The Cloud Doesn’t Exist. And That’s the Problem.

Reflections from Data Decoded London, April 2026.

I spent two sessions on stage at Data Decoded last week, one in the morning and one in the afternoon, with two very different conversations and a single argument running underneath both. By the end of the day I was more convinced than ever that the sustainability problem in front of large organisations is not a data problem, not a measurement problem, and not a strategy problem. It is a system problem. And until we treat it that way, we will keep producing reports about progress that is not happening.

The morning conversation: the behavioural ground

I opened the day in dialogue with Samantha Davies, Director of Global Product Insight and Analytics at Sky. Sam and I had started this conversation six months earlier at the same event. The question we left hanging in October was whether data could become part of the sustainability solution rather than just another footprint to manage. We picked it up in April with the volume turned up, because in the intervening six months AI has moved from a strategic discussion to a daily operational reality for most organisations in the room.

Sam’s frame from October was that the challenge is partly behavioural. Data teams are trained to optimise for financial efficiency, not environmental efficiency. Six months on, the gap is still there, but the awareness has shifted. People are asking whether the model needs to be that big, whether the query needs to run that often, whether the storage needs to be that hot. That is the start of a different conversation. It is not yet a different system.

What I argued, and what I find in every organisation I work with, is that the data is rarely the problem. The strategy is almost never the problem. What you find is a system that has not been designed to deliver. Procurement is measured on cost. Product on speed. Finance on margin. Sustainability advises but does not decide. Everyone is doing their job. Nothing is moving.

That is the diagnostic. And once you see it, you stop looking for the missing data point and start looking for the missing accountability.

The afternoon conversation: the physical reality

The afternoon session sharpened the same argument by giving it a physical address. I was joined by Andy Gomarsall, executive director of N2S and founding partner of Bioscope Technologies, and Mark Butcher, managing director of Posetiv. Both of them have spent serious time staring at what the cloud actually looks like when you take the metaphor away.

I started with a provocation. The word cloud is one of the most consequential acts of industrial rebranding in modern history. The metaphor did exactly what metaphors are designed to do. It made something disappear. An entire industry’s physical footprint, the water, the land, the energy, the rare earth materials, the labour, became invisible the moment we agreed to call the infrastructure something soft and weightless. And then we built AI on top of it.

The numbers the session surfaced should not be controversial, but they rarely make it into the conversations data leaders are having about their own infrastructure. Less than one per cent of rare earth metals are currently being recycled. Global e-waste has grown from 90 million tonnes to 120 million tonnes. For every pound spent on cloud, 50 pence is wasted, paid for, provisioned, and never used. Corporations are systematically under-reporting the water consumption of their data centres. And the global economy, far from becoming more circular, has gone from 9 per cent circular two years ago to 7 per cent today. We are moving in the wrong direction.

These are not numbers that exist in some separate ESG conversation. They sit underneath every dashboard, every model, every storage decision being made by the same people who were in the morning room with Sam.

The accountability question nobody wants

The moment that has stayed with me came from Mark. He made the point that corporations are not, on the whole, moving toward sustainability commitments voluntarily. Many have lobbyists actively working to dilute the legislation that would hold them accountable. Waiting for regulation is not a strategy. Waiting for the market is not a strategy. The pressure that actually moves these systems comes from inside organisations, and from the customers, regulators, and employees outside them who refuse to accept the current settlement.

That is uncomfortable to hear in a room full of people who, individually, are trying to do the right thing. But it is honest. And it points at where the leverage actually sits. Not in the sustainability function. In the data, engineering, procurement, and finance functions where the real decisions about cloud infrastructure are being made every day.

Andy made the related point on the circular side. The technology sector cannot decarbonise on a foundation of virgin extraction. Urban mining, the recovery of critical materials from existing devices and infrastructure, is not a romantic idea. It is an economic opportunity that is being left on the table because the procurement and design conversations are not connected to the end-of-life conversations. That disconnect is not a sustainability failure. It is a system design failure.

What responsible actually looks like

The phrase responsible data infrastructure gets thrown around a lot. In practice it is simpler than the language suggests. It looks like procurement contracts that price in lifecycle impact, not just unit cost. It looks like architecture decisions that ask whether the model needs to be that big, whether the storage needs to be that hot, whether the query needs to run that often. It looks like commercial teams whose scorecards include the sustainability outcomes they are actually driving, rather than outcomes set by one team and delivered by another.

This is what links the morning and afternoon conversations. Sam and I were talking about behavioural change inside data teams. Andy, Mark and I were talking about the physical and economic consequences of decisions those data teams are making. The two conversations are the same conversation. The behavioural shift only matters if it is wired into the system. The system only changes if the behavioural shift is supported by accountability, incentive, and authority sitting in the same place.

Where this leaves us

The cloud does not exist. It never did. It is land. It is water. It is energy. It is the labour of people mining materials in conditions most of us will never see. Every query, every model, every dashboard has a physical address.

The question is not whether organisations know this. Most of the people in those rooms knew it before they walked in. The question is whether they are willing to redesign the system that currently makes it nobody’s job to act on what they know.

That is the work. The gap between knowing and doing is not a data problem. It is a system problem. That is what we fix.