• New York Fed President John Williams says AI infrastructure demand is pushing up chip prices, construction costs and inflation, with supply unlikely to catch up for several years.
  • Rates: One more hike in 2026 may be appropriate; no additional hikes currently expected in 2027.
  • Inflation: Expected to approach 2% in 2027.
  • AI: Strong demand could keep chip prices elevated.
  • Economy: Labor markets remain stable, with no broad overheating.
  • Treasuries: Higher long-term yields are tightening borrowing conditions.

AI Boom Inflates Prices, But No Broad Overheating

The artificial intelligence investment boom is creating inflationary pressure in the U.S. economy, but it is not yet causing broad overheating, according to New York Federal Reserve President John Williams. In a speech late last month, Williams said that soaring demand for AI infrastructure—from semiconductors to power equipment and construction resources—is pushing up costs and contributing to inflation. He expects supply to catch up only after several years, allowing price pressures to eventually ease.

“AI-related demand is a new source of inflation,” Williams said, according to prepared remarks. “We are seeing upward pressure on prices for chips, electrical components, and construction services, but this is not translating into generalized overheating.”

Supply Constraints and Inflation Outlook

Williams’s baseline forecast sees overall inflation at about 3.5% in 2026, just above 2% in 2027, and 2% in 2028. That assumes energy prices normalize, tariff effects fade, and AI-related supply and demand become better balanced. He also projected real GDP growth of about 2.25% in both 2026 and 2027, with unemployment edging down toward 4%.

The central bank official emphasized that while AI investment is boosting demand for key inputs, it is not yet spilling over into broader wage or price inflation. “Labor markets remain stable,” he said, “and there is no evidence that these shocks have become broader or persistent.”

However, other Fed officials are increasingly concerned. Chicago Fed President Austan Goolsbee said in an interview updated October 13 that AI-related inflation is “jumping out of its lane,” citing competition for electricians, construction resources, and electrical components. Goolsbee stopped short of declaring the aggregate economy overheated, but his comments underscore growing unease about the breadth of price pressures.

Rates: One More Hike Possible, But No Commitment

Williams said one further rate increase “may be appropriate late this year,” conditional on the economy following his forecast. He added there was “no need for urgency” after September’s quarter-point increase, which brought the federal-funds target range to 3.75%–4%. His published forecast does not include additional hikes in 2027, but he cautioned that this is not a Fed commitment.

Market pricing and policymaker projections suggest the path is less settled. Bloomberg’s October 12 analysis reported an 85% probability of at least one hike by December and nearly an 80% probability of at least two hikes by March 2027. The September “dot plot” showed 16 of 18 participants anticipated at least one additional hike in 2026, with four expecting two. Eight participants saw rates 50 basis points higher by the end of 2027.

“The dot plot does not specify every move during the year,” Williams noted. “It is a conditional outlook.”

Market and Economic Implications

The AI investment race is massive. Bloomberg estimates that Meta (META), Amazon (AMZN), Microsoft (MSFT), and Alphabet (GOOGL) will together spend roughly $750 billion on capital expenditures this year. Data-center memory prices are more than five times year-earlier levels, and power-equipment backlogs extend beyond 2030. These pressures are concentrated in specific inputs, but Williams warned they can transmit to the wider economy because AI infrastructure goods are inputs into other consumer and business products.

“Higher costs for chips and construction can raise prices outside data centers,” he said. “That said, inflation is driven by multiple forces, including tariffs and energy disruptions, not AI alone.”

Long-term borrowing conditions are also tightening. Williams noted that higher long-term Treasury yields are contributing to tighter financial conditions. Recent reporting links the AI boom, energy shocks, and fiscal spending to a higher-rate global environment, though the exact contribution of AI to yield increases is unclear.

Industry and Policy Context

This is not a company-specific story, but the corporate spending race among large cloud and technology firms is central. Their investments extend well beyond software and chips into transformers, gas turbines, networking equipment, and real estate. However, large technology companies may be less sensitive to higher rates than ordinary borrowers, given their profitable existing businesses and strategic determination to build AI capacity.

Williams also highlighted other inflation drivers: earlier tariffs contributed to inflation, though they are no longer adding to goods-price inflation in his assessment. Middle East conflict and refining constraints are raising oil and refined-fuel costs, and he expects energy effects to be larger and longer-lasting than previously anticipated.

He also pointed to immigration policy and population aging as reasons labor-force growth is contributing less to U.S. economic growth than during the late-1990s technology boom. “Interest rates cannot directly manufacture chips or reopen refineries,” he said. “Tightening is a way to prevent these disturbances from becoming entrenched.”

Historical Parallels and Future Risks

Williams compared today’s AI boom to the internet expansion of the late 1990s. Then, GDP growth exceeded 4%, labor-force growth exceeded 1%, and productivity growth averaged about 3%. Today, he sees slower labor-force expansion and productivity growth only just above 2%, with a larger imported component of AI investment limiting its direct contribution to U.S. GDP.

Goolsbee recalled that internet infrastructure was built around optimistic growth assumptions, leading to overbuilding when growth fell short. “A useful technology can coexist with excessive investment and disappointing financial returns,” he said.

Two competing risks remain: inflation persistence if AI bottlenecks spread, and investment reversal if expected returns disappoint. Governor Lisa Cook said on October 1 that AI-buildout pressures might not resolve quickly and were a principal concern for 2027. Meanwhile, Bloomberg Economics expects the Fed to remain on hold for the rest of 2026 and cut by 25 basis points in the second half of 2027, contrasting with both the Fed’s projected additional hike and tighter market expectations.

As of October 13, 2026, Williams’s published forecast puts inflation just above 2% in 2027 and at 2% in 2028. His comments on long-term Treasury yields and multiyear supply bottlenecks are not established by his September 29 speech; separate reporting documents multiyear equipment bottlenecks, but that is not the same as verifying those exact remarks. The Fed did not respond to requests for comment.