这些协同政策把资源倾斜给 AI，推高已宣布的 AI 资本支出：hyperscalers 正把「数千亿美元」投入 data centers 内成千上万列的 servers、缆线与路由器。计算力预期到 2030 年将「加倍或更多」。随著 data centers 扩张，电力需求同步上升：McKinsey 预估 2030 年前新增上线的 data centers 需要「超过 600 terawatt-hours（TWh）」的电力，约相当于「接近 6,000 万户」家庭用电。供应链与成本压力叠加：transformers、circuit breakers、switchgear 等关键设备在需求超过产能后已出现投入成本上行；而 2025 年的关税进一步推高进口设备成本，其中对钢、铝、铜线的「50%」惩罚性税率特别打击金属占比高的 transformers、电力线路与输电塔；公用事业用的电力储能电池「几乎都来自中国」，因此面临更高税负。
移民政策使 AI 建设更慢且更贵：建筑工种中「25%」来自海外，「每 7 人就有 1 人」是无证移民；在更严的边境、ICE 突袭与加速遣返下，工地的临时与成队劳动力更难取得。承包商调查显示「超过 80%」有职缺，且缺工比近年记忆中的任何时期都更难补，缺工是当前工程延宕的「第一」原因。这种紧缺出现在其他建设走弱之际：新屋开工量下降「接近 10%」至五年低点，商用建设下降「13%」，但选举前的住房可负担性压力可能引发新一波住宅建设，与 data centers 竞逐同一批电工、HVAC 技师与焊工。文末主张把重点同样放在工种供给：增加一般 H-2B、扩大并加速 EB-3（建筑工种）申请，或建立新的建筑专用临时签证；即使 AI 能证明投资价值，成本与工期仍决定谁能受益、何时受益，而 Trump 的关税与移民政策正在压住美国的 models 与公司。
AI is pushing the S&P 500 index and the broader US economy forward, but the narrative around a few dominant firms has blurred the line between hype and reality. The argument is that what could burst the AI bubble is not necessarily circular financing, rising debt, or Chinese competition, but the unanticipated drag of tariffs and a fall in the number of migrants in the US. Trump promises to do “whatever it takes” to lead in AI: opening federal lands for data centers and power plants, fast-tracking permitting and environmental reviews, and taking equity stakes in Intel Corp. and x-Light Inc., while challenging state-level AI rules. The administration also exempts servers, semiconductors, circuit boards, and other electronics (roughly one-third of data center costs) from tariffs, even as imported building materials still face levies.
These coordinated policies tilt resources toward AI and amplify announced AI capex: hyperscalers are pouring hundreds of billions of dollars into data centers filled with thousands of rows of connected servers, cables, and routers. Compute power is expected to double or more by 2030. As data centers proliferate, electricity demand rises with them: McKinsey projects that new data centers coming online between now and 2030 will need more than 600 terawatt-hours (TWh) of electricity, enough to power nearly 60 million homes. Supply and cost pressures compound: key utility equipment such as transformers, circuit breakers, and switchgear already faces rising input prices as orders outstrip production, and 2025 tariffs further raise the cost of imported products, with punitive 50% rates on steel, aluminum, and copper wires disproportionately hitting metal-heavy transformers, power lines, and transmission towers; storage batteries used by utilities come almost entirely from China and therefore face even steeper levies.
Migration policy makes the AI build-out slower and more costly: 25% of building trades workers come from abroad, and one in seven are undocumented; with tighter borders, ICE raids, and stepped-up deportations, jobsite labor becomes harder to source. Contractor surveys indicate that more than 80% have openings and that these gaps are harder to fill than in recent memory, making worker shortages the number one reason for project delays. This scarcity grows even as other construction weakens—housing starts are down nearly 10% to a five-year low and commercial construction is down 13%—yet a likely home-building push would send more residential projects chasing the same shrinking pool of electricians, HVAC technicians, welders, and others. The prescription is to treat craft labor as strategically as engineers: offer more general H-2B visas, expand and speed EB-3 processing for building trades, or create a construction-specific temporary visa; even if AI proves the value of massive investment, cost and time set who gains and when, and tariffs plus immigration policy are constraining US models and companies.