Fed Governor Lisa Cook: Software Investment and Data Center Construction Delaying Inflation Return to 2% Target
Fed Governor Lisa Cook stated that software investment and data center construction are delaying the return of inflation to the 2% target. She pointed out that AI-related investments over the past year have contributed to inflation remaining persistently above target, with short-term demand shocks preceding productivity realization.
Cook noted that prices for chips, computers, and software have surged significantly, partly reflecting a concentration of demand in AI-related industries rather than a synchronized increase in overall economic demand. She emphasized that data center construction relies on cross-industry shared inputs such as construction labor and energy, and related investments will transmit price pressures to other sectors. Companies have announced about $2 trillion in related capital expenditure plans, of which only a small portion has been realized, with further demand still in the pipeline.
She pointed out that electricity and water prices have each risen by about 5% over the past year, and core goods prices have been running at an annualized rate exceeding 3% this year. A significant portion of the stock market's rise has been attributed to enthusiasm for AI, with the wealth effect seeping into household spending. Productivity improvements are expected to bring only moderate deflationary effects in the coming years, insufficient to offset the price pressures spreading in the latter part of this year.
She distinguished between two types of prices: relative price changes in industries are not the target of monetary policy, as the tools are too blunt; however, there are signs that demand expansion for shared inputs is spreading to the entire economy. The labor market may undergo a painful transition, and future adjustments to policy rates will depend on the effects of previous rate hikes and upcoming inflation and employment data.
She also mentioned that the oil price transmission and supply chain disruptions caused by the Middle East conflict will compound with AI construction, continuing to exert inflationary pressure in the coming months. Inflation has been above the 2% target for over five consecutive years, and she believes the employment sector can currently withstand interest rate hikes better.
In market mechanisms, buyers are large-scale cloud providers and tech companies, which are using equity premiums, investment-grade bonds, and operating cash flows to rapidly build computing power and software stacks; sellers include electricity, water, chips, servers, construction subcontractors, and utilities. This is an event-driven capital expenditure wave, with funds being drawn from general manufacturing, housing, and interest rate-sensitive sectors towards computing infrastructure. Beneficiaries include chips, power equipment, and the data center chain, while those under pressure include the power grid, residential electricity costs, core goods, and end employment that still needs to wait for productivity realization.
Additionally, the announced plans far exceed the portion already spent, indicating that the peak demand has not yet arrived; the 5% increase in electricity and water costs and the annualized rate exceeding 3% for core goods are observable interfaces of pressure spillover.
Source: Public Information
ABAB AI Insight
Lisa Cook has previously highlighted AI infrastructure as an independent inflation shock: in earlier speeches, she referenced over $1.5 trillion in data center plans with a very low realization rate, and listed the price increases of chips, high-tech equipment, software, and utilities as demand shocks outside the second layer of supply. She has consistently prioritized the credibility of the inflation target over employment risks, supporting the maintenance or increase of policy rates during periods of price stability, rather than waiting for the productivity narrative to automatically materialize.
The capital path is to "first seize physical bottlenecks, then discuss total factor productivity." Large-scale companies lock funds into land, transformers, cooling, chips, and software licenses using high-valuation equity and investment-grade debt, preempting the entire society's lendable funds and grid capacity for computing power. The motivation is not to maximize current profits, but to secure scarce quotas for training and inference; strategically, it is about discounting future cash flows to today's installation rights. The result is that neutral interest rates are raised during the construction period, squeezing out housing and traditional capacity.
Similar paths have been seen in the shale oil capital expenditure boom and the 5G base station cycle: the input side first pushes up prices for steel pipes, spectrum, electricity, and labor, while the output side's efficiency lags by several years. The current industry is in an expansion and installation phase, not yet entering the cost curve decline and control harvesting phase. NVIDIA, power equipment manufacturers, and independent power assets gain pricing power, while grid operators and residential users bear the spillover.
Structurally, this is a transfer of pricing power: scarcity temporarily shifts from labor to electricity, advanced packaging, and software licenses. The mechanism is that the supply curve steepens on shared inputs, while monetary policy targets aggregate demand, failing to address transformer delivery times, thus pushing the 2% target further back due to the construction cycle itself.
ABAB News · Cognitive Law
Before capacity is built, the tech narrative first adds inflation
Shared bottlenecks determine prices, not visions
Productivity is a long-term option, demand is immediate cash