The next US data-center wave needs community stakes and accountability. McKinsey estimates demand could rise from 25 gigawatts in 2024 to more than 80 gigawatts by 2030.

Hyperscalers, the large technology operators expanding computing infrastructure at scale, face a capital problem. Capital expenditure—the money spent on construction and equipment—is huge, but no spending total accompanies the forecast. Demand estimates do not guarantee operating capacity; financing and permitting still determine what gets built.

Spending on artificial intelligence (AI) accounted for almost one percentage point of real US gross domestic product (GDP) growth in the first half of 2025. Goldman Sachs warned: “Without it, U.S. GDP growth would have almost flatlined.” The measure is broader than data-center construction, but it shows why infrastructure policy now carries macroeconomic stakes.

Public acceptance is the constraint. A September Pew Research study found environmental concerns were the leading factor behind negative views of data-center construction, followed by home energy costs and quality of life for nearby residents. The survey identifies concerns, not the number of projects rejected.

Jeremy Siegel, emeritus professor of finance at Wharton, offers a more benign market view: if one community rejects a data center, another may welcome the investment, jobs and tax revenue. That outcome requires developers and governments to move across jurisdictions; neither mobility nor guaranteed tax gains is demonstrated.

A community-consent test for capital

Stavros Gadinis, a UC Berkeley law professor and faculty director of the Berkeley Center for Law and Business, says history offers a better playbook. In an exclusive interview at Web Summit in Lisbon, he argues that large investors should impose terms before committing capital.

• Start with outreach to affected communities, nonprofits and academics. Local communities matter, but consultation should not stop there.
• Give affected parties a stake in development, meaning a direct interest in project decisions and outcomes, rather than consultation alone.
• Condition the money. Large investors should set requirements before funding, and builders can proceed only after those terms are met.
• Require accountability. Companies should show how they respected local policies and goals, what progress they made and what they contributed.

Gadinis does not specify who grants a stake, what investors can demand or how compliance gets verified. Those omissions matter because he places capital behind the conditions; without standards, outreach could become another public-relations exercise.

Public pushback is spreading rapidly across the United States, including Texas, where Elon Musk's SpaceX merged with his AI-focused xAI earlier in 2026. Gadinis says conflicts and permit delays are emerging in Texas and could produce comparable gridlock elsewhere.

The author of Corporate Ordering argues neither AI companies nor the White House wants a breakdown, because continued compute capacity is a strategic advantage. He put the stake plainly: “The biggest advantage that we have against our global competitors is our ability to build compute, of which data centers are a very important part.”

Compute means processing capacity, and data centers underpin it. His prescription is negotiated common ground, not a construction freeze.

Liability becomes the regulatory fault line

The Democratic landslide in the midterm elections is likely to bring more scrutiny to companies that enjoyed significant freedoms under the Trump administration, analysts said. The market slid last week on those prospects. No index level or percentage decline is given, so the causal link remains an interpretation rather than a measured result.

AI companies and policymakers in Washington, D.C. have traded positions over regulation. Some founders seek government rules, while Treasury Secretary Scott Bessent has insisted that government should not become a liability shield, meaning legal protection, for technology giants. The exchange does not settle who should pay when AI systems fail.

Gadinis argues that whether government or industry moves first matters less than the foundations of the framework. He expects practice to settle first, after which Washington will turn the emerging norm into a common rule. He does not define the foundations or specify which practices qualify, leaving the sequence broad.

He goes further on systemic risk, the danger that one concentrated failure could destabilize wider finance. In his assessment, AI investment has become so large relative to GDP that AI companies became too big to fail almost instantly in 2023. That is an expert assertion, not a government designation, and no financial exposure is quantified.

He finds it very hard to imagine Washington refusing a backstop, meaning emergency support, during a crisis, particularly because government has elevated AI into a national-security concern. He compares the potential scale with subprime mortgages and says AI is looming large.

The demand for rules, in his account, reflects shared exposure: society is already involved whether it likes it or not. That supports a warning about concentration, not proof that public liabilities will follow.

What does this change for data-center developers and investors?

For developers, local acceptance belongs in planning before construction, not only in communications after announcement. The package includes stakeholder outreach, a community stake, investor-set conditions and evidence of compliance. Investors would need to test those conditions before committing capital, rather than treat goodwill as a public-relations line item.

For policymakers, the harder question is explicit: who bears legal responsibility when AI investment fails, and what would a government backstop mean in cost, timing and legal authority? Gadinis offers a sequence and a warning, not a legislative framework, leaving execution, taxpayer exposure and enforcement unresolved.

Until those answers become specific, the practical change is procedural rather than legal: capital should wait for a documented community bargain and a credible account of who bears the downside. The sequence remains a proposal, not an enacted rule.