About the Project
The problem
Data centers power much of the digital world, but they also consume large amounts of electricity and water. These impacts can conflict: a cooling system that reduces energy use may increase water consumption, while lower-carbon electricity may increase operating costs.
Most early-stage planning tools optimize one metric at a time. That makes it difficult for communities, planners, and operators to understand whether a proposed facility fits within local resource constraints.
Our solution
CoolShift is an interactive data center sustainability design lab. It helps users compare cooling systems and electricity sources while considering four connected impacts:
- Electricity consumption
- Site water consumption
- Carbon emissions
- Annual electricity cost
Users can explore three visually distinct facility designs:
- A dry-air reference facility
- A water-cooled alternative
- A future-ready closed-loop liquid design
The facility cutaway changes with each design, showing server racks, air handlers, cooling towers, water systems, manifolds, heat exchangers, and conceptual flow paths.
CoolShift also includes a community guardrail mode. Users can enter annual limits for water, carbon, and electricity cost. The application evaluates 12 illustrative combinations of cooling and electricity systems, removes designs that violate the selected limits, and ranks the remaining options according to the user's priorities.
If no design can support the requested workload, CoolShift estimates the maximum modeled IT load that fits within the constraints. It also identifies which single limit would need to be relaxed and provides hypothetical stress tests, such as a 30% electricity-price increase.
Why we built it
We were inspired by the growing environmental footprint of computing infrastructure and by the difficulty of discussing data center development in terms that both engineers and communities can understand.
A data center is not environmentally sustainable because of one number. Electricity, water, carbon, and cost must be considered together. We wanted to create a tool that makes those trade-offs visible, interactive, and understandable within a few minutes.
How it works
CoolShift uses transparent calculations that run locally in the browser.
For a workload with average IT load (P) in kilowatts and annual operating time (H) in hours:
$$ E_{\mathrm{IT}} = P \times H $$
Facility electricity is estimated using the power usage effectiveness value (PUE):
$$ E_{\mathrm{facility}} = E_{\mathrm{IT}} \times PUE $$
Estimated annual carbon emissions are:
$$ C = \frac{E_{\mathrm{facility}} \times I}{1000} $$
where (I) is the grid emissions intensity in kilograms of (CO_2e) per kilowatt-hour and (C) is measured in metric tonnes.
Estimated site water consumption is:
$$ W = \frac{E_{\mathrm{IT}} \times WUE}{1000} $$
where (WUE) is measured in liters per kilowatt-hour and (W) is measured in cubic meters.
Annual electricity cost is:
$$ K = E_{\mathrm{facility}} \times R $$
where (R) is the electricity price in dollars per kilowatt-hour.
The optimizer evaluates four cooling profiles and three hypothetical electricity supplies, producing 12 candidate designs. It normalizes the candidates across energy, water, carbon, and cost, then applies user-selected priorities to calculate a relative benefit score.
What we learned
We learned that sustainability optimization is usually a trade-off problem rather than a single-objective problem. A design can use less electricity while consuming more water, or reduce emissions while increasing cost.
We also learned that feasibility is as important as optimization. Sometimes the correct answer is not “choose the best design,” but “the requested workload cannot meet these community limits under the modeled assumptions.”
Building the interactive cutaway taught us that visual design can communicate engineering concepts faster than a table of numbers. Showing the cooling equipment and flow paths helps users understand why the metrics change.
Challenges we faced
One challenge was making the three facility scenarios meaningfully different instead of allowing every path to become an arbitrary copy of another. The first two scenarios are fixed reference designs, while the third can only be updated by applying an optimizer recommendation.
Another challenge was keeping the model honest. CoolShift uses illustrative assumptions rather than measured facility data, so the application clearly labels its results as a screening model. It does not claim to be a validated digital twin or a construction-ready engineering design.
We also had to make the interface understandable during a short demonstration. The final application combines an isometric facility visualization, interactive charts, hard community limits, Pareto comparisons, stress tests, and a methodology panel in one dependency-free web app.
Built with
CoolShift was built with:
- JavaScript ES modules
- HTML and CSS
- Programmatically drawn SVG
- Node.js for local development and testing
- The Node test runner for model validation
- OpenAI Codex for AI-assisted development
The core calculations run locally in the browser. No paid APIs, API keys, database, or account are required.
Limitations
All numeric presets are illustrative assumptions. PUE and WUE definitions and the calculation methodology are documented in the application.
CoolShift is an annual screening prototype. It does not model capital costs, weather, thermal capacity, reliability, embodied emissions, off-site water, or time-varying operating conditions. Its recommendations are the best choices among the 12 modeled candidates, not guaranteed globally optimal engineering designs.
Try it
Launch the live CoolShift demo
The strongest demonstration is to load the community challenge, find the feasible design, apply the recommendation to Scenario 03, and then run an electricity-price stress test to see how quickly feasibility can change.
Log in or sign up for Devpost to join the conversation.