The U.S. financial sector is undergoing a fundamental transformation in its approaches to project financing. The largest banks and investment firms on Wall Street can no longer ignore the voice of local communities when making decisions about lending for data center construction. Reputational and operational risks associated with resident protests are now woven into the fabric of credit analysis alongside traditional financial metrics.

My analysis shows that this is not a temporary measure but a structural shift. Lenders are beginning to de facto rank jurisdictions by their level of "social loyalty" to such projects, and states with a tougher stance from residents automatically receive less capital inflow. This creates a new geography for AI infrastructure investment.

Lenders tighten due diligence standards

A key criterion in reviewing applications is not only the technical readiness of the project but also its social license. As Karen Fang, head of infrastructure finance at Bank of America, emphasizes, support from the future neighbors of a facility is a mandatory condition for launch. JPMorgan notes that the loan approval process stretches over years, and any disruption during the construction phase results in enormous costs for the bank and a revision of covenants.

Morgan Stanley CFO Sharon Yeshaya rightly points to the capital intensity of the sector, which requires syndicated schemes and guarantees of full placement. However, now a mandatory analysis of social risks is added to these mechanisms, making deals more expensive and slower.

Scale of the problem: numbers and consequences

The statistics are telling: in the first quarter of 2026 alone, 75 projects with a total value of about $130 billion faced organized resistance from local communities. Meanwhile, the appetite of giants like Microsoft, Meta, and Amazon shows no signs of weakening—their combined future lease obligations have already exceeded $1 trillion, and Goldman Sachs' forecasts for AI spending through 2030 ($6 trillion) look frighteningly realistic.

Cases where banks are already involved in conflict situations are illustrative. The organization of a $12.3 billion bond issue for the BlackRock and Meta partnership in Texas, a $9.7 billion credit line for CyrusOne in Illinois—these are all examples of how financial institutions balance between profitability and reputational costs. The closure of the QTS project in Virginia served as a wake-up call for the entire market: even a powerful sponsor like Blackstone could not overcome resident resistance.

Adaptation strategies and new risks

The industry is seeking workarounds. Building their own power plants, including 7.65 GW gas giants from Amazon and small nuclear reactors from Crusoe, is an attempt to reduce the burden on local infrastructure and thereby weaken protest potential. However, this is only a partial solution. The introduction of the first state-level moratorium in the U.S. in New York demonstrates that the regulatory environment is becoming increasingly hostile.

Investors are already pricing the risk of project cancellations into asset values, understanding that the shortage of computing capacity will persist. But I see a systemic threat here: if banks begin to massively discount such projects, it could trigger a cooling of the entire AI infrastructure market, which would hit the pace of technology adoption. Ray Dalio's comparison to the bubbles of 1929 and 2000 no longer seems like an exaggeration—we are now witnessing how the social factor is becoming a catalyst for a potential correction.

My professional judgment: banks are effectively shifting the responsibility for social approval onto developers' shoulders but are not reducing the required returns. This creates a dangerous imbalance that, in the medium term, could lead to stagnation in supply amid growing demand for AI capacity.