The massive inflows of capital into data centers amid the frenzy of AI infrastructure investment have generated a significant pushback in the United States by local opponents, leading to the cancellation of data-center projects worth at least $85 billion over the past three years. In July, New York became the first state to enact a moratorium on new data centers, as other states and the federal government consider similar proposals. But concerns about data centers’ economic and environmental costs, while reasonable, are only one part of the story. Everyone still must ask: Will the US miss out by not building more data centers?
As Josh Zoffer, a former official at the National Economic Council under President Joe Biden and now an investor in the AI sector, recently put it, data centers are a “crucial test of US industrial resolve.” They highlight the uncomfortable truth that building strategic capacity in the future usually requires sacrifice in the present. Failure to recognise this reality has led US policymakers to bungle similar buildouts in the past. Two cases stand out.
In the 1970s, a global oil shock forced the US to acknowledge that its energy dependence was a source of strategic vulnerability. In a 1977 speech, President Jimmy Carter warned Americans that building a secure and diversified domestic energy base would require the US economy to go through a phase of “higher costs” and “greater inconvenience.” But Carter’s willingness to address reality, rather than sugarcoat it, did not have the desired effect.
As economist Jeffrey Currie argues, Carter’s successors drew a different lesson. They avoided admitting scarcity, and responded to supply shocks instead by talking prices down and drawing on strategic reserves. Still, they did not build real energy security; they bought time, but delayed necessary investments.
Today’s tensions around the Strait of Hormuz serve as a reminder of the need for the energy transition that Carter urged. Even though the vast reserves of shale gas that have been accessed since then mean that the US is no longer as energy dependent as it was in Carter’s time, it is still exposed to oil shocks, as rising petrol prices show.
By contrast, China has spent decades pursuing electrification, which Jeff Currie calls “the purchase of optionality.” The resulting flexibility—an electron, Currie writes, “can be sourced from coal, gas, sun, wind, or uranium”—has helped cushion China from the current energy crisis.
The US also missed an opportunity to avoid reliance on China for rare-earth minerals. In the second half of the 20th century, the Mountain Pass Mine in California was a key node in US domestic rare-earth production and processing capacity, ensuring a stable and self-sufficient supply chain. By 1999–2000, however, operations had been drastically scaled back, owing to environmental concerns, regulatory changes, and lower-cost Chinese producers, on which US industry was relying for more than 90% of its rare-earth needs, according to the US Geological Survey.
Both examples underscore an important shortcoming of US decision-making. Markets are efficient at pricing the costs of doing something, but very inefficient at pricing the costs of not doing it. As a result, markets fall short when returns have a social component, such as economy-wide investments in secure supplies of energy and rare-earth minerals.
To help economies successfully steer, coordinate, and underwrite investments with a cross-market scope, Mariana Mazzucato and Dani Rodrik have argued for public-private partnerships in their work “Industrial policy with conditionalities: a taxonomy and sample cases.” They clarify that conditionalities are crucial to success: the government must create incentives to drive private firms’ decision-making toward otherwise unsought outcomes.
Such a public-private deal could be designed around the government providing private firms with one or more benefits—loans, tax incentives, accelerated permitting—in exchange for two types of conditionalities: directionality and risk/reward. Directionality would mean requiring AI hyperscalers to address local concerns by internalising environmental and social costs (such as technical-data sharing, water management commitments, and co-investing in grid buildouts). Risk/reward would entail that, where the public underwrites part of the risk, it also shares part of the returns. For example, this could be done through excess-profit sharing, equity-like instruments, or access to computing power.
Compared to the recent past, policymakers are already in a better position to appreciate data centers’ strategic importance. The global economic regime has changed, clarifying the value of such infrastructure. After World War II, the world was focused on rebuilding together. Integration was the goal, and globalisation the means. Today, geoeconomics dominates a more fragmented landscape. Power now rests on the ability to control strategic nodes.
That means policymakers are already more inclined to see strategic investments through a cross-market lens. From that perspective, the strategic relevance of data centers is less uncertain than energy optionality or rare-earth processing once appeared to be. Nvidia CEO Jensen Huang has described AI as a five-layer cake: data centers sit inside the infrastructure layer, highlighting their economy-wide relevance.
Of course, the problem of channeling investment to strategic industries is not confined to the US. In his 2024 report on European competitiveness, former European Central Bank president Mario Draghi recognised breakthrough innovation, defence-industrial capacity, and cross-border grids as public goods “that will be undersupplied without common action.”
Building data centers must become a bipartisan issue. The success of public-private partnerships will hinge on their design. Too many conditions, and innovation might be stifled; too few, and private market players get to have their cake and eat it, too.
© Project Syndicate 1995–2026
.png)
