The narrative surrounding artificial intelligence is shifting from "software eating the world" to "hardware hitting physical limits." In the highly polarized U.S. political landscape, few issues can unite far-left Senator Bernie Sanders and far-right Governor Ron DeSantis—except for curbing the expansion of data centers. This phenomenon represents not just a political spectacle in Washington but a harsh "physical reality check" for Wall Street. As Silicon Valley giants brandish checkbooks larger than the Apollo moon program's budget, attempting to sustain the AI boom through computational power, they are colliding with barriers erected by political resistance and electrical grid limitations.
Lawmakers in New York have already proposed legislation to impose a minimum three-year moratorium on permits for new data center construction and operation. New York is at least the sixth state considering such a pause. From community protests in Florida to regulatory freezes on Texas's power grid, a market-overlooked risk is intensifying: if physical grid capacity falls short and political conditions turn unfavorable, the hundreds of billions in capital expenditures factored into valuation models may never materialize.
When Sanders and DeSantis Align Sanders and DeSantis stand diametrically opposed on most issues, yet they share a rare consensus on data center proliferation: it must be slowed. This cross-party opposition stems from tangible public grievances over AI's side effects. Across the U.S., 24/7 low-frequency noise from data centers disturbs neighboring communities, massive cooling demands strain local water supplies, and soaring electricity bills burden residents and small businesses, fueling growing public backlash.
Governor DeSantis's abrupt policy reversal exemplifies this political shift. In June 2025, he signed a major tax relief bill extending data center tax credits from 2027 to 2037. However, facing rising public pressure, DeSantis has since pivoted. "We don’t want to subsidize technologies that replace human experiences," he stated recently, advocating for an "AI Bill of Rights" and supporting legislation requiring data centers to cover their full utility costs. He emphasized that local communities shouldn’t bear the cost of expansion for "the richest companies in human history."
This rhetoric echoes Sanders, who previously warned that technology won’t improve workers' lives if decisions are made solely by billionaires focused on short-term profits. He explicitly urged Congress to pass a moratorium on new data centers: "We need to slow this process down." Legislators in Arizona, Georgia, Virginia, and other states are advancing bills to revoke tax incentives or ban non-disclosure agreements that hide details from the public. In Georgia, Oklahoma, and Vermont, lawmakers have proposed direct construction halts, mirroring Sanders's call. For tech giants, the era of red-carpet treatment is over.
Can Massive Capital Expenditures Be Deployed? If political resistance is a "soft constraint," grid limitations pose a deadlier "hard wall." Wall Street faces a perplexing paradox: Can the projected $600 billion in capital expenditures expected by 2026 actually be spent? Latest data shows Microsoft, Meta, Amazon, and Google alone plan $670 billion in AI infrastructure spending this year. Relative to U.S. GDP, this surpasses the Apollo program and the Interstate Highway System, second only to the 1803 Louisiana Purchase. Amazon alone aims to boost its capital expenditures by nearly 60% to $200 billion this year.
Most of these funds target data center construction, which requires enormous energy. BloombergNEF projects data center power demand will triple by 2035, surging from 34.7 gigawatts (GW) in 2024 to 106 GW—equivalent to powering 80 million homes. The problem: the current U.S. grid cannot meet this demand. In Texas, this physical constraint has triggered a regulatory crisis. As the nation’s second-largest data center hub after Virginia, the Electric Reliability Council of Texas (ERCOT) is implementing unprecedented freezes on projects.
ERCOT has proposed re-evaluating projects totaling 8.2 GW of power consumption—equivalent to eight nuclear reactors—including many previously approved initiatives. Its "Batch Zero" review mechanism assesses projects in phases to gauge grid impact. Meta’s energy project manager, Katie Bell, noted some submissions have waited 18 months without meeting Batch Zero criteria. This uncertainty undermines tech expansion plans: if the grid can’t connect, data centers won’t be built; if data centers aren’t built, the $670 billion budget goes unspent; and if funds aren’t spent, anticipated AI compute growth and commercialization will evaporate.
The "Physical Correction" in Wall Street’s Crowded Trade As the "unspendable money" risk gains recognition, financial markets are reacting sharply. U.S. stocks recently saw their fourth-largest single-day sell-off in momentum shares in a decade. Even independent power producers (IPPs) and nuclear energy stocks—once seen as AI boom beneficiaries—have slumped. The initial logic that "AI power shortages benefit utility stocks" has shifted: if grid connections stall, new electricity demand won’t materialize.
UBS analysis notes that IPPs like Constellation Energy have plunged 27% year-to-date over concerns that new loads can’t secure contracts with existing generation capacity. Markets now recognize that without grid expansion, power generation alone is meaningless. This panic has sparked a rise in "anti-AI trades," with capital flowing out of high-beta tech stocks into defensive sectors like chemicals and regional banks. The sell-off reflects deleveraging driven by quant and active funds.
Currently, markets must resolve a vexing paradox: either believe the grid will miraculously expand to absorb $600 billion in capex, or admit physical limits have been hit. If the latter holds true, it implies no grid build-out, no capital deployment, and no chip demand—ultimately bursting the AI super-cycle valuation bubble. With Sanders and DeSantis uniting politically and ERCOT shutting the door physically, Wall Street appears forced to accept the second scenario.