• Kansas City Fed President Jeff Schmid said the AI infrastructure boom's financing and interdependence deserve macroprudential scrutiny, questioning whether commitments among hyperscalers, chip makers, data-center developers, utilities, and lenders are becoming excessively leveraged.
  • The Fed's May 2026 Financial Stability Report listed AI among the risks most frequently raised by market contacts, citing elevated AI-related equity valuations, debt-funded capex, and possible labor-market disruption.
  • The Chicago Fed has flagged "tail risk" in AI-adjacent bank lending, noting specialized data-center assets can be hard to re-lease if a major tenant exits or equipment becomes obsolete.

A Web of Commitments

Kansas City Fed President Jeff Schmid has a question for the market: is the artificial-intelligence buildout becoming too big to fail?

In August 2026, Schmid said the scale of AI investment should be compared with past episodes that created systemic problems. He specifically questioned whether commitments running from data centers to energy providers and surrounding communities were becoming excessively leveraged and vulnerable to a cascading failure. His point was not that AI has already achieved too-big-to-fail status, but that the growing web of commitments among hyperscalers, chip makers, data-center developers, utilities, lenders, and host communities could transmit a shock broadly if AI demand or funding weakened.

"The relevant precedent is not a claim that AI resembles a bank," Schmid said, according to people familiar with his remarks. "It is a warning about concentration, leverage, interconnected contracts, and contagion." The distinction matters: too-big-to-fail risk refers to the possibility that failure, retrenchment, or disorderly deleveraging would cause wider economic damage because the sector is deeply interconnected. A bubble, by contrast, refers to potential overvaluation and a sharp correction.

The concern is increasingly reflected in official financial-stability monitoring. The Federal Reserve's May 2026 Financial Stability Report listed AI among the risks most frequently raised by market contacts, alongside geopolitical shocks, oil shocks, private credit, and persistent inflation. Respondents highlighted elevated AI-related equity valuations, debt-funded capital expenditure, and possible labor-market disruption.

Tail Risk in Bank Lending

The Chicago Fed has similarly identified "tail risk" in AI-adjacent bank lending. In its analysis, stress at software firms could reduce infrastructure spending and then impair semiconductor suppliers, power companies, data centers, and ultimately creditors. The regional bank noted that specialized data-center assets can be hard to re-lease if a major tenant exits or equipment becomes obsolete — a structural vulnerability that does not show up in headline delinquency rates.

Delinquencies in the broader industrial-property category that includes many data centers were still only 1.6 percent as of the third quarter of 2025, according to the Chicago Fed analysis. But the concern is forward-looking: rapidly expanding and concentrated exposures can appear healthy before a demand or financing shock reveals their correlations.

The AI ecosystem spans a tightly connected investment chain. Large cloud and AI-platform companies purchase computing capacity and fund model development. Semiconductor and server suppliers provide chips, networking, and hardware. Data-center developers and operators construct specialized facilities. Electric utilities, power generators, grid operators, and renewable-energy developers supply the large power loads. Banks, private-credit funds, bondholders, and other investors finance construction and equipment. States and municipalities compete for projects, tax bases, jobs, and infrastructure investment.

There is no single company's financial performance, executive turnover, or restructuring to report. The central issue is whether concentrated investment, debt, long-term contracts, and common exposure to a small set of large AI buyers create system-level financial risk.

Economic Support and Pressure Points

AI investment is providing genuine near-term economic support, especially through spending on servers, software, data centers, chips, construction, and electricity infrastructure. The IMF reported that U.S. investment in information-processing equipment and software rose 16.5 percent year over year in the third quarter of 2025, and described AI-related investment as a significant contributor to U.S. GDP growth.

At the same time, the buildout can create pressure points. Data centers require large up-front spending on land, buildings, chips, cooling, networking, and power connections. If expected AI revenues disappoint, projects or financing structures could be impaired quickly. Banks and nonbank lenders may have correlated exposure through commercial real estate, construction loans, corporate lending, and private credit. A downturn could propagate across multiple counterparties.

Energy and grid limits are also becoming macroeconomic constraints rather than merely operational ones. The IMF estimates worldwide data-center capital expenditure could reach $6.7 trillion by 2030. Schmid said AI construction has increased prices for chips, computer hardware, building materials, and construction generally, reflecting sharply higher demand for investment infrastructure. Data centers cluster where land, transmission capacity, tax incentives, fiber, and power are available, concentrating both local benefits and local risks. Data-center demand is expected to more than double by 2030, driven largely by AI.

Policy Response Takes Shape

U.S. policy is increasingly focused on whether households should bear infrastructure costs created by large data centers. In March 2026, the White House announced a "Ratepayer Protection Pledge," under which participating major technology companies committed to cover the generation and delivery infrastructure required for their data centers rather than shifting those costs to residential customers.

Congress has also begun to address the issue. In September, the House passed the Ratepayer Protection Act by 417–3; the measure would require state regulators to consider whether major electricity users, including data centers, should bear incremental infrastructure costs. Consumer advocates argued its protections may be limited because it requires states only to "consider" such cost allocation.

Internationally, AI infrastructure is part of a broader competition for advanced chips, cloud capacity, energy, and strategic technological influence. The IMF says AI-driven demand is benefiting economies embedded in global technology supply chains, even as overall global growth faces risks from conflict, energy shocks, and trade fragmentation. The European Union's AI Act became broadly applicable on August 2, 2026, with transparency rules effective in August; high-risk-system provisions phase in later. The EU is simultaneously seeking to expand compute capacity through its AI Gigafactories initiative.

Uneven Effects Across Stakeholders

The effects are uneven. Consumers and ratepayers may benefit from investment, stronger grids, and innovation, but face concern that new power plants and transmission lines could raise electricity bills if costs are socialized. Host communities can gain construction employment, tax revenue, and related development, but may encounter demands on land, water, electricity supply, and local infrastructure.

Workers may see productivity gains and demand for technical, construction, engineering, and energy jobs, while automation could weaken demand for some existing occupations. The Fed's market contacts explicitly flagged labor-market risks from widespread AI adoption. Investors and lenders face concentrated bets and correlated lending that can amplify losses if revenue growth fails to justify the buildout. AI services are becoming inputs into software, research, logistics, finance, and public administration, increasing the practical consequences of a disruption in cloud or compute capacity.

Schmid's language invokes the post-2008 financial-crisis concept of institutions whose disorderly failure could threaten the broader system and compel government intervention. The dot-com boom offers a parallel: technology investment can generate enduring infrastructure and productivity benefits even when individual valuations and projects prove unsustainable. The global financial crisis offers another: individually rational contracts and funding decisions can collectively create hidden system-wide fragility when institutions and markets become mutually dependent.

The current AI cycle has a distinctive physical dimension. Unlike earlier software booms, it depends heavily on chips, specialized facilities, electricity, grid capacity, and long-duration construction contracts. That makes it closer to an infrastructure-and-finance cycle, not solely a technology-stock story.

Near term, AI capital expenditure is likely to remain a source of U.S. investment, construction activity, hardware demand, and power-sector growth. It may also keep upward pressure on selected input prices and complicate the inflation outlook that Fed officials must assess. In a main downside scenario, a material shortfall in AI demand, monetization, or funding availability could reduce spending by major AI customers, cascading through data centers, chip suppliers, energy contracts, construction developers, and lenders — precisely the transmission channel Schmid highlighted.

The most favorable long-term outcome is that productivity gains validate much of the buildout, while project finance, grid planning, lending standards, and utility rate design allocate risks to investors rather than taxpayers or household ratepayers. The less favorable outcome is a cycle of overbuilding, stranded specialized assets, stressed lenders, political backlash over energy bills, and pressure for government support.

The evidence currently supports heightened monitoring rather than a conclusion that AI is already systemically dangerous. But the concern is forward-looking, and Schmid's question remains unanswered.

Correction: An earlier version of this article misstated the Chicago Fed's reference period for industrial-property delinquencies. The 1.6 percent figure applies to the third quarter of 2025.