Sustainable AI Group Wants Enterprises to Pressure Big Tech on AI’s Energy Use

• energy consumption of AI models
• emissions tied to data‑center electricity
• water use and cooling requirements
• infrastructure choices such as power sources

Another challenge is that different AI systems can have dramatically different environmental footprints. A smaller fine‑tuned model might perform a task using far less compute than a frontier‑scale model, depending on the application and deployment environment .

Despite these differences, buyers often default to the largest or most visible models without visibility into efficiency or environmental trade‑offs.

The Founders’ Key Argument: Enterprise Buyers Have Leverage

Most companies using AI are not building their own models or data centers. They rely on cloud providers and AI vendors.

Yet SAIG’s founders argue that enterprise customers still shape the entire infrastructure ecosystem.

Enterprise spending ultimately funds AI infrastructure expansion. As a result, procurement decisions—from RFP requirements to vendor contracts—send demand signals through the AI value chain .

In other words, while the power grid and data centers are the visible “supply side,” customer demand determines what providers build.

This dynamic gives large organizations leverage to influence:

• model efficiency expectations
• reporting standards for energy use
• infrastructure transparency
• where and how AI workloads are hosted

How Procurement and Contracts Could Shape AI Infrastructure

SAIG argues that enterprises can begin influencing AI sustainability through everyday purchasing processes.

Key mechanisms include:

1. RFP Requirements

Companies can include environmental criteria when evaluating AI vendors, asking for data on model efficiency, energy consumption, and infrastructure practices.

Procurement frameworks increasingly emphasize questions around model optimization practices and environmental reporting, allowing buyers to distinguish vendors with stronger sustainability commitments .

2. Vendor Selection

Organizations can choose AI providers that demonstrate lower energy intensity or better transparency around infrastructure.

If enough enterprise buyers prioritize these factors, vendors may begin competing not just on performance and price, but also on efficiency and sustainability.

3. Contract Clauses

Master service agreements and enterprise contracts can require disclosure about environmental impacts.

For example, contracts might require reporting about:

• the energy sources powering data centers
• whether facilities rely on off‑grid fossil‑fuel generation
• efficiency metrics for AI workloads

These requirements create direct incentives for providers to improve transparency and infrastructure choices .

What Services Sustainable AI Group Will Provide

SAIG positions itself as both a research organization and a strategic advisor helping companies operationalize sustainable AI practices.

Open Research

The firm plans to publish studies analyzing AI’s environmental impacts, including energy use from training and inference workloads as well as broader infrastructure effects .

Enterprise Advisory

SAIG will help organizations integrate sustainability into AI strategies through:

• procurement guidance
• internal workshops and strategy sessions
• measurement frameworks
• decision‑support tools
• stakeholder alignment across technical and sustainability teams

The firm also plans to help companies map where AI is being used internally and establish baseline environmental metrics for those workloads .

Measurement and Benchmarking Tools

A major focus is making AI efficiency measurable and comparable across systems.

One example is the AI Energy Score, a benchmarking initiative designed to rate the energy consumption of AI models across common tasks .

The framework provides standardized energy ratings and public comparisons across models, helping developers and buyers identify more efficient options . Earlier implementations of the system have already evaluated hundreds of widely used models across multiple AI tasks, creating a reference point for energy efficiency in AI systems .

Fit‑for‑Purpose Model Selection

Another planned capability is tooling that helps organizations route tasks to the most appropriate models—rather than defaulting to large frontier models for every use case.

Choosing smaller or specialized models when possible can significantly reduce compute demand while still meeting performance requirements .

Why the Timing Matters

The founders argue that the current moment is critical because AI infrastructure investment is accelerating rapidly.

SAIG estimates that spending on AI infrastructure could reach $5 trillion by the end of the decade, creating a narrow window for shaping how that infrastructure is built .

Once data centers, power contracts, and hardware supply chains are locked in, changing the environmental trajectory of AI becomes far more difficult.

Their strategy focuses on influencing the system before that lock‑in occurs—by encouraging enterprises to treat sustainability as a core purchasing criterion rather than an afterthought.

The Bigger Idea: Sustainability as a Market Signal

The broader vision behind Sustainable AI Group is simple but ambitious: turn enterprise demand into a market signal that rewards efficient AI systems.

If buyers begin requiring transparency and energy efficiency, vendors may compete on those dimensions just as aggressively as they compete on speed, capability, and price.

Whether that shift happens will depend on whether enterprises choose to exercise the leverage they already have in the AI supply chain.