For decades, Americans have grown accustomed to computers and other electronic devices getting cheaper, at least relative to their quality. Now, thanks largely to the artificial intelligence boom, that trend is reversing. This month, the Federal Reserve Bank of San Francisco reported that prices for information-processing equipment had risen 13 percent over the past year—the sharpest increase since the government began tracking the category in the 1960s.
At the heart of the problem is AI’s rapacious appetite for memory chips. Data centers have been amassing them in staggering numbers, squeezing supply for automakers, medical device manufacturers, cybersecurity firms, and consumer electronics makers. In 2020, data centers consumed 32 percent of dynamic random-access memory chips. By 2030, AI data centers are on pace to consume 60 percent of the world’s memory chips.
That competition is sending prices soaring. Memory chip prices have quadrupled in just the last year. In a letter to the Trump administration, industry and trade associations warned of an “urgent imbalance in the market for memory chips that could lead to significant and sustained near-term price increases for American households and disrupt critical U.S. supply chains.”
The effects of that “chipflation” are already reaching consumers. Apple raised prices by as much as 20 percent on some of its most popular laptops and tablets, which it attributed to the “hundred-year flood” of memory chip prices, as former CEO Tim Cook put it. Automakers have been losing out in the contest with big AI firms over memory chips, with some absorbing higher costs and signaling potential vehicle price increases. If worsening chipflation causes production delays and higher auto prices, it could lead to a replay of the post-pandemic years, when a semiconductor shortage sent car prices soaring and accounted for a full third of heightened inflation across the whole economy in 2021.
The scale of the AI infrastructure boom helps explain the shortage. Companies spent less than $900 million per month building data centers in December 2022, but by June 2026, they were spending more than $4 billion each month.
Meanwhile, memory chipmakers have raked in massive windfall profits. In September, Samsung reported a ninefold increase in quarterly operating profits from a year earlier, while Micron’s profits were up more than 1,000 percent. Those windfalls reflect the memory industry’s shift toward producing specialized chips for price-insensitive, flush-with-investment AI buyers. Industry analysts expect the memory chip shortage to last until at least 2027, when more domestic capacity is expected to come online at Micron’s new Idaho fabrication plant. Even then, there is little stopping rampant AI demand from consuming most of that new supply. Without intervention, chipflation could last for as long as the AI investment boom does.
The government need not settle for merely waiting out chipflation. As I detailed in a recent paper, it already has tools to tame AI-driven inflation. The Defense Production Act is one of the government’s most powerful and flexible industrial policy tools for addressing supply imbalances. The DPA gives the president authorities that could restrain chipflation by ensuring that hyperscalers do not consume a disproportionate share of scarce chips. For example, under the DPA, the president could give non-AI chip buyers priority for memory chips. Or he could create a government-facilitated “buyers’ club” to purchase chips on behalf of non-AI buyers. Or he could require memory chipmakers to allocate a certain portion of their output to non-AI buyers.
These demand-side policies to prevent AI buyers from distorting the memory chip market are essential in the near term—building new chip fabs or expanding existing ones to meet demand takes years and billions of dollars. But the DPA can also help chipmakers bring new supply to market. Under the law’s authorities, the government can expedite construction of chipmakers’ domestic capacity expansions and help them transition production lines to make standard dynamic random-access memory (DRAM) chips for non-AI buyers. By speeding up new capacity, both supply-side steps would also help cool chipflation.
Failure to act could lead to a double whammy for affordability, with consumers facing both higher inflation and higher borrowing costs. The Federal Reserve may respond to inflation fueled by AI investment by keeping interest rates higher for longer. In that scenario, because hyperscalers are overheating the economy, homebuyers and consumers would pay more in interest for big-ticket purchases.
Chipflation isn’t the only economic distortion of the AI buildout. Data center construction is increasing competition for land, electricians, construction workers, and HVAC technicians needed for new housing construction. Economists also warn that spending on AI data centers may be crowding out other investment and construction in the economy.
The danger resembles a form of “Dutch Disease,” in which a booming sector draws resources away from other parts of the economy. If the U.S. allows frenzied investment to concentrate resources in AI at the expense of other industries, it could face a similar problem.
But it’s not too late to change course. The Trump administration has enthusiastically promoted untrammeled AI development. But at what cost? As the AI boom begins driving up prices for everyone else, it may be time to assert some measure of democratic control over the pace and scale of AI development. The government has the tools to intervene. Until it does, Americans will keep paying for an AI boom whose benefits have yet to arrive.
The post How to Tame AI-Driven Inflation appeared first on Washington Monthly.
