Data centers power AI and green tech—but their diesel generators, energy and water use are surging. Here’s how smart permitting can clean that up.
Most people still think of AI as “digital” and therefore clean. The numbers coming out of California say otherwise.
Between 2019 and 2023, electricity use by California data centers jumped 95%. By 2028, it could be up 356%, pulling as much power as 2.4 million homes. Health costs tied to pollution from backup diesel generators have already tripled, from about $45 billion to $155 billion in four years.
This matters because data centers are the physical backbone of green technology and AI. If we get this wrong, the infrastructure that’s supposed to accelerate the clean energy transition ends up quietly undermining it.
In this post, part of our Green Technology series, I’ll break down how smart permitting and a planned phase-out of diesel generators can keep AI growth compatible with climate goals—and what city leaders, utilities, and data center operators can do about it right now.
The Hidden Environmental Cost of Data Centers
The key issue is simple: data centers are growing far faster than grid decarbonization and local planning rules can keep up.
A new analysis of California’s data centers shows three converging pressure points:
- Electricity demand is exploding. Data center power use in the state rose 95% from 2019–2023 and could rise up to 356% by 2028.
- Emissions are rising despite a clean grid. Two-thirds of California’s electricity comes from clean sources, yet data center-related carbon emissions nearly doubled from 1.24 to 2.38 million tons over that same four-year period and could hit 5.56 million tons by 2028.
- Water use is becoming a crisis. Cooling systems and power generation could require up to 116 billion liters of freshwater annually—in a state that already faces chronic drought.
The reality? You can’t call AI-powered infrastructure “green” if it’s driving local air pollution spikes, straining water supplies, and forcing utilities to lean on fossil-heavy peaker plants.
From a green technology standpoint, data centers are both the problem and the solution. They’re essential for:
- Optimizing renewable energy with AI-enabled forecasting
- Managing smart grids and demand response programs
- Running climate models, resilience analytics, and city-scale digital twins
But if we don’t set guardrails now, every incremental AI workload comes with a hidden tab in emissions, water, and health costs.
Why Diesel Backup Generators Are a Public Health Problem
If you only remember one thing from this article, let it be this: diesel backup generators are not just an emergency tool—they’re a steady source of pollution in many communities.
Nearly 90% of California’s data centers rely on diesel backup generators. On paper they’re for rare outages. In practice, they’re run regularly for:
- Compliance with strict testing and maintenance rules
- Supporting facilities during peak grid load or grid stress events
Every hour those generators run, they release fine particulate matter (PM2.5) and nitrogen oxides (NOx). These pollutants are directly linked to asthma, cardiovascular disease, and premature death.
The recent report estimates that from 2019 to 2023, health costs associated with data center emissions tripled, from $45 billion to $155 billion, and could reach $267 billion by 2028. Even under a conservative growth case, emissions tied to data centers could drive around 3,900 asthma symptom cases and 490 lost workdays per year by 2028.
Here’s the thing about this pollution: it’s hyper-local and deeply inequitable. The communities that live next to industrial zones, power corridors, and data center clusters—often lower-income neighborhoods and communities of color—take the hit, while the benefits of AI are distributed globally.
If you care about green technology and environmental justice, phasing out diesel in data center design isn’t a nice-to-have. It’s mandatory.
Smart Permitting: Reward the Cleanest Data Centers First
Smart permitting is one of the most powerful—and underused—tools cities have right now.
At its core, smart permitting means tying the speed and ease of approvals to the environmental performance of the project. For data centers, that can flip the current logic on its head:
- Today: approvals often go first to whoever secures land and capital fastest.
- Tomorrow: approvals go fastest to facilities that are energy- and water-efficient, avoid diesel, and are sited where the grid and water resources can actually support them.
What Smart Permitting Can Look Like in Practice
A smart, green-technology-aware permitting framework for data centers would:
-
Prioritize the right locations
- Near abundant clean electricity, not at the edge of an already stressed fossil-heavy grid.
- In regions with sustainable water supplies or non‑potable water reuse systems.
- Away from already overburdened communities with high cumulative pollution exposure.
-
Bake in performance requirements up front
Fast-track applications that commit to:- Aggressive power usage effectiveness (PUE) targets
- Low water usage effectiveness (WUE) or air/liquid cooling technologies that minimize water use
- On-site renewable energy or long-term clean power procurement
- Non‑diesel or ultra‑minimal diesel backup solutions
-
Tie permits to data transparency
Require operators to report annually on:- Electricity use
- Carbon intensity of their power mix
- Water withdrawals and consumption
- Backup power runtime and fuel type
Most companies get this wrong by treating permitting as a hurdle to clear, not a strategic tool. The smarter play is to design facilities that qualify for the fastest, lowest-friction pathway because they’re measurably better for the grid and the community.
For public agencies, this is where green technology policy becomes tangible. You’re not simply setting climate targets; you’re using the building and zoning code to decide what kind of AI infrastructure gets to exist in your city.
Cleaner Backup Power: Beyond Diesel Generators
The good news: we’re no longer stuck with a binary choice between diesel generators and unacceptable outage risk.
There are several viable, greener backup power options that data center designers and operators can deploy today:
1. Battery Energy Storage Systems (BESS)
Battery storage is the most direct alternative to diesel for short-duration backup and grid support.
- Can provide instant response during grid events
- Pairs naturally with on-site solar or nearby renewables
- Eliminates local air pollution during discharge
Paired with smart controls, batteries can also provide demand response and frequency regulation services, turning a sunk reliability cost into a revenue-generating grid asset.
2. Fuel Cells and Low-Carbon Fuels
Fuel cells running on green hydrogen, biogas, or other low‑carbon fuels can offer longer-duration backup without diesel’s particulate emissions.
- High efficiency, low local pollutants
- Can be deployed in modular form for redundancy
- Potential to integrate with future hydrogen infrastructure
This isn’t a free pass—fuel sourcing and lifecycle emissions still matter—but it’s substantially better for public health than conventional diesel.
3. Grid-Integrated Tiered Backup
A more advanced approach is to treat backup not as a separate system but as a tier of grid support:
- Critical loads backed by batteries and fuel cells
- Non‑critical loads shed or throttled during grid stress
- AI control systems predicting and smoothing demand spikes
This is where artificial intelligence and green technology intersect beautifully. The same AI models that optimize data center workloads can forecast grid conditions, orchestrate battery dispatch, and coordinate with utilities in real time.
The bottom line: cities should be phasing out diesel generators through standards and timelines, and permitting should explicitly reward operators who switch to cleaner backup architectures.
Why Standardized Reporting and Community Science Matter
You can’t manage what you can’t measure. Right now, data center reporting is fragmented at best.
The California research team argues for standardized reporting of:
- Electricity consumption
- Carbon emissions
- Water use
- Public health impacts tied to emissions
I’d go further: this should be the norm in any region courting data center investment, especially where AI infrastructure is marketed as part of a green technology cluster.
What Standard Reporting Unlocks
Once you have consistent, comparable data across facilities, regulators and communities can:
- Identify the highest-impact mitigation levers (e.g., targeting diesel runtime vs. tightening PUE)
- Adjust electricity tariffs and incentives to reward cleaner designs
- Integrate data centers into local climate action plans
- Benchmark operators and make that information publicly visible
That last point matters. When communities can compare two projects side by side—on emissions, water, and health impacts—political support shifts toward cleaner designs very quickly.
The Role of Community Science
The report also calls for community science and cross-sector research. I’m strongly in favor of this.
In practice, that looks like:
- Local air monitoring projects around data center clusters
- Public workshops explaining health impact data in plain language
- Collaborations between universities, health departments, and neighborhood groups
This isn’t just transparency for its own sake. When residents understand the trade‑offs and see that operators are actually reducing emissions and water use, resistance to new projects drops. That’s good for both sustainability and project timelines.
Practical Steps for Cities, Utilities, and Operators
Here’s how different players can act on all this right now.
For City and Regional Planners
- Adopt smart permitting criteria tying approval speed to:
- Siting near clean power and sustainable water
- Diesel‑free or diesel-minimized backup
- Strong energy and water efficiency commitments
- Require standardized reporting for all large data centers
- Integrate health impact analysis into environmental review, not as an afterthought
For Utilities and Grid Operators
- Prioritize clean capacity additions in emerging data center corridors
- Offer tariffs and programs that reward flexible, grid-friendly loads
- Co-design battery and demand response programs with large data center customers
For Data Center Developers and Operators
If you’re serious about participating in the green technology transition—and not just marketing it—your checklist should include:
- Designing for aggressive PUE and WUE targets
- Committing to 100% clean or near‑zero carbon power over contract lifetimes
- Replacing or minimizing diesel backup with batteries, fuel cells, and intelligent load shedding
- Sharing data on energy, water, and emissions with regulators and neighbors
The reality is that investors and enterprise customers are increasingly scrutinizing the full footprint of their AI infrastructure. Clean, transparent, smartly permitted facilities will win the best tenants and the cheapest capital.
Where Green Technology Goes From Here
AI and data centers aren’t going away. In fact, they’ll underpin almost every serious climate and sustainability solution over the next decade—from smarter grids to optimized transit systems.
The question isn’t whether we build more data centers. It’s how we design, permit, and power them.
Smart permitting, standardized reporting, and a rapid shift away from diesel backup generators are the fastest ways to align AI infrastructure with climate and public health goals. California is a warning and a blueprint at the same time: even with a relatively clean grid, unmanaged data center growth can overwhelm local resources and communities.
If you’re working in city planning, utilities, or data center development and want your projects to truly fit in a Green Technology future, this is the moment to tighten your standards—not loosen them.
The next wave of AI capacity will either lock in decades of unnecessary emissions and health damage, or become the backbone of a cleaner, more resilient energy system. The policies you adopt this year decide which path we take.