When the United Nations published Who Cares Wins in 2004, it gave the world a new framework and a vocabulary. Environmental, Social, and Governance (ESG) criteria arrived with institutional weight behind it. Financial firms, analysts, investors and boards were all addressed directly with a simple message – sustainability is your business too.
In reality, the response, for the better part of two decades, has been largely performative.
ESG 1.0 produced reports, frameworks, disclosure regimes and ratings, and very little else of substance. The data was always there, measuring the environmental cost of infrastructure is not technically difficult, but accepting that data meant confronting uncomfortable realities, and most organisations discovered they preferred their sustainability commitments at a comfortable distance from their operational decisions.
The financial case was not exactly compelling either. A University of Chicago analysis of over 20,000 mutual funds found that the highest-rated ESG funds attracted more capital than the lowest-rated but failed to outperform them financially.
While research from Columbia University and the London School of Economics revealed something more damning: companies which added to ESG portfolios subsequently showed worse compliance records on both labour and environmental regulations. High ESG scores, it turns out, are better at attracting capital than at changing behaviour.
Enter the machine that changed everything
Then came AI and with it a problem that no amount of polite indifference could dissolve.
Training large language models, running inference at scale, and moving data across cloud environments requires staggering amounts of compute.
Cloud spend, already measured in the billions, has been pushed further by AI workloads that most organisations were never architected to absorb. The energy demands are real, continuous, and growing faster than most infrastructure teams anticipated.
Here lies the central tension of our current moment. The organisations pulling furthest ahead with AI are also those experiencing the sharpest deterioration in environmental performance and energy efficiency.
AI-driven businesses are often running compute at a scale that would have alarmed any ESG analyst five years ago. The environmental and energy cost of competitive advantage has never been higher or more visible.
This creates a dilemma that cannot be resolved by a quarterly sustainability report. It demands something ESG 1.0 never required: operational data, measured in real time, integrated into the platforms where decisions are actually made.
ESG 2.0: sustainability as an engineering discipline
ESG 2.0 is a structural shift from sustainability as a reporting exercise to sustainability as a core operational signal. Whereas ESG 1.0 measured after the fact, ESG 2.0 integrates real-time data into daily business decisions. Where ESG 1.0 produced side reports reviewed by compliance teams, ESG 2.0 embeds cost and carbon signals directly into developer workflows.
Much as renewable energy stopped being a moral argument when it became a cheaper and more reliable option than pollutants, ESG 2.0 follows the same logic, when GreenOps lives in dashboards disconnected from developer workflows, organisations effectively hard-code a silo of waste, and expense, into their infrastructure.
When cost and carbon signals arrive weeks after the code that generated them is already in production, optimisation becomes slow, frustrating and, in practice, ignored.
GreenOps: efficiency is the new ethics
What has changed the conversation is GreenOps and its relationship to FinOps. In cloud environments, financial and environmental impact are driven by the same decisions around instance sizing, storage choices, data movement, and how long services remain running. These decisions are made constantly, at build and deploy time, by development teams who currently have little visibility into their consequences.
Put simply, cost and carbon are driven by the same infrastructure decisions made every day by development teams. In the cloud, financial and environmental impact cannot be separated.
This is where the moral position ends and the business case begins. When reducing energy consumption reduces cloud spend, which with AI workloads can be substantial, sustainability stops sounding like a constraint on innovation and starts looking like financial discipline.
Rightsizing, elasticity and automation reduce waste and idle resources, improve delivery speed and free up budget for work that actually performs. Sustainability is now about consuming better.
The transparency AI demands
ESG 1.0 failed to realise its well-intentioned ambitions because organisations had no structural incentive to act on it. AI has created that incentive as the compute costs are too large to ignore, the competitive pressures too intense to waste, and the infrastructure decisions too consequential to leave unmonitored until the next quarterly review.
The same transparency that sustainability professionals spent years advocating for has become a technical prerequisite for anyone running AI at scale.
That may be a less idealistic origin story than the one the 2004 UN report imagined. But it is a far more durable one.
