AI data centers are surging, but so are emissions and health costs. Here’s how smart permitting and cleaner backup power can turn data centers into true green tech.
Most companies racing to build AI data centers are ignoring the obvious constraint: the grid — and the communities — they depend on.
California is a useful warning light. Between 2019 and 2023, data center electricity use there jumped 95%, and it could grow as much as 356% by 2028. That’s the equivalent power demand of 2.4 million homes. At the same time, health costs linked to data center emissions in the state tripled from $45 billion to $155 billion in just four years. This isn’t a theoretical climate problem; it’s a public health and infrastructure problem that’s already landed.
Here’s the thing about green technology: it isn’t just about adding solar panels and batteries. It’s about aligning policy, permitting, and design so digital growth doesn’t crush local grids, water supplies, and air quality. The new UC Riverside and Next 10 report on California’s AI data centers makes that painfully clear — and also points to a roadmap cities and operators can actually use.
This article breaks down what that roadmap looks like and how smarter permitting, cleaner backup power, and better data can make AI infrastructure part of the green technology transition, not a drag on it.
The Real Impact of AI Data Centers: Power, Water, Health
AI data centers are already one of the fastest-growing sources of electricity demand in advanced economies, and California’s numbers show how quickly that demand can run ahead of planning.
By 2028, California data centers are projected to:
- Use 25.3 terawatt-hours of electricity per year (roughly 2.4 million households)
- Increase their carbon emissions up to 5.56 million tons (a 348% jump from 2019)
- Drive cooling and generation water demand up to 116 billion liters annually
This matters because AI workloads aren’t like traditional IT. Training large models and serving them at scale pushes facilities to the edge of their power and cooling envelopes. If the grid can’t keep up with clean capacity, operators fall back on diesel generators and high-carbon peaker power.
Why diesel-backed data centers are a health problem
Around 90% of California data centers rely on diesel backup generators. Even though they’re “backup,” they don’t sit idle:
- Operators must run them regularly to meet testing requirements
- They’re often used during grid peak periods or local constraints
- They emit fine particulate matter (PM2.5) and nitrogen oxides (NOx) that directly harm nearby communities
The report estimates that by 2028, even under a conservative growth scenario, data center-related emissions could drive about 3,900 asthma symptom cases and 490 lost workdays per year in California alone.
If you’re building or hosting AI services and you’re not looking at where your compute physically lives — and what backs it up — you’re missing the full picture of your environmental impact.
Smart Permitting: The Fast Lane for Truly Green Data Centers
The most powerful lever cities and states have right now is permitting. Not to block data centers, but to prioritize the right ones in the right places.
Smart permitting means tying project timelines and approvals to measurable sustainability criteria, not vague pledges. When you do that, you create a clear signal to developers: if you build greener and site smarter, you move faster.
What smart permitting can look like in practice
A smart permitting framework for AI data centers typically rewards:
-
Location in resource-advantaged zones
Areas with:- Abundant clean power (existing or planned wind, solar, geothermal, hydropower)
- Stable water supplies or non-potable water sources suitable for cooling
- Grid capacity that can absorb large loads without expensive upgrades
-
High performance on energy efficiency
For example, committing to:- Low PUE (Power Usage Effectiveness) targets and independent verification
- Advanced cooling (liquid cooling, rear-door heat exchangers, hot/cold aisle containment)
- AI-based controls that optimize temperature setpoints, airflow, and workload placement
-
Aggressive carbon and water management plans
Including:- Long-term contracts for renewable energy matched hourly, not just annually
- On-site storage to reduce reliance on peaker plants and backup diesel
- Cooling designs that minimize freshwater withdrawals
-
Transparent reporting and community engagement
Requiring:- Standardized reporting on electricity use, emissions, and water consumption
- Health impact assessments
- Clear channels for nearby communities to receive information and raise concerns
When permitting bodies make these criteria explicit and link them to faster approvals and predictable timelines, sustainable projects become commercially attractive — not just morally appealing.
From a green technology standpoint, this is exactly where AI can help cities: using predictive models to map grid constraints, forecast cumulative load, and identify the best zones for new data centers before problems show up on electricity bills.
Phasing Out Diesel Generators: Practical Alternatives
If your data center relies on diesel backup, you’re carrying both a climate liability and a business risk. As more states follow California’s lead on air quality, diesel run hours will get capped, permits will tighten, and local opposition will grow.
The report is blunt: it’s time to transition away from diesel and toward cleaner backup systems.
What can realistically replace diesel today?
Several technologies are ready for prime time, especially in combination:
- Battery Energy Storage Systems (BESS)
- Provide instantaneous backup for short and medium durations
- Support grid during peaks, reducing strain and emissions
- Pair well with on-site solar or wind
- Ideal for reducing generator run hours to near-zero
-
Fuel cells (especially with green hydrogen or biogas)
- Deliver longer-duration, lower-emission power
- Lower NOx and particulate emissions than diesel
- Can be integrated into microgrids with renewables and storage
-
Advanced alternative fuels
- Renewable diesel or biodiesel can be a transition step, especially where full battery or fuel-cell deployments aren’t yet viable
- They’re not a perfect solution, but they reduce lifecycle emissions and some pollutants compared to conventional diesel
-
Grid-interactive data center design
- Facilities can shed non-critical workloads during grid stress
- AI-driven workload orchestration can move compute across regions where power is cleaner and cheaper in real time
- This reduces the need to fire up any backup at all
The most resilient and climate-aligned approach is a hybrid architecture: batteries for instantaneous support, fuel cells or renewables-powered microgrids for longer events, and software that reshapes demand.
If you’re an operator or a corporate buyer of AI infrastructure, you should be asking vendors specific questions:
- How many hours of battery storage back each MW of IT load?
- What’s the plan to phase down diesel generator run hours by year?
- How is backup power integrated with renewables or low-carbon fuels?
Why Standardized Reporting Is the Backbone of Green AI Infrastructure
You can’t build smart policy or credible climate claims on fragmented data. Right now, one of the biggest barriers to effective regulation is that data center reporting is inconsistent across operators and jurisdictions.
The UC Riverside / Next 10 report calls for standardized metrics covering:
- Electricity consumption (facility-level and IT load)
- Carbon emissions (scope 2 from grid usage, plus backup power fuel use)
- Water consumption by source and use (cooling vs other)
- Public health impacts, especially in communities near multiple facilities
What standardized reporting enables
For policymakers and utilities:
- Clear visibility into where and when data centers stress the grid
- Better long-term capacity planning and clean energy procurement
- Ability to target incentives or constraints to specific zones
For cities and communities:
- Evidence to understand and negotiate local impacts
- A way to compare facilities and push for stronger conditions on new builds
For operators and cloud buyers:
- Comparable benchmarks across facilities and regions
- Proof points for sustainability claims investors and customers will actually trust
From a green technology perspective, this is where AI again is useful: ingesting standardized data, modeling health and climate impacts, and helping regulators design smarter, more targeted permitting rules instead of blunt bans or moratoriums.
Community Science and Cross-Sector Collaboration: The Missing Ingredient
California is showing that even in a state with strong climate policy, local impacts can still be severe if communities aren’t part of the planning process.
The report argues for “community science” — structured collaboration where residents, researchers, and regulators share data, tools, and decision-making.
Here’s what that can look like:
- Air quality monitoring near data center clusters, with data shared publicly
- Workshops where operators explain load growth, backup power plans, and mitigation strategies in plain language
- Joint risk mapping, combining community observations with academic modeling
- Co-designed mitigation measures, like vegetative buffers, noise barriers, or localized grid upgrades
When you combine this with transparent reporting and smart permitting, you get a loop where communities aren’t just reacting to projects — they’re shaping what “acceptable” AI infrastructure looks like.
From a business standpoint, that collaboration reduces project risk. Projects that community groups see as aligned with clean energy, clean air, and green jobs get built faster and face fewer legal or political surprises.
How This Fits Into the Broader Green Technology Transition
The broader theme of this Green Technology series is simple: AI can either strain climate goals or accelerate them. Data centers sit right on that fault line.
California’s experience makes two things clear:
- Unmanaged AI data center growth is incompatible with climate and health goals. Left alone, loads and emissions grow faster than grids and communities can adapt.
- Smart permitting and clean backup technologies can flip the script. By rewarding efficient, well-sited, transparently run facilities, states can support digital growth while driving investment into cleaner grids and more resilient infrastructure.
If you’re a policymaker, utility planner, or sustainability leader in a tech-heavy organization, now’s the time to move from broad sustainability statements to hard requirements:
- Tie fast-track permitting to location quality, efficiency, and clean backup power
- Require standardized reporting on energy, emissions, and water
- Support pilots that replace diesel with batteries, fuel cells, and microgrids
- Fund community science and engagement around major digital infrastructure hubs
The reality? Aligning AI infrastructure with green technology goals is absolutely doable — but only if we treat permitting, backup power, and transparency as core design choices, not afterthoughts.
The next wave of AI capacity is being planned right now. Cities and companies that get these fundamentals right will be the ones that can scale AI confidently, without sacrificing the health of their people or the stability of their grids.