The conversation evolves from aspirational to accountable

The most profound shift over the past year has been the forced elevation of the sustainability dialogue directly onto the executive committee’s core risk portfolio. This movement is not voluntary; it is driven by impending regulation and the sobering realisation that environmental failure now carries direct, auditable financial penalties and board-level liability.

Only a year ago, discussions circled around unquantifiable reputational benefits. Today, the lexicon is dominated by acronyms signalling mandatory compliance: CSDDD, CSRD, and the tightening of the SBTi Net-Zero Standard V2. These frameworks compel executives to move past narratives and confront the granular, auditable data attached to every asset, vendor, and cloud usage.

For the CIO, this manifests in two critical areas. First, energy efficiency is decisively reframed as a cost of doing business, crucial for operational expenditure control amid volatile global energy markets. Second, the sudden energy demand of generative AI has triggered a rapid, internal debate on responsible compute architecture.

Leaders are increasingly compelled to justify AI investment not solely on traditional ROI, but via a nascent “return on compute” model that necessarily integrates and accounts for carbon expenditure. This makes the environmental cost of IT an integrated input in the total cost of ownership calculation, rather than a polite footnote.

Despite this high-level engagement, progress remains complicated. The IT function often lacks the authority to enforce change across complex internal silos, and the necessary budget and risk tolerance for truly transformative shifts remain stubbornly limited.

Genuine progress where the green shoots are taking hold

Despite systemic inertia, 2025 delivered solid, tangible progress in certain operational domains, offering a partial blueprint for future net-zero efforts. Our confidence is bolstered by three examples, though it is crucial to understand that wide-scale adoption across the average enterprise remains nascent and often confined to pilot programs:

1. Decoupling cloud growth from carbon: Hyperscale cloud providers have largely won the battle for renewable energy procurement. The next frontier — optimising physical operations — has seen enterprise engagement. We saw accelerated adoption of advanced liquid cooling technologies (still primarily concentrated in hyperscale environments, but critical for future AI scaling). Enterprises optimising workloads for low-carbon regions and utilising serverless architectures successfully decoupled rapid cloud expansion from a proportional rise in emissions. This success belongs predominantly to the hyperscalers, and enterprise optimisation remains an ongoing campaign.

2. Maturing the circular IT model (As-a-Service): The year 2025 saw the Managed Device-as-a-Service (MDaaS) model transition into a critical environmental enabler. By outsourcing the entire device lifecycle, enterprises commit practically to refurbishment and robust reverse logistics. Successful enterprises leverage these contracts to guarantee asset re-entry into the value chain via certified refurbishment, drastically reducing e-waste. The caveats are two-fold: MDaaS adoption is far from universal, and the verification of these circular chains still lacks necessary, robust third-party scrutiny.

3. The nascent rise of green software engineering: The formal emergence of green software engineering (GSE) is perhaps the most encouraging development. For too long, the environmental focus was only on hardware. This year, organisations began measuring code energy consumption — optimising algorithms and refactoring applications to reduce reliance on resource-intensive computing.

An important development this year was the publication of the W3C Web Sustainability Guidelines (WSG) Draft Note. Developed through a global, collaborative effort — in which I was pleased to participate — the guidelines offer a structured and internationally relevant set of best practices for reducing the environmental footprint of web products and services. While the scope focuses specifically on the web rather than the full breadth of enterprise IT, the Draft Note nonetheless represents a significant step forward for the industry.

The persistent gaps undermining net-zero momentum

For all the genuine acceleration, 2025 was equally defined by two persistent, critical gaps that threaten to derail net-zero pathways and demand urgent attention.

1. The Scope 3 emissions chasm: The most pervasive and frustrating gap remains the measurement and meaningful reduction of Scope 3 emissions, particularly from purchased goods and downstream asset end-of-life.

Despite regulatory urgency, the vast majority of enterprises still rely on highly aggregated, industry-average supplier data (spend-based or activity-based), which is neither auditable nor sufficient for mandatory disclosure. The necessary mechanism — detailed, granular product carbon footprints (PCF) provided by every vendor — is simply not available at scale or with sufficient fidelity.

The problem persists because it requires collaboration across complex, often proprietary global supply chains. Suppliers are reticent to disclose granular data, citing competitive concerns, while buyers lack the leverage to mandate it. The result is a ‘Scope 3 plateau’: targets are set, but underlying emissions remain stubbornly high, creating a significant credibility risk. We are still largely measuring a reflection, not the reality.

2. The generative AI energy debt: While AI is a powerful tool for sustainability optimisation, the immediate, unmanaged energy demand of Large Language Models (LLMs) represents a profound and growing gap. The speed of AI adoption, combined with the inherently expensive High-Performance Computing (HPC) required, creates an “energy debt” that offsets hard-won gains elsewhere.

The challenge is governance. Enterprises are deploying AI solutions without robust, mandatory policies on model selection, inference efficiency, or resource decommissioning. Crucially, most organisations remain focused on achieving initial ROI metrics, relegating energy efficiency to an optional performance tweak. Failure to enforce a framework for ‘responsible compute’ risks the transformative power of AI being negated by its own expanding environmental impact. This is the single greatest risk to the IT sector’s net-zero journey.

Strategic priorities for 2026 and beyond

As the IT Sustainability Think Tank looks towards 2026, the focus must shift from identifying the problem to systematically closing the remaining gaps with institutional discipline. We must treat these priorities as non-negotiable elements of future business resilience:

1. Mandate data granularity for Scope 3: Leverage procurement influence to force supplier compliance on verifiable Product Carbon Footprints (PCF). The mandate must be non-negotiable, enforced with clear vendor scorecards and contractual requirements.
2. Institutionalise green software engineering: Invest heavily in training and tooling to embed energy efficiency into the software development lifecycle (SDLC). Software architecture must be treated with the same environmental scrutiny as data centre cooling, making efficiency an audited requirement.
3. Govern the AI energy cost: Implement a Responsible AI framework that includes mandatory energy consumption metrics and resource allocation policies for all Generative AI deployments.

The year 2025 was when IT sustainability moved into the board’s audit file. Next year must be the year we finally gather the granular data, enforce the necessary discipline, and manage the rapidly growing energy appetite of our own invention. The time for aspirational statements is definitively over; the urgent task now is to move these nascent efforts into full, verifiable accountability.