The global artificial-intelligence boom has entered a more difficult phase. After years in which investors treated enormous spending on chips, data centers and electricity as evidence of an unstoppable technological cycle, markets are now asking a less comfortable question: who will pay for the build-out, and how much public resistance can it withstand?
The immediate trigger was political. Democrats’ decisive gains in the US midterm elections—at least 40 House seats and seven Senate seats, according to Reuters reporting—raised expectations that Congress could impose tougher scrutiny on AI companies and slow data-center construction. The result was a technology sell-off that Reuters said erased roughly $3 trillion from the S&P 500’s market value, with infrastructure-related companies hit particularly hard.
From enthusiasm to an investment test
The market reaction does not mean the underlying technology has stopped advancing. It reflects a reassessment of the financial assumptions supporting it. Companies have committed vast sums to computing capacity before clear evidence that revenues will match the cost of electricity, specialized chips, land and debt.
Meta, for example, announced a further $15 billion bond offering to finance its AI expansion, while Marvell Technology agreed to acquire Lumentum in a $115 billion all-stock transaction, including debt. Such deals show that the sector still has access to capital. They also underline the scale of the wager: AI expansion increasingly depends on balance sheets, credit markets and infrastructure that communities must approve.
That dependence makes the sector vulnerable to higher interest rates and weaker growth. Reuters reported that the technology rout was broadening beyond individual stocks, while investors questioned whether central banks could continue raising rates if an energy shock weakened economic activity. The International Monetary Fund has estimated global growth at 3.1% this year and warned that a sharp fall in AI investment could reduce global GDP by 1.1% by 2028 in a downside scenario.
Why politics is becoming central
Opposition to data centers is not necessarily opposition to AI itself. Local residents and policymakers have focused on electricity demand, water consumption, noise, tax incentives and whether communities receive durable employment after construction ends. Democrats made data-center moratoriums or restrictions part of campaigns in several areas, turning infrastructure planning into a national political issue.
Industry supporters counter that the facilities can attract investment, strengthen domestic computing capacity and help the United States compete with China. They argue that delaying construction could push innovation and high-value jobs abroad. For technology companies, the central challenge is therefore not simply to persuade investors that AI is profitable, but to persuade voters that its benefits justify the local costs.
“The acute cause was fear that Democrats would slow the AI capex boom,” Reuters reported, describing the market’s reaction after the election.
That interpretation captures the market’s short-term fear, but it should not be mistaken for a complete explanation. Valuations were already vulnerable, returns on AI investment remained uncertain, and some expansion was being funded with substantial borrowing. The election exposed those concerns rather than creating them.
War and energy complicate the outlook
The AI debate is unfolding alongside an energy shock. Conflict involving Iran and attacks affecting shipping through the Strait of Hormuz have disrupted energy flows and introduced the possibility of higher oil prices. Reuters reported that oil rose after an Iranian official said enriched nuclear material would remain in the country, while outages across the region were still estimated at about 3.8 million barrels per day.
Energy prices matter directly to data centers, which require reliable electricity, and indirectly to the entire economy. Higher fuel costs can keep inflation elevated even as technology investment and consumer demand weaken. That combination would leave central banks facing a difficult choice: tolerate inflation, or maintain restrictive policy while growth deteriorates.
There are signs that the energy shock could ease. Oil prices had fallen sharply on hopes that US-Iran talks might prevent renewed attacks, and Saudi and Iraqi production had reportedly returned toward pre-war levels. But the volatility itself is significant. A technology sector that depends on long-term infrastructure planning is being forced to operate in a world where energy security and geopolitics can change project economics within days.
What comes next
The next phase is likely to separate durable AI businesses from speculative infrastructure spending. Investors will watch whether companies can generate cash rather than merely announce capacity, while lenders will scrutinize debt-heavy data-center operators. Policymakers will face pressure to establish predictable rules on permits, power use and environmental standards instead of relying on abrupt moratoriums.
Three outcomes remain possible. A negotiated easing of the Iran conflict and continued productivity gains could restore confidence. A prolonged energy crisis combined with tighter US regulation could delay construction and reduce global growth. Or the sector could adapt through more efficient models, improved chips and smaller systems that deliver useful services without requiring unlimited new capacity.
The central lesson is that AI is no longer only a technology story. It is now a test of political legitimacy, financial discipline and energy resilience. The boom may continue, but its next stage will be slower, more regulated and more dependent on proving that extraordinary investment can produce ordinary economic value.