AI
Jul 15, 2026
The Climate Cost of AI: Reconciling Rapid AI Innovation with Environmental Stewardship
By Ellza Malok, Chief R&D Officer, AMPRESTA
At AMPRESTA, our mission is to translate complex climate data into actionable financial intelligence. As we advance our model and platform, we will need to rely heavily on advanced Artificial Intelligence and Large Language Models (LLMs). This creates an apparent paradox: we are using energy-intensive technology to solve the problems caused by energy-intensive industries.
Critics often frame this as an "either/or" choice: either we innovate rapidly with AI, or we protect the environment. This is a false dichotomy. The solution is not to abandon AI, but to fundamentally rethink where and how we compute. By differentiating our computing needs and choosing infrastructure that aligns with our values, we can prove that innovation and environmental stewardship are not just compatible—they are mutually reinforcing.
Differentiating Computing Needs
Before selecting a data center, we must distinguish between the two types of computational workloads AMPRESTA performs. Treating them identically leads to inefficiency and unnecessary environmental impact.
Training & Heavy Modeling: These are high-intensity, batch-processing tasks used to refine our climate risk models. They require massive GPU clusters but are intermittent in nature.
Inference & User Workflows: These are the real-time, lower-latency tasks that generate a user’s financial risk report. They require consistent, reliable availability.
Purpose-Driven AI Functions: We strictly avoid implementing "AI for the sake of AI." Every algorithmic function must serve a specific product goal:
Criteria: Does this function provide a quantifiably better value than a non-AI alternative? Does it significantly enhance the customer experience or unlock a decision-making capability that was previously impossible?
Partner Specialization (No One-Size-Fits-All): We recognize that no single provider excels at every task. Our infrastructure strategy is modular.
By segmenting these workloads and aligning them with purpose-built partners, we avoid running "always-on" servers on dirty grids and ensure that our AI deployment is both ethically sound and economically superior.
Not All Data Centers Are the Same
The physical reality of data centers varies wildly depending on their location and design philosophy.
The Standard US Build: Specific Cases of Strain
Many standard US facilities are built in regions chasing tax breaks and cheap land, often with devastating local consequences.
Prince William County, Virginia: Known as "Data Center Alley," this region hosts a massive concentration of hyperscale facilities. Recent reports indicate these centers consume billions of gallons of water annually, straining the Potomac River basin and local aquifers. In some cases, data centers have drawn millions of gallons of water per day for evaporative cooling, directly competing with residential and agricultural needs during drought periods.
Arizona & Georgia: In the Southwest, facilities like those operated by QTS and other major providers have faced intense scrutiny. A single facility in Georgia was reported to have drained over 30 million gallons of water unnoticed, highlighting the lack of transparency and the sheer scale of consumption in arid regions. In Arizona, data centers are exacerbating critical drought conditions, locking communities into long-term water scarcity while offering short-term tax revenues that do not offset the environmental degradation.
In these models, server heat is treated as waste, vented directly into the atmosphere. Furthermore, despite marketing claims of "100% renewable energy" via offsets, many still rely on natural gas "peaker" plants during periods of high demand.
The Ethical Model (Swiss/European)
In contrast, providers like Infomaniak (Switzerland), Hetzner (Germany), Green Cloud (France), and Leafcloud (Netherlands) operate on a circular economy model. Located in regions with abundant hydroelectric power or strict efficiency regulations, these facilities do not just minimize harm; they create value.
Infomaniak’s D4 (Geneva): Captures 100% of its waste heat to warm residential neighborhoods.
Green Cloud (France): Integrates directly with district heating networks, turning server output into community warmth.
Leafcloud (Netherlands): Designed from the ground up to reuse heat for local buildings, achieving a near-zero waste profile.
Notice that these options are completely redesigning what is net-zero from the foundation of their products and not just offsetting those costs through credits or other carbon accounting instruments.
The True Cost: A Four-Dimensional Analysis
To make an informed decision, we must look beyond the invoice price. The true cost of computation includes externalities often ignored in traditional accounting.
Header 1 | Header 2 | Header 3 |
|---|---|---|
Cell 1-1 | Cell 1-2 | Cell 1-3 |
Cell 2-1 | Cell 2-2 | Cell 2-3 |
Cost Dimension | Standard US Model | Global Standard Model |
|---|---|---|
To the Data Center (OpEx) | Lower upfront CapEx, but volatile energy prices; exposure to future carbon taxes and water restrictions. | Higher initial infrastructure investment (heat exchangers), but stable, localized energy costs; immune to carbon taxes. |
To the User (AMPRESTA) | Cheaper per compute-hour initially. Hidden Risk: Reputational damage, future regulatory penalties, misalignment with brand values. | Slightly higher direct cost; though in the last 6-12 months has leveled off and more competitive. Value Add: Aligns with brand promise; mitigates regulatory risk; marketable |
To the Local Population | Negative: Water scarcity, strain on local grids, heat pollution, noise. Communities bear the environmental burden. | Positive: Free/cheap residential heating, grid stability, no water depletion. The community benefits from the data center's presence. |
To the Environment | High: Significant carbon emissions, massive water evaporation, and thermal pollution. | Net Neutral/Negative: Renewable energy + waste heat reuse = computation that actively reduces fossil fuel heating demand elsewhere. |
Comparative Analysis: The Real Cost of Computation
The following chart compares specific infrastructure providers to illustrate the tangible differences in their operational models.
Feature | Standard US Provider A | Standard US Provider B | Global Standard Provider A | Global Standard Provider B |
Primary Energy Source | Mixed Grid (Gas/Coal + Wind offsets) | Mixed Grid (High Solar, but Gas peakers) | 100% Local Hydro & Solar (Real-time) | High Renewable Mix (Wind/Hydro) |
Cooling Method | Evaporative (High Water Loss) | Evaporative (Critical in Drought) | Air-side economization (Low Water) | Advanced Air/Water Hybrid |
Heat Management | Vented to Atmosphere (Waste) | Vented to Atmosphere (Waste) | 100% Recovered (Heats homes) | Highly Efficient (Site dependent) |
Water Consumption | High (Liters per kWh) | Critical (Competes with agriculture) | Negligible (Closed loop) | Low (Strict regulatory limits) |
Community Impact | Strain on local grid & water | Exacerbates drought conditions | Net Positive (Free heat) | Positive (Grid stability) |
Hardware Lifespan | 5–7 Years (Rapid Obsolescence) | 5–7 Years (Rapid Obsolescence) | 10+ Years (Modular Upgrades) | 10+ Years (Modular Upgrades) |
Other Notable Ethical Providers:
Scaleway (France): "Smart DC" initiatives with heat recovery projects in Paris.
OvHcloud (France): Strong commitment to water cooling recovery and renewable energy mix.
Luminus (Belgium): Focus on integration with local energy grids.
The Urgency: Why This Matters Now
The window to act is closing. Recent investigations by environmental activist Erin Brockovich have exposed the lack of transparency surrounding data center construction, highlighting how communities are often kept in the dark about the massive water and energy demands until it is too late. Her work underscores the need for corporate accountability in infrastructure selection.
Furthermore, the economic argument for "cheap" US hosting is eroding. Industry analysis suggests that the effective lifespan of a standard data center facility is shrinking to merely 5 to 10 years due to rapid hardware obsolescence and the inability to upgrade cooling systems for next-generation AI chips. This short lifecycle creates a "build-demolish-rebuild" cycle that generates immense electronic waste and carbon emissions.
Regulatory bodies are beginning to respond. As noted in a recent New York Times report, lawmakers in New York and other states are moving toward moratoriums on new data center construction due to their strain on power grids and water resources. The era of unchecked, environmentally extractive computing is rapidly closing.
Strategic Reconciliation: The AMPRESTA Path Forward
AMPRESTA is not waiting for regulation to force our hand. We are proactively adopting a Green-First Infrastructure Policy:
Primary Providers: Our core infrastructure partners follow the global standard and align to our environmental alignment. These providers align with our requirement for renewable energy, water stewardship, and long-term hardware sustainability.
Workload Segmentation: We schedule heavy training during peak renewable generation and host inference on low-carbon baseloads.
Transparency: As our capabilities grow, we will publish our own infrastructure metrics, applying our own risk intelligence to our operations. We believe our clients have a right to know the carbon and water footprint of their intelligence.
Computing with Conscience
I want to be unequivocally clear about our position: Our main technology providers follow these global standards for high-level computing. We have made this choice deliberately, and we will continue to keep our promise to operate sustainably.
There is no "either/or" situation here, as we have perhaps been led to believe. Innovation and rapid development are not counter-intuitive to environmental stewardship; in fact, true innovation requires stewardship. Building on infrastructure that depletes water and burns fossil fuels is not "fast" — it is a liability waiting to happen.
Our aim at AMPRESTA is to present the financial outcomes when Profit, People, and Planet align. By choosing infrastructure that aligns to our core purpose, we prove the technology solving climate change doesn’t have to come at the planet’s expense. We are building a future where intelligence drives resilience, fueled by renewable energy and community value, not resource extraction.
References & Further Reading:
Erin Brockovich Research: The Brockovich Report, "If Data Centers Are So Great, Why Are They Being Built in Secret?" (2026); TechCrunch, "Erin Brockovich takes aim at data center secrecy" (May 2026).
Financial & Environmental Risk: Sustainalytics, "Data Centers and Water Scarcity: Assessing Corporate Risk" (2026); Ceres, "Drained by Data: The Cumulative Impact of Data Centers on Regional Water Stress."
Regulatory Context: The New York Times, "The Movement to Stop Data Centers," June 9, 2026; Politico, "A data center drained 30M gallons of water unnoticed" (May 2026).
Infrastructure Lifespan: JLL Research, "Why data centers could hit obsolescence sooner than you think" (2025); CIRKLA, "Taking a look at the lifespan of Data Centre components."
Provider Sustainability:
Infomaniak Ecology & Heat Recovery
