The Last Factory: When Data Centers Stop Needing Communities

In my inaugural TAIIP essay, I outlined how examining AI from the “outside/in”—looking at power structures, infrastructure, and governance rather than just capabilities—reveals dynamics that implementation guides miss. This essay applies that lens to a specific case: how we’re using outdated industrial development frameworks to govern data center infrastructure, and why that governance failure has profound implications for both communities and accountability.

In 1983, I transferred to Intel’s New Mexico site to help start up the company’s first 6-inch wafer fab. Like many who’d worked in Silicon Valley’s fabs, I understood the pitch Intel was making to communities—construction jobs, local tax revenue, and long-term integration. Engineers would buy homes, suppliers would move nearby, schools would adapt their programs.

I only stayed a year before returning to California, but the inequity of it stayed with me. Over decades—from Santa Clara to Rio Rancho to Kern County—I’ve watched which promises held and which dissolved. Communities kept courting infrastructure, hoping each boom would stick.

Forty years later, the pattern hasn’t just repeated — it’s metastasized. In the 2025 U.S. elections, data centers became a defining issue in states like Virginia, New Jersey, and Georgia, where candidates sparred over who was to blame for soaring electricity
bills.¹

In Virginia, Democratic gubernatorial candidate Abigail Spanberger campaigned on making tech companies “pay their fair share” for grid upgrades, while her Republican opponent, Winsome Earle-Sears, argued that data centers were essential for economic growth and blamed clean energy mandates for rising costs.

The debate wasn’t just about jobs or tax revenue. It was about who bears the cost when a single data center can consume as much power as 400,000 homes, and residents see their monthly bills jump by 11–22% in a single year.

The political theater masked a deeper truth: communities are no longer passive hosts. In Georgia, voters ousted utility commissioners who had approved six rate hikes in two years, electing Democrats to the Public Service Commission for the first time in two decades — an explicit rebuke to policies that let data centers offload costs onto ratepayers.²

Meanwhile, in Arizona and Ohio, local councils rejected massive data center projects after residents organized against the strain on water and power grids. The message was clear: the old playbook—promise jobs, downplay costs—isn’t working. The question now is whether governance can catch up.

---

Part 1: What Manufacturing Promised (and Sometimes Delivered)

The manufacturing-era playbook wasn’t arbitrary — it emerged from how factories actually integrated with communities, however imperfectly.

Fabs employed thousands: operators, trainers, frontline supervisors, engineers and technicians, maintenance and facilities staff. You couldn’t run a fab remotely; you needed local people who could handle emergencies and maintain institutional knowledge. Factories required chemicals, gases, precision equipment. Suppliers clustered nearby because response time mattered. Once built, billion-dollar fabs didn’t easily move; sunk costs gave communities at least nominal leverage.

The tradeoffs were visible, if not exactly accountable. Factories brought jobs and pollution to the same geography. Santa Clara County ended up with 23 Superfund sites—the toxic residue of the semiconductor boom. Cleanup took decades and billions, often after companies built in other states or moved offshore.³

Proximity didn’t ensure accountability, but it made consequences legible. The people benefiting and the people breathing contaminated air at least lived in the same place.

Those flawed systems were still reciprocal at their core — factories needed communities as much as communities needed factories. Data centers dispense with even that pretense.

---

Part 2: How Data Centers Invert Everything

Consider what’s happening right now in California’s Central Valley. The Elk Hills gas plant in Kern County released 88.5 tons of nitrogen oxides in 2023. It’s a 550-megawatt facility that California Resources Corporation is now retrofitting and marketing as infrastructure for AI data centers—not in Kern County, but in Silicon Valley, hundreds of miles away.⁴

One community bears the emissions. Another captures the economic value. The environmental costs and the computational benefits exist in entirely different political jurisdictions.

Data centers sometimes look like factories — large, capital-intensive buildings demanding massive infrastructure. The surface similarities mask a structural inversion. Every assumption that made the manufacturing playbook work collapses here:

Employment vanishes. A billion-dollar facility might employ 50–100 people, mostly specialized imports, not locals. Construction jobs are temporary; automation is extensive. The middle-class employment base that once anchored factory towns never materializes.

Supply chains globalize. Their supply chains are global, not local. The multiplier effect never happens because co-location incentives no longer exist. There’s no ecosystem of suppliers, no network of service providers, no reason for complementary businesses to cluster nearby.

Resources flow one direction. A large AI training facility can use 500 megawatts continuously — enough to power 400,000 homes.⁵ But unlike manufacturing, these resources don’t support local employment or economic integration. Power flows in. Profit flows out. The community is left with stressed grids and depleted aquifers.

Geography as a loophole: What we’re seeing in Elk Hills isn’t an outlier — it’s what happens when we decouple the benefits of industrial growth from the places that bear its costs. This is the predictable outcome of exporting impact while importing advantage.

In the old model, however imperfect, communities and industry shared a feedback loop. If a factory polluted, the people who ran it had to breathe the same air, although economic imbalance allowed many who polluted to avoid having to drink the same water. Accountability didn’t always live next door.

The data-center era breaks that chain entirely. When the beneficiaries are distant and the burden is local, responsibility dissolves. And without shared stakes, there can be no true stewardship—only consumption and cleanup.

---

Part 3: Why Our Governance Can’t Adapt

The policy failures aren’t accidental. They’re structural, emerging from the mismatch between the frameworks we’re using and the infrastructure we’re trying to govern.

We’re still measuring the wrong things. Economic development incentives focus on job creation metrics, local hire requirements, and tax revenue projections — all the measurements that made sense for manufacturing. But data centers optimize entirely different variables: power costs, cooling efficiency, network connectivity. A company can capture every benefit a community offers while delivering almost none of what the metrics measure. The incentive structures don’t just misalign, they reward exactly the behavior we should be discouraging.

Communities have lost leverage. As Natalia Cote-Munoz documents in her technical analysis, the opacity is systemic: “Companies don’t have to disclose how much energy or water they actually use. We’re making 30-year infrastructure bets based on 3-year technology cycles, with no plan for what happens if the AI boom fizzles.”⁶

But this isn’t a bug in the system — it’s what happens when governance frameworks built for sticky, employment-dense manufacturing get applied to footloose, automation-heavy server farms.

Unlike manufacturing, data centers can chase incentives with minimal switching costs. Virginia gave over $100M in sales tax exemptions; other states responded by offering more. The result: communities offering more while getting less, unable to enforce accountability because the company can credibly threaten to build elsewhere. When infrastructure doesn’t need local workforces or supply chains, communities have nothing left to bargain with.

The transparency vacuum. When California lawmakers tried to regulate data centers in 2025, the results illustrated exactly how leverage has collapsed. Four bills entered the legislature; two died outright. The survivors were gutted by amendments in late August — lobbyists for data centers and Big Tech citing trade secrets and competitive concerns.⁷

One bill that would have protected ratepayers from shouldering infrastructure costs was reduced to merely letting regulators assess whether cost-shifting is happening—authority they already had. Another requires water use disclosure, but only to water suppliers, not the public.

This regulatory vacuum creates space for a different kind of policy: bipartisan legislation enabling gas plants with carbon capture to serve as “clean infrastructure” for data centers. The technology may eventually work. The problem is we’re locking in 30-year commitments based on speculative carbon capture while communities bearing current emissions have no say in whether those bets pay off.

The 2025 elections laid bare the consequences of this regulatory vacuum. In Virginia, where data centers contribute $9.1 billion to the state’s GDP, both parties acknowledged the industry’s outsize energy demands but neither offered a coherent plan to ensure communities benefit.

Democrats won by hammering Republicans on affordability, but their solutions—like Spanberger’s call for data centers to “pay their own way”—remain vague on enforcement.

Meanwhile, Republicans in states like West Virginia passed laws creating “microgrid districts” free from local zoning and rate regulations, explicitly prioritizing corporate flexibility over community input.

The result? A patchwork of incentives and pushback, with no national framework to ensure that the communities bearing the costs—higher bills, stressed grids, depleted aquifers—see any meaningful return.

---

Part 4: What Accountability Actually Requires

After four decades of watching these cycles repeat, I’ve learned we can’t ban data centers — but we can govern them. Not through aspirational frameworks, but through enforceable mechanisms.

Transparency as prerequisite. Public disclosure of actual energy, water, and employment data before any incentives are granted. If resource use is proprietary, infrastructure access should be treated as a privilege requiring public justification, not a right. Communities shouldn’t be making 30-year commitments to facilities that won’t reveal their resource demands.

Accountability with teeth. Incentives tied to audited outcomes — measured economic benefit, not projected jobs. Clawbacks that reflect how easily these facilities relocate. If a company closes or automates further within 10 years, the tax breaks get repaid. If water use exceeds disclosed estimates, rates increase accordingly.

Reconnecting costs and benefits. If Kern County bears the emissions that power Silicon Valley’s compute, its residents deserve both voice and share. That means representation in planning decisions and direct benefit from the infrastructure they’re subsidizing, whether through revenue sharing, energy credits, or environmental remediation funds.

None of this is radical. It’s basic infrastructure governance: the people bearing costs get a say in the decisions. The frameworks exist. The precedents exist. What’s missing is the political will to enforce them. And that begins with communities asking the right questions.

---

Closing: The Symmetry Is Cruel

This isn’t nostalgia for factory jobs without accountability. Silicon Valley’s semiconductor boom left its own toxic inheritance — communities living above poisoned groundwater, decades of cleanup still unfinished. Manufacturing’s reciprocity was always asymmetric, its accountability more promise than practice.

Now we’re making the inverse mistake: separating industries from the consequences they create entirely. Silicon Valley once polluted its own aquifers; now it outsources its emissions to Kern County and calls it clean. We learned that proximity doesn’t guarantee accountability. But we’re now discovering that distance guarantees its absence.

Forty years of watching these patterns has taught me that accountability doesn’t emerge from legislation alone—it emerges from communities who understand what they’re being asked to give up, and who demand something real in return.

Cote-Munoz offers three essential questions for communities evaluating data center proposals:⁸

1. Can you see the numbers?

2. Who’s paying for grid upgrades?

3. Are there real community benefits?

Start there. Ask those questions before the contracts are signed, before the incentives are granted, before the thirty-year commitments lock in. Because once the data center is built and the power is flowing, the leverage is gone.

The transition from factories to data centers tests whether we can govern infrastructure we can’t see. Not through nostalgia for factory jobs, but through insisting on the basic terms that made any industrial arrangement accountable: transparency about costs, enforcement of commitments, and consequences that flow both ways.

We know how to do this. The question is whether we’ll insist on it — one community, one proposal, one set of hard questions at a time.

The 2025 elections proved that voters are done waiting for trickle-down benefits. From Georgia’s utility commission races to Virginia’s gubernatorial battle, the message was the same: if data centers are here to stay, the terms must change. The symmetry isn’t just cruel — it’s politically unsustainable. Communities are no longer asking for reciprocity. They’re demanding it.

Dee McCrorey spent twenty years in the trenches of Silicon Valley’s early semiconductor ecosystem, an inside/out perspective shaped by work across National Semiconductor, Monolithic Memories, Intersil, Fairchild, and Intel’s Santa Clara and New Mexico fabs and by watching how infrastructure decisions rippled through surrounding communities. This essay pairs that lived experience with an outside/in analysis, building on Natalia Cote-Munoz’s technical analysis, “The Data Center Reckoning.”

---

Sources

1. “Nine States Face Public Utility Commission Elections Ahead of 2026,” MultiState, October 2025. The article details 11–22% utility bill increases in VA/NJ/GA URL: https://www.multistate.us/insider/2025/10/27/nine-states-face-key-public-utility-commission-elections-ahead-of-2026 ↩

2. “Trump’s focus on cultural issues amid high energy bills zaps Republicans”, Axios, November 2025. URL: https://www.axios.com/2025/11/05/trump-election-energy-utility-bill-increase ↩

3. Santa Clara County’s 23 Superfund sites resulted from decades of semiconductor manufacturing. Cleanup efforts continue, often long after companies relocated production offshore. “Silicon Valley’s Toxic Past Haunts Sunnyvale Neighborhood,” KQED, December 2017. URL: https://www.kqed.org/futureofyou/388730/silicon-valleys-toxic-past-haunts-sunnyvale-neighborhood ↩

4. “Insatiable energy demands of data centers could increase fossil fuel emissions in California,” Capital and Main, October 2024. The article details how California Resources Corporation operates the Elk Hills plant and markets it to AI data center companies. URL: https://capitalandmain.com/the-insatiable-energy-demands-of-data-centers-could-increase-fossil-fuel-emissions-in-california ↩

5. Natalia Cote-Munoz, “The Data Center Reckoning: Power, Water, and Who Really Pays,” Artificial Inquiry (Substack), 2024. URL: “The Data Center Reckoning” ↩

6. Ibid. ↩

7. “California lawmakers wanted to get tough on data centers. Here’s what survived,” CalMatters, September 2025. URL: https://calmatters.org/environment/2025/09/data-centers-california-electricity-rates/ ↩

8. Natalia Cote-Munoz, “The Data Center Reckoning.” URL: “The Data Center Reckoning” ↩

Additional Resources

“Carbon management coming soon to California: Governor Newsom signs legislation paving way for climate pollution-cutting technology,” Governor’s Office Press Release, October 10, 2025. URL: https://www.gov.ca.gov/2025/10/10/carbon-management-coming-soon-to-california-governor-newsom-signs-legislation-paving-way-for-climate-pollution-cutting-technology/ ↩