Tracking the AI Prize: Trends and Updates

Welcome to Dispatch Markets! The public conversation about AI in America is saturated with worries about how data centers affect nearby communities and what automation might mean for the job prospects of upcoming graduates. These concerns are legitimate and deserve attention. Yet they shouldn’t be the central question.

When AI was merely an idea on the horizon, discussions around it carried a tone of awe and optimism. We pictured all the transformative possibilities technology could unlock for the world and for human progress. As we’ve moved into the era of usable AI, the mood has shifted. Things like wonder or even guarded optimism are less common in today’s dialogue.

If we fix our gaze solely on constraints or the downsides tied to AI, we miss the bigger picture. Meanwhile, China appears to fully grasp the scale and seriousness of the technological upheaval we’re facing. What matters most is recognizing that whoever leads in the AI race will shape the trajectory of the future. The ultimate objective should be American AI leadership with the broader aim of steering humanity toward a liberal framework rather than an authoritarian one.

The usual policy debate around AI starts by identifying tangible harms and asking how to lessen them. Data centers impose burdens on local communities. Advances in AI model sophistication escalate cyber risks. Automating clerical “laptop” tasks threatens millions of white-collar workers. Each concern is real and important, but addressing them with the current approach yields suboptimal results.

Our traditional playbook aims to minimize harm and regard foregone capability as the price of risk mitigation. Yet sufficient capability isn’t guaranteed. We should pursue a national mission that prioritizes maximizing the growth of capabilities, treating harms as constraints to be measured, mitigated, engineered around, or compensated for, rather than letting them halt progress.

The central question is whether AI will empower free and open societies—grounded in liberal ideals (in a broad sense)—or bolster a censored, surveilled, totalitarian order rooted in rigid ideology. This may sound like alarmism, but it isn’t.

The world is increasingly divided into two dominant powers, the United States and China, each pulling in its own direction. The advance of artificial intelligence raises the stakes in this already-existing divide, not because AI might go wrong in countless ways, but because the scope of things that can go right is vast.

All other concerns—though real and valid—should be secondary to this overarching issue. Our primary orientation toward AI should emphasize its potential successes rather than its possible failures.

AI lowers the cost of thinking. When it functions well, AI can shorten every gap—from the moment we suspect a problem to taking effective action, and from a spark of inspiration to its full realization. That kind of leverage makes AI a force multiplier, perhaps greater than any innovation since the invention of the printing press.

Ukraine has, for nearly two years, relied on AI-assisted guidance to keep strike drones on target despite Russian jamming attempts. The U.S. Department of Defense’s Project Maven uses AI to transform vast streams of data into one coherent battlefield picture. This enables commanders to identify threats more quickly, enhancing speed, precision, and ultimately the effectiveness of responses. In cyber operations, the Cybersecurity and Infrastructure Security Agency is applying cutting-edge models to hunt bugs that hostile foreign services would otherwise uncover.

Let’s examine the most concrete debate about how AI could falter: where data centers should be located. In the race for AI dominance, these questions have become chokepoints. They will determine whether America’s industrial base for AI grows quickly enough to stay ahead or whether delays and cancellations will allow rivals to overtake us. Any discussion of the harms associated with data-center development is incomplete unless it is paired with a plan for mitigating those harms while maintaining the maximum feasible rate of compute expansion. Today’s approach is almost the opposite, and policymakers at every level lack both a roadmap and the political will to update it. The following three cases illustrate how well-meaning measures to prevent harm can complicate the path for developers. The United States must treat each as a constraint on the pace at which capital, energy, labor, and land can be converted into national computing power.

Northern Virginia, the epicenter of the U.S. data-center surge with roughly 250 facilities, cannot escape local resistance. Early last year, Loudoun County ended its longstanding rule that allowed construction to begin once county administrators certified compliance with zoning and building codes. Now every project must undergo a legislative review that could add months, or even years, to the approval timeline. A process that once guaranteed compliance now invites public hearings, discretionary votes, and all the unpredictability of local politics.

In San Jose, California, a proposed Microsoft data-center campus faced more than five years of delays, with repeated hearings before the local planning commission and city council, and demands to revise the environmental impact study mandated by state law.

At the Susquehanna nuclear facility in Pennsylvania, Amazon attempted to address the AI power issue by placing the data center adjacent to the generator. Yet, statutory requirements required the Federal Energy Regulatory Commission to assess whether such an arrangement could divert power from the regional grid, who would bear the losses if that power were diverted, and whether it would affect reliability or consumer costs. In November 2024, the commission rejected the amended interconnection agreement, leaving the project stranded.

Taken together, these three cases show how reasonable harm-prevention measures can turn the path to data-center development into a labyrinth for developers. The United States must treat each instance as an additional constraint on the pace at which it can turn capital, energy, labor, and land into national computing capacity.

Nearly every capability of AI can be used for offense or defense. The same models that enable a criminal syndicate to orchestrate cyberattacks on a large scale also allow a company to scan its code for vulnerabilities before they become breaches. Both sides reap the multiplier. The task is to turn that multiplier into a lasting strategic edge before the other side does.

Trying to curb the future by constraining model capabilities will weaken our defenses, while adversaries—whether state actors like China or other criminals—will have a more effective toolkit. Cyber policy must ensure that defensive deployments outpace offensive exploitation.

Likewise, concerns about worker displacement should be addressed, in part, by policies that align with the proportional productivity gains that typically accompany workforce reductions. If a task that once required 10 workers can now be done with seven, that is both a displacement issue and a 30 percent improvement in output per worker.

Historically, higher productivity has been associated with higher wages. But suppose the transition is so rapid and widespread that workers can’t adjust quickly enough to capitalize on those gains. Even in that scenario, higher productivity still translates into greater GDP, which in turn yields more tax revenue. A country of displaced workers will still need to generate tax income.

Retraining carries a price tag. So do the health and income support that workers may require during the shift. Returning to work may demand transportation, broadband access, relocation aid, or a combination of those. The funds must come from somewhere. A policy aimed at limiting displacement will end up constraining its own ability to address labor-market pressures that will inevitably surface.

Even postponing America’s AI progress imposes strategic costs. The models and industrial base that reach scale first will shape the infrastructure and standards the rest of the world follows. If the U.S. cannot overcome domestic hurdles to AI advancement—or, frankly, to AI dominance—we risk ceding the future to a competitor that can prevail.

The way forward.

We require a framework aimed at ensuring U.S. AI leadership, with the objective of growing our AI capabilities faster than China’s. Indicators like data-center capacity and the lead time for frontier models provide clues, though they don’t capture every factor, they signal whether policy is moving in the right direction.

The potential harms should stay part of the analysis. They impose discipline and act as constraints on how quickly the U.S. can sustain progress. That’s substantial. But they are not, by themselves, a set of independent targets that we inadvertently pay for by slowing down our AI capability growth.

Even economists, trained to weigh good and bad outcomes, are not accustomed to thinking in terms of national imperatives. For example, an environmental economist studying climate policy typically frames the goal as reducing carbon emissions as efficiently as possible, even if the final reduction isn’t exactly what policymakers might wish.

Such an economist has never been taught to prioritize minimizing emissions relative to the maximum possible expansion of accessible energy. Yet that is precisely the type of directive we need to secure American dominance in AI.

Sixty-five years ago, Americans watched in astonishment as Yuri Gagarin became the first human to leave Earth’s atmosphere. Our response to that event became legendary. We did, after all, beat the Soviets to the Moon. Yet achieving victory in the space race required an extraordinary mobilization of resources and a singular focus.

We are now engaged in a comparable race with China. The stakes today are far higher.

If America is to prevail in the AI competition, AI policy must undergo a broad re-prioritization. There are many potential pitfalls as AI advances, as there were during every major technological shift—from steam power to the Internet.

There are moments when the fate of liberalism itself hangs in the balance. One could argue that this was true during the threat of war with Japan and Nazi Germany, certainly during the Cold War with the Soviet Union, and it is the case today.

A China that becomes AI-dominant will soon wield military and economic supremacy. It will eventually control the infrastructure that underpins modern life and, with that power, may impose censorship and surveillance around the world. We cannot allow that outcome.

*Author’s note: Today we understand the term “gung ho” to be a U.S. Marine Corps battle cry used in popular speech to signal vigorous support. Historically, it stems from the Chinese phrase gōngyè hézuòshè, meaning “industrial cooperation.” During World War II, its shortened form gōnghé, meaning “work together,” was overheard by Maj. Evans Carlson, who adopted it as his battalion motto, and the phrase spread through the Marine Corps and American society.

America’s major airline hubs are operating at capacity, and the incumbents are urging Congress to expand the $12.5 billion already allocated to upgrade the aging air-traffic control system in the wake of the fatal collision near Reagan National Airport involving an Army Black Hawk helicopter and an American Airlines jet last year. Electra.aero isn’t waiting for Congress or the FAA to act.

Electra has created and successfully tested a 17-passenger point-to-point aircraft that uses blown-lift tech to take off from runways as short as 150 feet. Its hybrid-electric propellers run quieter than typical airplanes, making it feasible to take off and land from airfields embedded within urban areas. Such point-to-point travel could eliminate the large-scale congestion issues associated with major hubs.

The 2022 CHIPS and Science Act authorized $280 billion in spending to briefly triple total investment in manufacturing facilities. Yet employment in manufacturing hardly moved. This reinforces another lesson: industrial policy is not typically effective growth policy.