America's $500bn wager on yesterday's technology
Trump's AI Action Plan builds massive data centers just as China proves efficiency trumps scale
As "God Bless the USA" echoed through Washington's Mellon Auditorium, President Trump took the stage flanked by technology luminaries including Nvidia's Jensen Huang and declared America would become an "AI export powerhouse" through eliminating regulatory red tape. The audience of venture capitalists erupted as Trump signed three executive orders accelerating artificial intelligence development whilst rolling back Biden-era safety requirements.
Yet beneath this political theatre lies a contradiction threatening the very technological leadership the administration claims to champion. The AI Action Plan promises American dominance through deregulation whilst simultaneously proposing infrastructure investments that historically required sustained government intervention—the precise institutional approach the plan rejects.
The scale of ambition versus historical reality
The numbers reveal staggering ambition. The Stargate project promises $500 billion over four years—requiring electrical generation capacity equivalent to 30 nuclear power plants. Goldman Sachs forecasts global data center power demand rising 165 percent by 2030, demanding $6.7 trillion investment according to McKinsey. Deloitte projects AI data centers consuming 123 gigawatts by 2035, up from 4 gigawatts today.
Yet delivery track records suggest systematic overreach. Wisconsin's Foxconn deal promised $10 billion investment creating 13,000 jobs before being scaled back at $231,000 per job created. The CHIPS Act authorised $52 billion for semiconductors but has disbursed $7 billion thus far, with projects delayed by workforce shortages.
Water requirements for cooling will strain watersheds whilst the specialised workforce barely exists. MIT's Vijay Gadepally notes data centers already account for 1-2 percent of global energy demand—similar to aviation—with projections reaching 21 percent by 2030.
Silicon Valley's inconvenient history
The plan's core premise contradicts American technological development history. Margaret O'Mara's research documents how DARPA funding, 1965 immigration reform, and 1970s capital regulations created Silicon Valley's conditions for success. ARPANET, Defence Department funded, became the internet foundation. The National Science Foundation's NSFNET provided commercial internet's "backbone."
"Silicon Valley was built on the government's largesse," notes Foreign Policy's Julian Zelizer. Computer mice to semiconductors emerged from federal research programs. Current deregulation risks dismantling the institutional foundations enabling American technological leadership.
The workforce contradiction intensifies this problem. Institute for Progress analysis shows 60 percent of leading AI companies have immigrant cofounders, yet immigration restrictions discourage global talent acquisition. MIT Technology Review's Asad Ramzanali warns that "the US is already losing its AI talent edge through anti-immigration actions."
China's efficiency advantage
The plan may be solving yesterday's problem. China's DeepSeek R1 recently matched sophisticated AI capabilities with dramatically lower computational requirements than American companies assumed necessary. This suggests the future belongs to efficiency rather than brute-force computing power, potentially rendering massive infrastructure investments obsolete before completion.
America's strategy of building vast data center complexes may resemble investing in telephone switching infrastructure just before mobile technology rendered centralised systems obsolete.
Industry lobbying reveals aligned financial incentives. Technology sector spending reached $36 million quarterly — $320,000 per congressional session day according to Issue One. Meta spent $13.8 million and hired 86 lobbyists; Nvidia increased spending 388 percent year-over-year. These companies profit from infrastructure buildouts regardless of productivity gains.
Commercial reality versus nationalist rhetoric
Export controls and technology sharing restrictions with allies risk creating a technically sophisticated but economically unviable domestic industry. Semiconductor and aviation history shows global market leadership, not domestic protection, provides sustainable competitive advantages.
Current constraints cannot be overcome through regulatory changes. Northern Virginia maintains data center vacancy rates below 1 percent. Power limitations force utilities to offer capacity in small tranches whilst infrastructure develops. The plan acknowledges these challenges but offers limited solutions beyond streamlined permitting.
Meanwhile, the EU's AI Act prioritises safety over speed, potentially attracting talent concerned about American oversight gaps.
The human cost of technological nationalism
Pew Research reveals the expert-public divide: 56 percent of AI experts expect positive impacts whilst only 17 percent of the public shares this optimism. Customer service representatives, data entry clerks, and administrative assistants face displacement with minimal retraining support, whilst infrastructure subsidies flow to existing technology companies.
Investment concentrates in already-advantaged coastal regions despite national benefit rhetoric. Federal funding criteria considering state AI regulatory climates may further concentrate resources in business-friendly jurisdictions rather than areas with displaced workers.
The foundations of failure
The AI Action Plan embodies a fundamental misunderstanding of how technological leadership develops. By treating innovation as linear progress accelerated through deregulation, the administration ignores the institutional relationships that historically enabled American success.
Competing with China through infrastructure spending misses China's actual advantage: iterating quickly on efficiency improvements whilst American companies pursue computational maximalism. DeepSeek's achievements suggest smart engineering trumps massive investments, potentially leaving America with expensive infrastructure serving obsolete approaches.
The plan's "bias-free" AI requirements create competitive barriers disguised as neutrality, concentrating market power among firms that modified products to meet political requirements. This represents regulatory capture the plan ostensibly opposes.
The ultimate irony: promising to restore American technological dominance through policies undermining past success foundations. Silicon Valley required decades of patient government investment, immigration openness, and regulatory frameworks encouraging both innovation and responsibility. Current approaches risk dismantling these advantages whilst China and allies build comprehensive capabilities.
Success in the AI era demands sustained, bipartisan commitment to institutional development that enabled previous breakthroughs. Instead, the Action Plan offers political theatre disguised as industrial policy, substituting regulatory rollbacks for the hard work of building competitive advantages through education, research investment, and international cooperation.
Whether America maintains technological leadership may depend less on artificial intelligence capabilities than recognising that innovation requires institutions, not just infrastructure.