Examine public statements from U.S. Treasury officials, SEC commissioners, congressional leaders, and international regulatory bodies…

Former U.S. Treasury Secretary Janet Yellen publicly described “bubble-like activity” in the AI sector in statements around late 2025/early 2026, explicitly comparing it to crypto excesses, while current FSOC (chaired by Treasury Secretary Scott Bessent) has launched targeted monitoring via an interagency AI working group and Innovation Series without declaring an imminent systemic threat.[1]

This reflects a pattern across U.S. officials: acknowledgment of valuation and debt pressures in AI infrastructure (data centers, hyperscaler capex) alongside emphasis on governance frameworks and innovation support rather than alarmist warnings of imminent collapse.

• Yellen’s remarks (e.g., on CNN’s Erin Burnett) highlighted circular financing and hype-driven valuations reminiscent of prior bubbles, noting risks from concentrated investment without proportional near-term revenue realization.
• In December 2025, FSOC announced an interagency AI working group to explore AI for financial resilience while monitoring stability risks from adoption “within and outside” the sector; a January 22, 2026 Senate Democrats letter urged it to formally investigate AI-related debt growth and potential bubbles, partnering with the Office of Financial Research.[2]
• Treasury released an AI Lexicon and Financial Services AI Risk Management Framework in February 2026 focused on secure adoption, lexicon standardization, and risk controls (e.g., identity, fraud, explainability). A November 20, 2025 letter led by Rep. Bill Foster (D-IL) and 21 House members called for FSOC to assess vulnerability to a “sharp drop in the value of AI-related assets and infrastructure” and include findings in its 2025 Annual Report.[3]
• FSOC’s AI Innovation Series began with a March 4, 2026 roundtable on strategy and governance principles, stressing practical scaling of AI while preserving safety.

These actions signal proactive vigilance on concentrated AI capex and debt (hyperscalers projected ~$700 billion in 2026 spending) without labeling it a full systemic bubble, implying regulators view it as a monitorable concentration risk rather than an uncontrollable one.[4]

SEC leadership has focused on disclosure integrity and operational risks rather than broad valuation-bubble warnings, with former Chair Gary Gensler highlighting a stark spending-revenue gap while current Chair Paul Atkins stresses governance and AI-enabled oversight tools.[5]

Gensler (speaking March 31, 2026 at MIT Sloan) noted AI infrastructure capex reached $400 billion in 2025 and was projected at $500–600 billion in 2026, against only ~$50 billion in direct generative AI revenues, creating “disequilibrium” with “more downside risk than upside potential” due to high fixed costs and uncertain payback.

• Current SEC Chair Paul Atkins, at the March 4, 2026 FSOC AI roundtable, outlined a posture of understanding AI’s market implications, deploying it for risk assessments, fraud detection, and efficient disclosures, while forming an internal AI Task Force in August 2025.[6]
• SEC 2026 Examination Priorities dedicate attention to verifying AI capability claims (“AI washing”), model governance, human oversight, and risks in automated tools, trading algorithms, and AML functions—prioritizing alignment between representations and actual processes over macro valuation critiques.[7]

This approach treats AI valuation exuberance primarily through the lens of investor protection and market integrity rather than immediate systemic stability threats.

Congressional leaders, primarily Democrats, have pressed FSOC for formal assessments of AI investment and debt risks, framing potential sharp corrections as a financial stability concern warranting contingency planning.[3]

• The November 2025 Foster-led letter explicitly sought expert analysis of “vulnerability of the U.S. financial system to a potential sharp drop in the value of AI-related assets and infrastructure.”
• The January 2026 Senate Democrats letter referenced the new FSOC AI working group and called for investigation into AI debt dynamics, citing circular financing and self-reinforcing investment loops.

These efforts have contributed to FSOC’s creation of the working group and roundtable series, indicating congressional influence in elevating the issue without standalone legislation or hearings producing major new findings by mid-2026.

International bodies (FSB, IMF, BIS) have integrated AI-related risks into broader financial stability monitoring, emphasizing concentration, model opacity, and potential correction channels while noting AI’s productivity upside; none issued a dedicated 2025–2026 white paper solely on AI valuation bubbles.[8]

• FSB’s October 2025 report Monitoring Adoption of Artificial Intelligence and Related Vulnerabilities in the Financial Sector updates its 2024 analysis, flagging third-party dependencies, market correlations, cyber risks, and model governance as stability-relevant vulnerabilities. It calls for better data and monitoring but does not quantify valuation or overinvestment risks.[8]
• IMF’s October 2025 Global Financial Stability Report notes AI investment supporting U.S. growth but warns that reassessment of productivity expectations could trigger abrupt investment declines, market corrections, negative wealth effects, and contagion—implicitly acknowledging stretched valuations without using “bubble” terminology.
• BIS publications (including its 2025 AI-for-policy report) highlight AI’s value for early-warning indicators of asset-price bubbles and stress, while noting supply-chain concentration risks that could amplify synchronized behaviors; a related April 2026 speech by ECB’s Philip Lane references “speculative growth” and the “AI bubble” in a euro-area context.[9]
• Related ESRB (December 2025) analysis examines AI’s potential to amplify systemic risks via common exposures, herding, and opacity.

For market participants or new entrants, the regulatory posture implies continued tolerance for AI innovation provided firms strengthen governance, disclosures, and resilience against concentration or model risks; explicit bubble warnings remain largely from former officials and market analysts rather than sitting regulators, suggesting any correction would likely prompt supervisory scrutiny rather than immediate intervention. No single formal report exclusively titled around “AI valuation risk” has emerged, but the topic is embedded in FSOC, FSB, and IMF workstreams. Additional primary-source releases from the FSOC working group (expected post-March 2026 roundtables) would further clarify official risk calibrations.

---

Recent Findings Supplement (May 2026)

ESRB’s December 2025 report explicitly flags high AI-provider valuations as a trigger for sharp market corrections and questions whether AI overinvestment creates durable infrastructure.[1]

The ESRB Advisory Scientific Committee’s December 2025 paper (No. 16) details how concentrated high valuations in major AI firms could amplify systemic stress through sudden repricing, while noting that AI capex—unlike past infrastructure booms—often lacks long-lived physical assets that retain value after a correction. It identifies AI-specific mechanisms (model monocultures, overreliance on identical systems, and rapid herding via automated trading) that could turn a valuation shock into correlated failures across institutions.

• High market valuations of major AI providers “could trigger sharp market corrections” posing “additional financial risk.”[1]
• Overinvestment in AI “may not create tangible infrastructure with long-term social benefits” compared with railroads or fiber optics.[1]
• Model uniformity and overreliance risks are highlighted as new systemic amplifiers not fully captured in traditional stress tests.

This raises the bar for any new AI-finance entrant or regulator: monitoring must now include real-time valuation stress tests and concentration metrics rather than relying solely on traditional leverage or liquidity indicators.

BIS March 2026 Quarterly Review quantifies the shift to debt and off-balance-sheet financing for AI infrastructure, raising the stakes for any valuation correction.[2]

In its March 2026 Quarterly Review, the BIS documented that hyperscalers issued $120 billion in bonds in 2025 (a fivefold increase from 2024) and moved another $120 billion off-balance sheet through SPVs and private credit to fund data centers. This marks a clear transition from cash-flow-funded capex to leveraged structures, directly linking AI-sector spending to broader credit and shadow-banking channels.

• Hyperscalers’ 2025 bond issuance reached $120 billion; additional $120 billion routed via SPVs/private credit.[3]
• Elevated valuations and rising capital-spending guidance have already produced volatility and sector rotation away from AI leaders.[4]
• P/E ratios for the broader market approached dot-com levels even after corrections in big-tech names.[4]

For market participants, this means any AI-driven earnings disappointment now carries direct transmission risk through private-credit and SPV structures that were previously absent.

FSB’s February 2026 work programme elevates AI monitoring and “sound practices” guidance to a 2026 priority.[5]

The FSB’s 2026 work plan explicitly schedules a report on sound practices for AI adoption, use, and innovation in financial services, while continuing assessments of third-party dependencies, concentration, and cyber risks first flagged in its 2024–2025 analyses. This moves AI from “watch item” to active supervisory workstream.

• 2026 deliverables include a dedicated report on AI sound practices and ongoing evaluation of financial-stability implications.[5]
• Emphasis on closing data gaps around AI model concentration and third-party criticality.[6]

Regulators and firms entering AI-related finance must now prepare for forthcoming FSB guidance that will likely require enhanced disclosure of model provenance and concentration exposures.

IMF May 2026 analysis shifts focus to AI-enabled cyberattacks as a new systemic channel, while January 2026 WEO flags technology-expectation reevaluation as a downside risk.[7]

In a May 2026 blog, the IMF warned that advanced models (citing Anthropic’s Claude Mythos) can dramatically reduce the time and cost of exploiting financial-system vulnerabilities, enabling correlated failures across banks, payment systems, and cloud providers. The January 2026 World Economic Outlook separately lists “reevaluation of technology expectations” among key downside risks to growth.

• Extreme cyber-incident losses could trigger funding strains, solvency concerns, and market disruption.[7]
• Interconnected digital infrastructure amplifies single-point failures.[8]

This reframes AI-sector risk from pure valuation to operational-cyber transmission, requiring entrants to demonstrate robust AI-specific cyber-resilience controls.

U.S. Treasury and SEC statements post-November 2025 have centered on AI cyber and governance risks rather than direct valuation-bubble warnings, with limited new quantitative guidance.[9]

Treasury Secretary Scott Bessent and Fed Chair Powell convened bank CEOs in April 2026 to discuss risks from advanced AI models, and Bessent publicly highlighted AI cyber threats in May 2026. SEC exam priorities and Chairman Atkins’ March 2026 remarks emphasize governance, disclosure accuracy, and oversight of AI tools used by registrants, but do not introduce new sector-wide valuation-stress metrics.

• No new Treasury or SEC white paper or formal guidance specifically quantifying AI overinvestment systemic risk has been released since late 2025.
• FSOC under Bessent has narrowed its systemic-risk focus in favor of growth-oriented policies.[10]

For U.S.-focused players, the immediate regulatory pressure remains on operational resilience and accurate AI-related disclosures rather than macro-prudential valuation caps.