The Invisible Trillion · What You Can Do That Actually Matters

The Specific Levers The Last Several Chapters Suggest

The middle of this book — the people who live next to the data centers, the demand picture, the architecture that would actually work, the financial bill, the mechanisms by which decisions get made — produces a few concrete levers I want to name plainly. Some you can act on this week; some you could choose as the focus of a career. None of them are the levers you’d have come up with from the standard environmental-impact discussions.

If you live near a proposed or under-construction data center, the window is roughly years zero through three of the Chapter 10 trajectory. Not later. By the time the bills get bad and the for-sale signs start sitting unsold, the people with options have left and taken the capacity to organize with them.

The window is while the neighborhood is still strong enough to fight: find the others who are paying attention, learn the names of the developers and the shell companies, demand the NDAs be disclosed, show up to the early planning-board meetings before most people know what’s coming, and refuse the framing that the project is inevitable. Some communities that have done this have stopped projects. Most that haven’t have ended up hosting them. I won’t pretend the odds are even — they aren’t — but the difference between trying early and not trying is most of what determines the outcome.

If your community is being approached, ask whether the boundary-site trick from Chapter 9 is in play. Developers often place a facility right on a jurisdictional line, so the impacts land on residents in one place while the approving body sits in another. If a proposed site is unusually close to a county or municipal line, ask why. The answer usually tells you something.

If you have any say over how your family’s savings are invested, look at the concentration. I’m not telling you to sell anything — I’m telling you to look. As of 2026, the default target-date funds that hold most American retirement money are heavily weighted toward the AI-infrastructure companies, by way of being weighted toward the S&P 500, which is itself weighted toward them.

The history of speculative infrastructure buildouts is that the resolution tends to involve large write-downs, and those have historically fallen hardest on the people closest to retirement when the reckoning comes. Whether your exposure is appropriate for your timeline is a financial-planning question, not a political one, and reasonable people answer it differently. I’m not your advisor. I’m only saying it’s a question you can answer in fifteen minutes, and most people haven’t asked it.

Evaluate the AI narrative piece by piece, not as a whole. When you read a news story or an industry pronouncement about the buildout, don’t ask is this whole story true. Ask, for each specific claim, what’s the evidence for this one, and what does the answer tell me about which parts of the bigger story are tracking reality.

Ukraine using AI-enabled drones at unprecedented scale: well-supported. The claim that the US must build at the current scale for national security: partly true, substantially exaggerated, functionally serving the narrative. The claim that data centers are good for host communities: false in a way the Chapter 8 evidence makes overwhelming. Piece-by-piece is the move that lets you keep thinking after the news cycle moves on, and it’s a skill that compounds — the more you do it, the harder it gets for any single narrative to capture you.

If you have technical skills, or are building them, build for the architecture that actually works. Chapter 13 made the case that the right shape for general AI — heavy centralized training, distributed edge inference, occasional updates rather than constant round-trips — is already known and already demonstrated in self-driving, and is being slowed mostly by business-model preference.

Edge AI tooling, smaller and more capable open-source models, better distillation, privacy-preserving local inference, federated systems that actually scale: the tools are public and improving, the funding is thinner than the hyperscale side, and the work is genuinely consequential. (There’s also the world-models bet from Chapter 12 — spatial and embodied intelligence rather than language. Earlier-stage, deeper open problems, fewer people, more direct path from competence to contribution. If you’re drawn to robotics and the physical world, that’s where a lot of the interesting work will be.)

Support the journalism, specifically. The local-news collapse from Chapter 10 — private-equity buyouts, ad-revenue dependence on the very industries the papers should investigate — has weakened this exactly as it’s gotten more important. The reporters doing the real work on siting, the NDA pattern, utility rate cases, and supply chains are mostly at understaffed outlets kept alive by subscribers. The marginal subscription — Wisconsin Watch on Beaver Dam, the Minneapolis Star Tribune on Minnesota’s NDA pattern, ProPublica on supply chains, the capital-markets desks that actually cover the money — does more to constrain bad behavior than most consumption choices ever will.

Support the legal mechanism, specifically. The Southern Environmental Law Center brought the xAI Memphis case. Earthjustice, Public Citizen, the regional Riverkeepers, the state public-interest law clinics, the environmental-justice legal groups in affected regions — these are the entities that turn community organizing into actual consequences. They run on grants and donations, chronically out-resourced by the firms they face, and a single successful case can constrain industry behavior across many later decisions. The leverage per dollar is high.

All 19 chapters of The Invisible Trillion · Library · All courses