Overview
The AI buildout is reshaping the global economy, but it also complicates monetary policy. Businesses are spending heavily on acquiring chips, building data centers and consuming energy. Yet, the technology’s full productivity gains depend on firms integrating it into their workflows, a process unfolding unevenly across the economy. That mismatch in spending and adoption could influence inflation and employment outlooks facing central banks, whose policy choices then influence the financing and customer demand supporting further AI development and adoption. AI is shaping the economic conditions that will, in turn, shape its expansion.
The risk is that the scale of the AI buildout creates economic pressures that monetary policy alone cannot resolve. If central banks raise rates to combat an overheated economy caused in part by AI spending, it remains possible that high expected returns and substantial internal funding could sustain expenditures by major AI investors, shifting the slowdown to businesses and households whose spending responds more readily. If AI automation displaces workers faster than new jobs emerge, lower rates may not quickly restore employment or resolve the mismatch between workers’ skills and available jobs. And if revenues disappoint after years of heavy investment, debt distress could spread through lenders and investors, restricting credit and weakening spending enough to turn an AI bust into a global downturn and potentially prompt emergency central bank support.
The AI–Central Bank Feedback Loop
To develop the AI economy, businesses spend on infrastructure and workers, generating income and demand throughout their supply chains. Competition for scarce resources, including electricity, advanced semiconductors and skilled construction labor, can push up costs before AI’s full productivity benefits emerge. As businesses integrate AI into their operations, productive capacity and employment change. Firms can potentially produce more with fewer inputs, while automation reduces demand for some tasks and creates or expands others. These shifts alter the inflation and employment outlook. Spending that grows faster than productive capacity can increase inflationary pressure; productivity gains can ease it, although expectations of higher profits and incomes can also encourage spending. In a November 2025 op-ed, now Fed Chair Kevin Warsh argued that AI would be a “significant disinflationary force,” urging the Fed to reconsider its forecasts for weak growth and persistent inflation. How far productivity gains reduce inflation also depends on competition in the industries adopting AI, as firms with pricing power may retain more of the cost savings as higher margins instead of passing them on through lower prices.
Central banks develop their monetary policy, in part, in response to the impacts of AI on the economy, which could increase global GDP by up to 15% above its baseline by 2035. Central banks seek to make the distinction between AI sectoral price increases and persistent, economy-wide conditions before deciding on policy. For instance, at the Fed’s September 2026 meeting, participants argued that higher rates could keep price pressures from AI demand and energy disruptions from spreading across the economy and becoming persistent. Central banks also consider AI’s ability to alter the economy’s productive potential and the neutral interest rate: the rate that, after accounting for inflation, neither stimulates nor slows the economy. As AI shifts that benchmark, keeping the policy rate unchanged can still make monetary policy more or less restrictive. The banks then set policy on interest rates, asset-purchase policies and communication about future decisions, all of which influence financial conditions.
Central bank policy choices then feed back into the financing and demand supporting the AI economy. AI developers face changes in the funding available for research and infrastructure, while their customers face changes in sales and budgets for software, integration and training. The effect depends partly on funding structure, with cash reserves and previously secured fixed-rate finance being able to cushion immediate borrowing-cost changes, while floating-rate debt and upcoming refinancing are more exposed to changes in policy. These conditions influence which AI projects firms accelerate, postpone or cancel.
These decisions then reshape AI’s economic impact. Changes in AI spending affect demand now, while changes in useful development and adoption affect productive capacity and employment over time. Slower investment could reduce immediate inflation pressure while also postponing future productivity gains. As that balance changes the inflation and employment outlook, central banks face a different set of AI-related pressures, and their subsequent decisions again influence AI financing, development and adoption, completing the loop.
AI’s Economic Stability Risk
The feedback loop becomes difficult for monetary policy to manage when AI creates inflation, employment and financial pressures at the same time. The following three risk scenarios could unfold separately or reinforce one another.
Scenario One: Heavy spending on AI infrastructure can push up demand for resources shared with other industries beyond available supply, spreading cost pressures across the economy and potentially prompting central banks to raise rates if those pressures threaten persistent inflation.
Higher rates, however, could slow the rest of the economy before they slow the AI buildout. There is a broad market concern that AI spending is more rate-insensitive than other portions of the economy. Many of the large AI spenders currently have substantial cash flows, secured financing and high expected returns, so they may keep spending even as households and other businesses cut back. Their reliance on debt is nevertheless growing, with the largest five hyperscalers in the US accounting for $222 billion in debt issuance in 2026 through September 22. If that spending sustains broader inflation pressure, bringing inflation down could require a sharper slowdown elsewhere. Businesses could consequently face high input costs and falling sales at the same time.
The slowdown could also delay the productivity gains expected to ease inflationary pressures, with businesses facing weaker sales or tighter budgets deciding to postpone adoption. Infrastructure construction could then outpace adoption, leaving providers with more capacity than customers are ready to use. Developers would then wait longer for revenues, and their customers would wait longer for cost savings.
Financing pressure could also increase dependence on a small number of technology companies. Independent developers struggling to raise capital elsewhere may turn to firms that supply both funding and computing capacity. For example, cloud providers often require the AI developers they fund to spend a substantial share of that funding on their cloud services. If tighter financing strengthens that dependence, developers could have less freedom to choose suppliers, and their customers could become more reliant on the same few platforms. If this reduces competition, AI providers may face less pressure to lower prices, leaving other businesses with higher adoption costs and smaller savings.
Scenario Two: If AI displaces workers faster than new jobs are created, lost income could weaken household spending and business sales, potentially prompting central banks to lower rates to support demand and hiring.
In this scenario, some unemployment may reflect a gap between workers and available jobs. Displaced workers and new labor market entrants may lack the skills employers need or live far from growing industries. Lower rates can encourage spending, but they cannot quickly resolve these mismatches in a structurally different economy. Some businesses could struggle to fill jobs while others lose sales as household spending weakens. If central banks overestimate how much of this unemployment reflects weak demand, they could ease rates too aggressively, pushing up spending faster than firms can expand production. Fed Governor Lisa Cook has warned that lowering rates to address unemployment caused by a mismatch between workers’ skills and available jobs could fuel inflation.
Scenario Three: An AI investment bust could prompt emergency central bank action if financial losses threaten to trigger a broader economic downturn.
AI investment could fall sharply if investors lose confidence that future revenues will justify the cost of the buildout. The projected US AI infrastructure buildout, for example, must generate an estimated $3.55 trillion in annual revenue by 2032, equivalent to roughly 9% of projected US GDP, to deliver a 10% annual return before leverage. Open-weight models could make AI services cheaper for customers while intensifying price competition for developers of proprietary models, making it harder for those developers to recover their investment. And while data centers can be built in one to two years, the transmission and power generation to get them online take longer, resulting in data centers housing “dark GPUs” that are not operational, tying up capital before the equipment can generate revenue. Firms with large debt and lease payments could then have to cancel projects, sell assets or default.
Potential losses could spread through the financial relationships connecting AI companies. A Bank of International Settlements study found 46.4% of the disclosed value of investment deals between AI companies from 2021 to 2025 involved firms that were also suppliers or customers of one another. A company’s failure could hit its partners through lost sales, investment losses and, where contracts require them, payments under guarantees. Banks and investment funds could also suffer losses on several connected borrowers. Growing use of private credit and complex financing structures can also make it harder to identify debt burdens and who would bear losses if AI projects disappoint. This could complicate central banks’ assessment of the scale of potential losses and the financial system’s ability to absorb them. If those losses become large enough to make lenders cut credit to businesses outside AI, the downturn would spread. Falling stock prices could further weaken spending, while international investors and suppliers carry the effects abroad. If an AI bust occurs while inflation remains high, central banks could face pressure to support spending while still needing to contain price increases, likely through targeted emergency lending.
