Why America’s AI Dream Is Failing to Launch

Ryan Cunningham is the founder of Edgerunner Ventures, an early-stage energy-compute fund, and publishes the Machine Yearning Substack.

Kristy Loke is a MATS research fellow focusing on China’s AI strategy and international AI governance, and an incoming doctoral candidate at the Lau China Institute, King’s College London.

In 2022, American dominance in AI looked almost unassailable. The U.S. led in advanced models, semiconductors, capital and talent. Its closest competitor, China, appeared to be reeling from export controls that cut it off from the world’s most advanced process nodes, chips and memory technologies.

Four years later, the tide seems to be turning. Chinese models have gained major global token share; American technology giants are eyeing Chinese logic and memory chips; and America’s grand AI ambitions face growing public backlash.

During the first three months of 2026, local opposition blocked or delayed roughly $130 billion in data center projects. This spring, 71% of Americans surveyed opposed the construction of a data center near where they live. Texas, the fastest-growing state in data center construction, has frozen new grid-connected projects until further impact studies are completed. Data centers have become a bipartisan sleeper issue for the election cycle.

Meanwhile, American leaders appear disconnected from voters whose consent they ultimately require to build. As frontier labs prepare for initial public offerings and President Donald Trump urges communities to “let Data Reign,” firmwide AI adoption and visible productivity gains remain limited.

America’s problem is not a shortage of capital, chips or technical talent. It is the absence of an AI strategy that connects frontier capability to broadly shared economic gains, public legitimacy and the energy infrastructure required to deploy it.

We write as an energy-compute analyst and a researcher focused on China’s AI policy. From the vantage points of the bottlenecks America’s physical AI build-out currently faces and China’s multifaceted approach to AI development, the U.S.’s huge capital expenditures on AI mask a dire situation: The country is financing an industrial revolution most citizens currently want no part of.

That could still change. It will not change on the present terms.

The American AI Dream(s)

In the U.S.’s clearest articulation of AI priorities to date, America’s AI Action Plan focuses simply on “winning the race.” Policymakers are seeking to establish the U.S. AI stack as the “gold standard for AI worldwide,” though relevant measures of victory remain unclear. Advocates gesture toward job creation and higher standards of living but offer few tangible expectations for how Americans’ lives will improve.

What, then, is the “American dream” for AI?

Up to now, there have been two dominant versions. The first, from Silicon Valley, promises a world of infinite wealth and abundance achievable only through unprecedented investment in frontier labs and model training infrastructure. The cost of intelligence will fall so low that most forms of labor will no longer be required, freeing humans to bask in a life of leisure, intellectual curiosity and self-direction. The second, prevalent in Washington, is less a vision of what the dream is than what it cannot be: the nightmare of losing an arms race with China.

These visions emerge from different institutional cultures but converge on the same message: the monopoly of power. In the former, Silicon Valley leaders are anchored by the economic belief that “competition is for losers” and the quasi-religious belief that whoever first creates the machine God will capture “the light cone of all future value in the universe.” The latter vision treats AI principally as a tool for preserving American hegemony and fears apocalypse if that hegemony erodes.

In short, you have visions for a techno utopia if we get our way and a techno dystopia if we don’t. Each works well within its respective echo chamber. Neither speaks persuasively to the lived experience of people outside these systems.

For many Americans, the fundamental social compact—“if you work hard, you will be rewarded”—has been fractured by financial crises, wageless recoveries and a global pandemic. They have watched platform companies build business models linked to serious damage to children’s mental health. Now many of the same firms are asking to put a data center in their backyard. When labs pitch a “post-labor” economy, workers do not hear a promise of leisure; they hear consistent threats of a “permanent underclass.” After repeated broken promises that technology would universally improve their lives, they have little reason to accept the claim that this time will be different.

American leadership is promising either the moon or damnation. Americans are saying, “Prove it.”

Four Orientations, Not One

Compared to the U.S.’s fixation on frontier AI, China’s strategy is more multidimensional. Since 2020, Chinese leaders have organized science and technology policies around what Xi Jinping calls the “Four Orientations”: the technological frontier, economic development, major national needs, and the life and health of the people.

The framework appears repeatedly in high-level AI policy and discourse, signaling what Beijing is optimizing for—and just as importantly, what goals it considers insufficient in isolation.

China seeks frontier capabilities and has pursued legal and bureaucratic reforms intended to improve its innovation environment. But it has not treated frontier leadership as a singular AI strategy. That may have been out of necessity: Constrained in terms of advanced chips, computing power and capital, China needed a “good enough” strategy that balanced pragmatism with technological ambition. Yet the approach continued after the 2025 DeepSeek moment. Rather than treating frontier or frontier-adjacent model building as the sole national objective, Beijing continued to pursue industrial diffusion, technological sovereignty and governance alongside them.

Economic development is arguably where much of Beijing’s policy attention lies. The “AI Plus” initiative, reiterated and expanded in 2025, has little to do with superintelligence. It sets industrial targets, directing local governments and firms toward productivity gains and industrial upgrades.

The third orientation focuses on building secure core technological supply chains. After the first Trump administration’s 2019 campaign against Huawei, Xi told Chinese scientists and engineers that core technologies could not be “borrowed, bought or begged for,” but must be secured through domestic innovation. Chinese officials and bureaucrats subsequently intensified efforts to localize key parts of the semiconductor supply chain. China’s recent chip gains reflect in part the experimentation and investments this push set in motion.

The fourth orientation mandates that tech development has to improve both public well-being and quality of life. In the post-ChatGPT period, governance has remained a formal policy priority. As concerns about automation and job displacement grew, the State Council incorporated research and preparation for those risks into high-level planning, while regulators strengthened accountability requirements for workers managed by algorithmic systems. When AI companions came under scrutiny for harms they caused to young people and vulnerable users, China’s cyberspace regulator moved quickly to finalize some of the world’s most stringent rules.

While the government’s record on AI governance is far from perfect, and not all four orientations are equally transferable to the U.S., Beijing’s framework treats industrial development, sovereignty and governance as concurrent objectives rather than subordinating each to frontier competition. That contrast warrants reflection.

A Race Is Not a Strategy

China’s example shows that inclusive AI policy cannot consist merely of restricting competitors and accelerating domestic firms. A technology capable of creating significant economic gains, while imposing real social and physical costs, requires planning, coordination and harmonizing of competing goals.

For all of America’s advantages—and they are considerable—the U.S. faces substantial execution risk. The world’s best chips are of little use if developers cannot permit facilities to house them or build the generation capacity to power them.

Nor should AI dominance be pursued in isolation. It may conflict with openness, safety, privacy, labor protections and international restraint. A credible American AI dream cannot be solely defined by winning an ambiguously defined race: It must be measured by whether ordinary Americans gain capability, security, opportunity and genuine freedom, including from avoidable technological harm.

The U.S. needs a more holistic, pro-social definition for its AI dream. This is not anti-innovation; this is antifragile. Domestically, it means transparently incorporating local feedback into infrastructure planning, preparing workers for disruption rather than treating displacement as an abstraction, and giving the public concrete reasons to believe AI’s gains will be broadly shared. Globally, it means engaging constructively with partners and competitors, including China, to establish a shared account of AI’s most desirable outcomes and a common framework for managing its most consequential risks.

If the U.S. wants its promise of AI-enabled abundance to survive contact with voters, it needs a pragmatic and inclusive strategy fast—or its dream will die on the launchpad.