While communities across Sub-Saharan Africa face critical water stress — a reality I lived growing up in Zimbabwe without reliable clean water — the infrastructure powering artificial intelligence is now competing for that same scarce resource in the United States. This project maps exactly where that collision is happening and how severe it will become by 2030.
This is Project 2 in a connected research portfolio. Project 1 modeled water stress across Sub-Saharan Africa. Project 2 shows the same crisis emerging inside the world's most technologically advanced nation — driven by AI data center expansion.
- A single large AI data center can consume up to 5 million gallons of water per day
- US data centers consumed 411.1 billion gallons of water annually
- By 2030 that figure is projected to reach 979.2 billion gallons — a 138.2% increase in 6 years
- California and Texas — the two states with the highest data center water consumption — are also among the most water stressed states in the nation
This model combines two datasets — US state water stress indices and data center water consumption figures — into a single Collision Score that identifies where water scarcity and data center expansion are on a direct collision course.
The core research question: Are the regions building the most AI data centers the same regions already facing critical water stress? The answer, as this data shows, is yes.
- California — Water Stress Index 0.88, 49.2 billion gallons annual DC water consumption. Collision Score: 0.800 — Critical Collision
- Texas — Water Stress Index 0.72, 49.0 billion gallons annual DC water consumption. Collision Score: 0.718 — Critical Collision
- Virginia — Despite lower water stress, hosts 68.4 billion gallons of DC water consumption with 2030 projections approaching 192 billion gallons
- Arizona and Nevada — Both face Critical water stress while simultaneously expanding data center infrastructure rapidly
- 138.2% projected growth in data center water consumption by 2030 nationally
- Collision Score by State (Top 20) — bar chart ranking states by combined water stress and data center water burden
- Water Stress vs Data Center Consumption Scatter Plot — bubble chart where bubble size represents number of data centers
- Current vs 2030 Projected Consumption (Top 10) — side by side comparison showing the scale of growth coming
The Collision Score is a composite index combining:
- Water Stress Index (50% weight) — ratio of water demand to available supply per state
- Normalized Data Center Water Consumption (50% weight) — annual data center water consumption normalized against the maximum observed value
Risk Classification:
- Critical Collision: Score >= 0.70
- High Collision: Score >= 0.50
- Moderate Collision: Score >= 0.30
- Low Collision: Score < 0.30
| State | Water Stress | DC Water (BG/yr) | Collision Score | Risk Level |
|---|---|---|---|---|
| California | 0.88 | 49.2 | 0.800 | Critical Collision |
| Texas | 0.72 | 49.0 | 0.718 | Critical Collision |
| Virginia | 0.38 | 68.4 | 0.690 | High Collision |
| Arizona | 0.95 | 28.6 | 0.684 | High Collision |
| Nevada | 0.91 | 18.4 | 0.590 | High Collision |
| Colorado | 0.79 | 9.8 | 0.467 | Moderate Collision |
| Utah | 0.83 | 6.2 | 0.460 | Moderate Collision |
| New Mexico | 0.87 | 2.1 | 0.450 | Moderate Collision |
| Idaho | 0.71 | 3.2 | 0.378 | Moderate Collision |
| Georgia | 0.45 | 18.6 | 0.361 | Moderate Collision |
- USGS Water Resources — State water withdrawal data
- World Resources Institute Aqueduct — Water stress indices
- Datacenter Map — Facility count by state
- US Energy Information Administration — Energy consumption data
- Published research on data center Water Usage Effectiveness standards
This project applies Professor Braden Allenby's framework of Earth systems engineering — the idea that technological development must be understood at planetary scale, accounting for long-term ecological consequences. AI data center expansion is precisely the kind of technological system that appears locally beneficial but creates systemic environmental stress at scale.
Professor Margaret Garcia's research on water infrastructure resilience under climate stress provides the analytical lens for understanding why the collision identified here is not just a resource competition problem — it is an infrastructure resilience problem.
This project is the second in a connected research portfolio:
- Project 1: Water Stress Risk Index — Sub-Saharan Africa modeled water stress across 15 African countries under current and future climate scenarios
- Project 2: This project maps the same water stress collision inside the United States, driven by AI infrastructure expansion
Together they argue that water stress is not a developing world problem. It is a global systems problem — and artificial intelligence is accelerating it.
- Python 3.14
- NumPy — numerical computation
- Pandas — data manipulation
- Matplotlib — visualization
- Seaborn — statistical graphics
Marcus Mashanda Mechanical Engineering — Trinity College Graduate Studies in Business Analytics — Bentley University Co-founder, Zi-Farm (2020) — Food security initiative, Zimbabwe
"Water insecurity is not an abstraction. It is a daily reality for millions — and the infrastructure we build either alleviates or accelerates it."