Too big to pause: Could an AI slowdown crash the economy?
Calls to slow development of the most powerful AI models are colliding with an estimated $800 billion buildout by US hyperscalers, raising questions about the economic fallout of any coordinated pause. Industry leaders, some lawmakers and international figures are split between proposals to ‘pace’ frontier models for safety and political pushes to continue rapid development to maintain competitiveness, particularly with China.
Why It Matters
AI investment is already a large and growing driver of US GDP and markets, so a sustained pause in frontier model development could trigger market repricing, reduced corporate investment and tighter credit with measurable effects on growth. Conversely, continued deployment of existing models could sustain productivity gains even if training of new models slows.
Key Facts
- Estimated AI hyperscaler investment: $800 billion this year (Goldman Sachs estimate)
- SoftBank funding moves: Launched a $10 billion and €1 billion bond sale to fund OpenAI investment; had invested ~$54.6 billion in OpenAI by end of July
- St. Louis Fed finding: AI-related investment contributed 0.97 percentage points to US real GDP growth in the first three quarters of 2025, accounting for 39% of total GDP growth through Q3 2025
- Legislative action: ‘Ban Artificial Superintelligence Act’ introduced Sept. 23 would prohibit superintelligence and pause advanced AI development until federal safety rules are set
- Voluntary safety accord: On Sept. 29 President Trump and leading AI executives signed a voluntary safety agreement emphasizing internal controls, audits and oversight
The US AI sector is facing a tug-of-war between safety-focused proposals to slow frontier model development and political and commercial pressures to accelerate deployment. Anthropic CEO Dario Amodei has proposed measures to ‘pace the frontier’—including independent evaluations, shared safety standards and limits on developers in democratic countries—with endorsements from figures such as Sam Altman, Demis Hassabis and Elon Musk. By contrast, President Donald Trump has rejected a coordinated slowdown, arguing rapid development is necessary to compete with China and establishing a ‘Super Intelligence Force’ led by former SEC chair Jay Clayton.
The stakes are large: Goldman Sachs projects roughly $800 billion of spending by major US AI hyperscalers this year, and major corporate financings continue to flow—for example, SoftBank’s recent $10 billion and €1 billion bond offerings to back its OpenAI stake, after having invested about $54.6 billion through July. A St. Louis Fed analysis attributed nearly one percentage point of US real GDP growth to AI-related categories in the first three quarters of 2025, representing 39% of total GDP growth through that period, underscoring how entwined economic activity and AI investment have become.
Analysts and international institutions warn that halting or reversing AI investment could have material macroeconomic effects. The IMF in April estimated that an AI-investment reversal could knock US equity markets down about 20%, tighten credit and leave US GDP 1.5% below baseline, while a scenario presented by Fitch envisages a 35% equity shock coupled with capital expenditure retrenchment producing a US recession. Those risks hinge on changes in investor and corporate expectations: a slowdown could prompt cancelled infrastructure plans, repricing of AI assets and reduced lending.
There is also a countervailing view that much economic value will continue to flow even if frontier training slows. Executives in the AI ecosystem argue that widespread adoption and inference demand for existing models and agentic systems will continue to drive productivity gains across industries for the near to medium term. Whether the net outcome is contractionary or merely a reorientation of investment depends on how policymakers, firms and markets respond to safety incidents, regulatory moves and evolving expectations about AI’s productivity trajectory.
Keep Reading
AI chatbots recommend shopping at Walmart over just about anywhere else, analysts find
AI glasses face their first major government crackdown
Formula One to probe software bug behind Bahrain GP chaos in Sepang
