What makes an industrial complex is frontier technology, with feedback loops powerful enough to steer both markets and public policy.
Economists describe five telltale traits: mutual dependency between state and suppliers; capital costs so high that only a few firms can play; standards written by insiders; external costs shifted to the public; and a strategic framing (“national security,” “global competitiveness”) that evades scrutiny.
By that definition, we are witnessing the construction of an AI industrial complex, and its socioeconomic wake is already visible.
The early externalities: Power and water
The International Energy Agency projects that, by 2030, global data‑center electricity demand will double to roughly 945 TWh — a trillion watts used per hour. That is about the annual consumption of Japan. Much of that growth is traceable to AI inference and training workloads. The strain is leaking into household bills.
- In Virginia, Dominion Energy has requested a $10.92/month fuel‑rate increase for the “typical” home customer, citing higher gas prices and PJM capacity charges driven in part by data‑center growth. The request is still pending.
- Prices in PJM’s latest capacity auction cleared 22 % higher than last year; the East Coast grid operator estimates that will add between 1.5% and 5 % to retail bills beginning next summer.
Electricity is only half the story. Google’s environmental report shows its hyperscale facilities used nearly 6 billion gallons of fresh water in 2024, up 8 % year‑on‑year, a jump Google attributes largely to AI. The company says it “replenished” 64 % of that total, but the replenishment often happens far from the aquifers it drains. Out West, drought‑stressed cities respond by digging deeper municipal wells or raising water rates — expenses that never appear on a quarterly earnings call.
Labor and looming AI divide: Whither the middle class?
Generative AI in its first iteration targets the heart of the middle‑class office and clerical economy. Brookings finds more than 30 % of U.S. workers could see at least half of their tasks disrupted, with mid‑skill professional roles most exposed. Public anxiety is tracking the data: A 2024 Governance AI survey shows about half of Americans expect AI to widen income inequality, and two‑thirds want federal action to cushion job losses. The paradox: The same families paying higher utility bills may also face stagnant wages or displacement in the very sectors AI automates.
Access to front‑of‑the‑line models already carries a price tag — ChatGPT Plus costs $20 a month, and enterprise “Pro” tiers are tenfold higher. When homework help and job‑application prep migrate behind priority tokens, the complex threatens to hard‑code a two‑tier digital citizenship: seamless for those who can pay, throttled for those who cannot.
The new and irreparable digital divide will be between those who know how to work with AI and those who do not. Unlike the paternalistic and naive exhortations to learn to code, this is not about coding or even about software at all — this divide is about critical thinking, first-principles reasoning, and the ability to harness an emergent capability that can manifest both cloying sycophancy and Machiavellian deception.
The predictable response: UBI, with Big AI as the adjudicator
Inevitably, the drums will begin to beat across the nation for so-called universal basic income and its ilk; the state will have to devise economic mechanisms to address unemployment and the increasing replacement of both white and blue-collar jobs; but at what cost?
The social contract will have to be rewritten; AI will increasingly assume the roles of fact-checker, arbiter and judge; and Big AI will loom large in the enforcement and governance of this new contract, as the state is forced to reinvent itself. “Need more training data” is the cry, and intellectual property is collateral damage; we are witnessing the vestiges of the publishing industry attempting to make a quick buck as copyrights are being railroaded. History is written by winners, and it remains to be seen how the winning AI models and their trainers shape what will count as downstream history.
The American spirit owes its striking characteristic to the frontier — Americans have always shaped our environment and our tools, not the other way around. But we are at a critical juncture. Are we going to let AI shape our socioeconomic future? Are we about to voluntarily swap our frontier spirit for a gilded cage of our own making?
The first meaningful step in addressing this is the rolling back of the proposed moratorium on state and local AI laws in H.R.1 (aka the “Big Beautiful Bill”).
Considering the deep socio-economic impact of AI, it’s crucial that regulatory authority stays with local governments. A decentralized approach enables state-level experimentation, tailors oversight and risk management to local needs, and increases the likelihood of harnessing AI’s potential for the common good.
About the author: John deVadoss: is a former general manager at Microsoft Corp. in Redmond. He led architecture strategy for .NET, SOA and the early SaaS/Cloud initiatives at Microsoft.
Loading recommendations…
