Not Every Data Center Is the Same
A Cal Bay AI℠ Essay
⬇ Download this essay as a PDF
A Word That Ends Conversations
In October 2026, Oakland’s City Council voted 7–0 to pause approvals for new data centers for 45 days. Oakland joined San Francisco, Richmond, Gilroy, and other Bay Area cities, part of a wave of pauses and bans spreading across the country.
The trigger was a single proposal: part of one downtown building, once home to a government supercomputer lab, converted into computing space for AI companies. Residents filled the meetings. Petitions circulated. Some speakers called for keeping data centers out of the city for good.
Right now, “data center” may be the most loaded phrase in local politics. Say it at a council meeting and the conversation is over before it starts. People hear it and picture the worst version. That reaction is understandable, but it is leading communities toward decisions that may cost them more than they realize.
The Worries Are Real
Let’s start with what the people at those meetings get right.
- Power. The largest AI facilities can draw as much electricity as a small city.
- Water. They can use large amounts of water for cooling.
- Air. Many keep diesel generators on site for backup, in neighborhoods that already breathe more than their share of pollution.
- Bills. When utilities build new lines and substations to serve them, ordinary customers can end up sharing the cost.
There is also a deeper worry: who decides? Communities like Oakland have watched industries arrive with big promises, take what they needed, and leave behind empty buildings. When a developer promises thousands of jobs from a building full of computers, skepticism is earned.
These concerns deserve real answers, not dismissal. Anyone who wants communities to embrace new technology has to start by taking them seriously.
What One Word Hides
The problem is that “data center” now covers almost everything. It describes a room of computers in a school basement. It describes a single floor of servers in an empty office building. And it describes a campus the size of an airport, drawing hundreds of megawatts. These are not the same thing, any more than a corner store and an industrial food plant are the same because both handle food.
When every project gets treated like the biggest one, communities lose twice:
- They shut the door on small, local computing that could serve their own colleges, training programs, and small businesses — the same tools wealthier communities already use to teach their kids to build with AI instead of fear it.
- They miss what the label leaves out. Some of the most power-hungry new uses, like robotics and advanced manufacturing, aren’t called data centers at all, so a data center ban never touches them.
Meanwhile, many downtowns are full of empty buildings with deferred maintenance, dragging down the blocks around them. A blanket ban doesn’t fill those buildings. It just takes one option off the table without asking whether that option might have helped.
A Better Question
The useful question isn’t whether something is a data center. It’s this:
How much power and water will it use, what will neighbors hear and breathe, and what does the community get back?
That question points to a simple approach: match the rules to the size of the impact.
| Size | Think of It As | Rough Power Use | How a Community Might Treat It |
|---|---|---|---|
| Neighborhood-scale | A corner store | Under about 3 MW, on the building’s existing power | Clear standards and a faster path, in exchange for a community benefit |
| Mid-scale | A commercial kitchen | About 3 to 30 MW | Full public review, with standards for water, noise, and generators |
| Large-scale | An industrial food plant | Above about 30 MW, or needing a new substation | Kept out of dense neighborhoods; considered only in industrial areas |
The exact numbers should fit each community’s grid. What matters is the principle: be strict where the impact is large, and workable where it is small.
Making It Practical
Verify with Data People Already Have
Once a year, a facility shares its electric bill’s peak demand. One number tells the city whether it is staying within its category, without inspecting anyone’s software or customers. A public dashboard shows everyone the results.
Make Sure the Community Gets Something Real Back
A small facility that wants the faster path could choose from a menu of benefits worth roughly 10 percent of its capacity:
- free computing time for students and small businesses,
- servers dedicated to a local college,
- a ground-floor training space,
- apprenticeships and local hiring, or
- hands-on help for schools and nonprofits adopting AI tools.
Because the operator chooses this in exchange for speed, it is a partnership, not a penalty.
Be Honest About Jobs
A building full of servers employs a few dozen people, not thousands. What these facilities actually produce is computing power, so that is what the community’s share should be. Not a check that can disappear into a general fund, but computing time that only works for learning, research, and building.
Each unit can be issued as a community compute credit and recorded on a public record no one can quietly change, so residents can see where every credit went and companies can see their contribution reached the people it was meant for.
How Value Moves in the New Economy
Value still reaches the community. It just doesn’t arrive as directly as a factory payroll.
Grid, Land, and Water → Shared Computing Power → Local Colleges and Training → Trained Residents → Jobs and Growing Businesses
The facility uses the community’s grid, land, and water. In place of the jobs it doesn’t create, it shares computing power. That power goes to local colleges, training programs, and small businesses, which use it to teach residents and adopt new tools. Trained residents find work, and local businesses grow and hire.
The facility doesn’t create those jobs itself. It supplies the tool that helps the community create its own.
A Toolbox, Not a Mandate
None of this has to be adopted all at once, or at all. It is a set of tools any city, county, or state can take apart and use as it sees fit.
| Community | What It Might Adopt |
|---|---|
| A small town | Only the yearly power check |
| A college town | Only the benefit menu |
| A large city | The whole approach |
The benefits can fit the local economy too:
| Local Economy | Where Donated Computing Might Go |
|---|---|
| A city | Community colleges and job training |
| A farm town | Agricultural co-ops modeling weather, soil, and crop yields |
| A manufacturing region | Trade schools teaching robotics |
Communities can add their own protections as well:
- Protect household bills. Large facilities pay the full cost of the grid upgrades they need, so residents’ electric rates don’t rise to serve them.
- Turn strain into support. Facilities agree to cut power use during heat waves, and their backup batteries help keep a neighborhood resilience center running during outages.
- Put waste heat to work. Some cities pipe server heat into nearby buildings, pools, or greenhouses.
- Plan for the exit. Operators commit to recycling their equipment and restoring the building if they leave, so the community isn’t left holding another empty shell.
- Start small. Try the approach in one district for two years, then review what worked.
The details will differ everywhere, and they should. Each community knows its own grid, its own needs, and its own history best.
Deciding on Our Own Terms
The choice in front of communities isn’t fear or blind acceptance. It isn’t “ban everything” or “let anyone build anything.” There is a middle path:
Say No to What Doesn’t Fit → Say Yes to What Helps → Make Sure the Community Gets Something Lasting
A pause can be a wise thing. It gives a city time to think. But a pause only helps if it ends in a plan. Otherwise the same fight happens again with the next proposal, and the one after that, while the opportunity goes somewhere else.
This technology is going to be part of how people learn, work, and build businesses for a long time. The question is whether communities like Oakland help shape it, or sit it out and let others decide for them.
Complaining is easy. Building a better path is harder — and it’s the only thing that actually moves anything forward.
Related Reading
- A Tiered Framework for Computing Facilities in Oakland — the detailed policy proposal behind this essay
- The 25-Year Energy Wall
- The Physical Hierarchy of Compute
- Additionality or Extinction
Sources
- Local News Matters, “Oakland becomes latest in Bay Area to enact moratorium on new data centers”
- SFist, “San Francisco and Oakland Pass 45-Day Moratoriums on New Data Centers”
- NBC Bay Area, reporting on the Richmond and Gilroy data center moratoriums
- Data Center Dynamics, “Former LBNL supercomputing facility in Oakland, California targeted for 20MW data center”