The rush to invest in artificial intelligence could leave a more lasting mark on chip and equipment prices than expected. John Williams, president of the Federal Reserve Bank of New York, believes that the production capacity needed to meet demand for AI infrastructure may not reach equilibrium within two years. It is a significant signal for US monetary policy, placing the AI boom firmly on the terrain of physical supply: semiconductors, components, and equipment required to build and operate compute capacity.
In an interview with Bloomberg, Williams explained that the expansion of AI infrastructure is impacting the prices of chips and the goods that incorporate them. Asked whether supply and demand could balance out within two years, he said he expects it to take longer, though he conditioned this outlook on the current demand momentum holding up. The distinction is substantial: this is not a definitive estimate of how long shortages will last, but rather an assessment of a scenario in which spending on AI capacity continues to grow aggressively.
An industrial constraint within the inflation picture
The remarks from the New York Fed president shift attention away from an exclusively financial narrative around AI and toward its physical infrastructure. Deploying these systems requires specialized hardware and a supply chain capable of delivering components and equipment in adequate volumes. When orders concentrate around supply that cannot scale at the same pace, prices can remain under pressure for longer, and the impact can ripple out to products that rely on those components.
For a central bank, this does not automatically translate into a fresh, broad-based flare-up in inflation. However, it does mean that a segment of the economy tied to tech investment can generate supply-side friction, right as the Fed assesses how long monetary policy must remain restrictive to bring inflation back to target. Williams explicitly linked the scarcity of AI equipment to potential price pressures, without quantifying the extent or singling it out as the sole deciding factor for interest rate decisions.
Williams's perspective also matters because of his role on the Federal Open Market Committee, the body that sets Federal Reserve interest rates. His view is that the economic impact of AI should not be evaluated solely through the promise of future efficiency and productivity. During the infrastructure buildout phase, demand can instead run into tangible industrial constraints, with knock-on effects on costs. Expected productivity gains and price pressures can therefore coexist over time, without the former immediately offsetting the latter.
The message on rates remains cautious
The commentary on supply constraints came alongside a reaffirmation of Williams's monetary policy stance: in his view, a further rate hike later this year could be appropriate. He did not frame this move as a certainty. The phrasing leaves the door open to incoming data on inflation and economic activity, but signals that the risk of financial conditions not yet being sufficiently restrictive remains part of his assessment.
Looking to 2027, Williams did not say he currently sees reasons to anticipate further increases. Here too, the message is not a binding forecast: the subsequent outlook will depend on how economic conditions evolve and how prices actually behave. The distinction between a potential move before year-end and the lack of guidance thereafter helps frame his words for what they are: conditional guidance, rather than the announcement of a preset path.
The combination of resilient AI demand and constrained hardware supply thus adds another layer to the rate debate. If investments continue unabated, a potentially higher cost of borrowing may have a limited impact on spending by the companies and investors most exposed to this transformation. Williams noted that investors in the sector appear relatively insensitive to the cost of capital, as they are targeting productivity gains and profound shifts across the economy.
Expensive capital, sustained demand
This assessment does not imply that interest rates are irrelevant to the sector. Williams himself ruled out that demand is completely indifferent to monetary policy. His reasoning is more specific: the conviction that AI has vast economic potential can make a portion of investments less sensitive to changes in the cost of money compared to other projects. In a market with insufficient production capacity, the persistence of orders can lengthen the time needed to normalize availability and prices.
This leads to a practical takeaway for businesses and markets. Chip availability is not the only variable to monitor: equipment tied to computing power and other physical components required for projects also matter. Those planning expansions in AI may have to grapple with supply constraints and high costs for longer. For investors, Williams's thesis suggests that the intensity of infrastructure spending should not be confused with the speed at which such investments will yield available capacity on more favorable terms.
Supply pressure can also affect operators attempting to repurpose existing infrastructure for AI computing services. Cointelegraph notes that Bitcoin miners moving in this direction find themselves competing for chips and electrical equipment, while also facing substantial financing needs and tighter credit. Last July, IREN announced new contracts for AI cloud services worth $2.8 billion; the case illustrates how the line between digital infrastructure and capacity dedicated to AI is becoming an industrial issue, not just a technological one.
However, an important caveat remains in Williams's assessment: his scenario depends on continued demand. A slowdown in investments, shifts in the plans of major companies, or faster-than-expected supply growth would alter the balance. Currently, his statements provide neither an estimate of the shortage's scale nor a precise timeline for overcoming it. The core message is that two years might not be enough, not that the market is destined to remain in imbalance for a fixed duration.
In the coming months, attention will therefore focus on two sets of signals: the trajectory of AI investments and the data the Fed will use to gauge inflation. If infrastructure demand remains robust and supply continues to play catch-up, the issue could weigh on the inflation outlook and the debate over the monetary trajectory. If, on the other hand, production capacity or investments change pace, Williams' hypothesis will also need to be recalibrated.
