Why Public Health Must Lead the Push for “Health-Informed AI”
Artificial intelligence is often celebrated for saving lives, predicting disease outbreaks, and improving healthcare access. But beneath that innovation lies a lesser-known truth: the infrastructure powering AI—massive data centers—can harm the very public health it aims to protect.
By 2028, U.S. data centers are expected to consume 6–12% of the nation’s electricity, costing $20 billion annually in health damages linked to fine particulate pollution (PM₂.₅) and nitrogen oxides. These emissions exacerbate asthma, heart disease, and cancer—especially in communities already burdened by environmental inequities.
Recent headlines have made this local and urgent. In Boxtown, Memphis, residents reported skyrocketing asthma rates and a 79% spike in nitrogen dioxide after an AI supercomputer complex, Elon Musk’s xAI “Colossus,” came online. Boxtown’s experience isn’t unique; it’s a preview of how the AI boom may reproduce classic patterns of environmental racism, concentrating pollution in low-income, Black, and brown neighborhoods.
AI’s “Digital Smog”: Why It Matters for Public Health
Like vehicle exhaust or industrial smoke, “digital smog” comes from two main sources:
- On-site diesel or gas generators are used to ensure uninterrupted data center power.
- Pollutant-intensive electricity grids (still heavily reliant on coal and natural gas).
While companies often highlight “net-zero” carbon goals, these address only climate impacts—not acute respiratory and cardiovascular risks. Unlike carbon dioxide, PM₂.₅ and NOₓ have no safe threshold; even brief exposures can trigger asthma attacks or heart events.
That means public health professionals—not just computer scientists—must play a leadership role in shaping the growth of AI infrastructure.
From Carbon-Aware to Health-Informed AI
Some technology companies, like Google, are experimenting with “carbon-aware” computing, dynamically shifting AI workloads to data centers where renewable energy is available. However, the next evolution is “health-informed AI”: integrating air pollution exposure, population density, and local health data into decisions about where and when to train AI models.
Research shows that such load-shifting strategies could cut AI’s health impact by up to 26% while also reducing costs and emissions. When combined with rigorous siting policies and transparent pollutant reporting, this approach could transform AI from a silent polluter into a genuine partner in population health.
What Public Health Professionals Can Do
1. Include AI in Environmental Health Monitoring
Public health departments already track industrial emissions and transportation pollution. Add AI data centers—often hidden behind nondescript buildings—to your environmental health assessments. Partner with academic or citizen-science groups to install low-cost air quality monitors near data center clusters.
2. Advocate for Health Impact Assessments (HIAs)
Before new facilities are approved, push for mandatory HIAs that evaluate impacts on air quality, respiratory disease, and equity. States like Washington have started requiring this for large energy-intensive projects—public health agencies can replicate and expand these models.
3. Build “Healthy AI Procurement” Standards
Just as hospitals now purchase sustainably sourced food or PPE, public agencies and universities can adopt procurement guidelines that prioritize cloud vendors with transparent pollution data, Tier 4 generators, and renewable-heavy grids. Encourage partners to report not only carbon but also criteria pollutants (PM₂.₅, NOₓ, SO₂).
4. Educate Staff and Partners on Digital Energy Use
Every AI-driven workflow—whether it involves chatbots, predictive analytics, or image processing—runs on computing power. Develop quick staff trainings or infographics showing how to reduce compute intensity (e.g., use smaller models for pilot projects, schedule processing during off-peak hours, or select cloud regions with cleaner grids).
5. Center Equity in AI Governance
Boxtown’s story highlights the importance of integrating environmental justice into digital transformation. When your organization deploys AI, map who bears the energy and pollution costs—and ensure community voices shape mitigation plans. This is the same equity lens public health applies to housing, transportation, and climate.
The Opportunity Ahead
If carbon-aware computing is the first wave of sustainable AI, health-informed AI is the second—and public health can lead it. Just as epidemiologists learned to trace viruses, we now must learn to trace emissions from algorithms.
The tools already exist: EPA’s COBRA model can estimate county-level health costs associated with emissions, and public satellite data (such as NASA’s TEMPO) can reveal hotspots in near real-time. The next step is collaboration—between technologists, environmental health scientists, and community advocates—to ensure AI’s expansion doesn’t widen health gaps.
AI may help predict pandemics, accelerate drug discovery, and personalize care. But unless we manage its invisible byproducts—pollution, inequity, and health burden—it risks doing more harm than good. Health-informed AI provides a roadmap to ensure that technological progress also leads to cleaner air, stronger communities, and a truly healthier future.
