Durabilité.aiurabilité.ai
Beta

On the business approach to sustainable AI

By Chieh Hsu

8 December 2025

Introduction

AI is scaling fast enough that its energy and emissions footprint is becoming a material business variable beyond just a technical detail. Bain & Company estimates that, in a high-growth scenario, AI and data centers could emit 810 MtCO₂ per year by 2035, about 2% of global emissions and 17% of industrial emissions (Bain & Company, 2025). At the same time, the electricity demand of data centres is projected to roughly double by 2030 in the IEA’s base case (to ~945 TWh), making it among the fastest-growing sources of power demand (IEA, 2025). This article argues that “sustainable AI” is increasingly driven by core business incentives: cost, regulation, customer demand, and, most crucially, the cost of capital.

Where the footprint shows up

Two places matter most for business decisions:
Electricity for training and inference (data centers). The IEA’s “Energy and AI” analysis highlights rapid growth in data-centre electricity consumption this decade (IEA, 2025).

Upstream supply chains (equipment and services). In corporate GHG accounting, purchased goods and services are typically captured in Scope 3, Category 1, with formal methods for estimating these emissions (GHG Protocol, 2011).

Motivation 1: Minimise operating cost, energy becomes a competitive input

Power and cooling are major operating costs for AI infrastructure. A key economic point is that renewables are increasingly the most cost-competitive source of new electricity in many regions (IRENA, 2025).

Strategy: Procure low-cost clean electricity through PPAs (Power Purchase Agreements), on-site generation, or grid-based procurement where renewables are abundant, and prioritise efficiency measures that reduce total MWh per unit of compute delivered.

Motivation 2: Compliance and disclosure, tightening regulation

In the EU, sustainability reporting rules are expanding and standardising. The European Commission describes corporate sustainability reporting requirements as focusing on both risks companies face and impacts companies have (European Commission, n.d.).

Strategy: Treat sustainable AI practices as part of compliance readiness, such as disclosure controls, auditable metrics for energy, carbon footprint, or water, and credible transition plans, rather than as optional CSR (Corporate Social Responsibility).

Motivation 3: Brand and customer demand, sustainability as a product attribute

Sustainability can affect demand, especially for customer-facing AI products and for enterprise buyers with their own reporting obligations. For example, PwC reports consumers say they are willing to pay an average premium for sustainably produced or sourced goods (PwC, 2024).

Strategy: Compete on sustainability as a measurable feature, e.g., model-level energy or carbon reporting, region-aware routing, “low-carbon inference” tiers, while avoiding vague claims that risk greenwashing scrutiny (PwC, 2024).

Motivation 4: Lower climate risk can raise valuation via the cost of capital

A common building block in valuation of a company is the Gordon–Shapiro perpetuity growth model, often used as the terminal value component inside a DCF, more in note 1 (Gordon & Shapiro, 1956). When valuing the whole firm using WACC, the cash flow is free cash flow to the firm (FCFF), and the formula is:

  • EV = FCFF₍t+1₎ / (WACC − g)

Where:

  • EV: enterprise value
  • FCFF₍t+1₎: next period free cash flow to the firm
  • WACC: weighted average cost of capital, the required return demanded by debt + equity providers (HBS Online, 2022).
  • g: long-run growth rate

Example inputs:

  • FCFF₍t+1₎ = $50m
  • WACC = 9%
  • g = 6%

Base valuation:

  • EV = 50 / (0.09 − 0.06)
  • EV = 50 / 0.03
  • EV = $1.67B

Scenario A — cash flow rises 10% (to $55m), same WACC and g:

  • EV = 55 / (0.09 − 0.06)
  • EV = 55 / 0.03
  • EV = $1.83B

Scenario B — cash flow stays $50m, but WACC falls by 0.5 percentage points (9.0% → 8.5%):

  • EV = 50 / (0.085 − 0.06)
  • EV = 50 / 0.025
  • EV = $2.00B

This is why reducing risk, in particular climate-related risk, can matter as much as improving near-term cash flows, sometimes can even determine the direction of the company’s growth.

Why climate risk can move WACC

There is a growing finance literature showing that climate risk and carbon exposure are reflected in asset prices and risk premia, mechanisms that can feed into the cost of capital. Examples include evidence that carbon-related risks are priced in equity returns and that carbon risk shows up in tail-risk pricing in options markets (Bolton & Kacperczyk, 2020/2021; Ilhan, Sautner, & Vilkov, 2020). On the debt side, climate-related regulatory risks can affect corporate bond pricing (Seltzer, Starks, & Zhu, 2025 revised).

Strategy: Build a credible climate-risk story that capital providers can underwrite, such as audited emissions data, decarbonisation levers tied to capex, and operational resilience to physical risk because markets can price these risks.

This motivation is inspired by a talk by Mr. Philippe Langlet during the 2025 RedPeak Climate Day.

Motivation 5: ESG / Scope 3 accounting turns “AI usage” into a buyer constraint

Many companies quantify indirect emissions across their value chain using the GHG Protocol framework, including Scope 3 categories such as purchased goods and services. (GHG Protocol, 2011). As corporate reporting expectations expand in regions such as Europe, enterprise customers may increasingly ask AI vendors for verifiable footprints because vendor choices affect the buyer’s own reported emissions and risk narratives (European Commission, n.d.).

Strategy: Offer enterprise-grade footprint transparency in parameters such as energy or carbon footprint per workload, contractual renewable coverage claims, or auditable methodology aligned with widely used standards.

What is a Discounted Cash Flow (DCF)?

A DCF values a company by:
forecasting future cash flows over an explicit horizon (e.g., 5–10 years)
discounting them back to today using a discount rate reflecting risk (often WACC for enterprise valuation), and adding a terminal value to represent cash flows beyond the explicit forecast period. (Damodaran, n.d.).

The Gordon growth formula above is one common method for that terminal value, assuming cash flows grow at a stable perpetual rate g.

Conclusion: a practical checklist for a business-first approach to environmentally-responsible AI

  • Cost: treat clean power + efficiency as unit-economics drivers, not extras.
  • Compliance: build auditable sustainability data pipelines early.
  • Demand: make sustainability measurable at the product level.
  • Capital: manage climate risk as a cost-of-capital lever.
  • B2B fit: align reporting with common standards so enterprise buyers can use the numbers to make wise decisions.

References