The backlash against data centers evident in this week's US midterm elections raises new risks for a key source of economic growth. But restrictions on development could ultimately reduce the odds of a downturn by keeping the AI investment boom in check.

Around half of planned expansion nationwide is vulnerable to increased scrutiny and pushback from Democrats who clinched historic victories in states like Texas, Georgia and Ohio. Many of the candidates campaigned against unfettered data-center construction and will likely seek to quickly deliver on what has become a wildly popular position across the country this year.

The new mandate is arriving just as technology investment has become more important for the economy than ever before. But the scale and speed of the boom have also inevitably raised concerns about an eventual bust.

"If some of these moratoriums or regulatory pressures slow some of this buildout, it almost prevents you from blowing too big of a bubble," said Brett Ryan, an economist at Deutsche Bank Securities. "This is like a good excuse to kill some projects here that were already looking shaky."

The paradox illustrates a central tension in the US economy of the post-pandemic era: Growth has become unusually reliant on a handful of sectors pouring enormous sums into a new technology with a payoff that's so far been difficult to quantify.

Thursday's selloff in AI-linked stocks also offered a reminder that the boom has helped support consumer spending through rising household wealth, a channel that could work in reverse if the trade unwinds quickly.

Over the last year, capital expenditures on information processing equipment and software alone have given the US economy the biggest boost on record, eclipsing even their contribution at the peak of the 1990s dot-com boom.

A report Thursday also showed broader economic growth is being propelled almost entirely by productivity enhancements, though economists said the prospective boost from AI is largely yet to appear in the numbers.

Monthly construction spending on data centers, meanwhile, has almost doubled over the last 12 months, to nearly $100 billion on an annualized basis.

For now, economists don't foresee an immediate pullback because much of the capacity that will come online next year is already under construction. This week's election results, however, could have a bigger impact on the outlook for 2028 to the extent that groundbreaking on new facilities begins to slow.

Ryan said Deutsche Bank had already been projecting slower growth in investment by then. Following Tuesday's elections, "my knee-jerk reaction would be, OK, I don't have to raise my 2028 forecast anytime soon," he said.

The impact will depend on the specific shape any new restrictions might take. About 45% of planned data-center capacity is in states where newly elected Democrats campaigned on such limits, according to Goldman Sachs estimates.

Texas alone accounts for 25% of the total, and Gina Hinojosa, the first Democrat to be elected governor there since 1990, has vowed to halt new approvals. Georgia and Ohio are also getting Democratic governors for the first time in decades. In the former, Keisha Lance Bottoms campaigned on a statewide pause, while in the latter Amy Acton has proposed stricter conditions on new projects.

States with restrictions in place prior to the midterms have so far opted for temporary measures rather than permanent bans. In July, New York became the first to announce a pause on new environmental permits for hyperscale data center construction for up to one year.

Many smaller localities, however, have gone further. Monterey Park, California; Harford County, Maryland; and East Rutherford, New Jersey have all imposed some form of outright prohibition on data centers this year, and several Ohio municipalities this week voted to adopt similar measures.

Sarah Bianchi, a strategist at Evercore ISI, said she expects state moratoriums would last no more than a year as policymakers decide what restrictions to put in place, including whether to grandfather existing projects or impose rules based on project size.

Given the breadth of the backlash, though, "It's not like it's going to be super easy to go somewhere else in this environment," she said.

Uploaded by Siow Chen Ming