Why the future of cloud will be closer than we think
For much of the past decade, the direction of travel in digital infrastructure has been towards centralisation. Public cloud and hyperscale data centres have transformed the way organisations consume technology, offering extraordinary scale, flexibility and access to computing power.
That model is not going away. But increasingly, it is no longer the whole answer.
AI, growing volumes of data, new sovereignty requirements and the need for greater resilience are changing the infrastructure conversation. Organisations are beginning to ask not simply how much compute do we need?, but where should that compute actually sit?
The government’s new Data Centre Covenant adds another dimension to that conversation, linking infrastructure growth with national resilience, environmental performance and benefits for local communities. For the sector, this raises a practical question: how can investment in computing capacity serve both the organisations using it and the places hosting it?
This is why the next chapter of cloud is likely to be much more distributed.
From cloud-first to workload-first
The debate is sometimes presented as a choice between cloud and physical infrastructure. In reality, most organisations are already operating somewhere in between.
Infrastructure increasingly spans public cloud, private cloud, SaaS platforms, on-premise technology and colocation. The challenge is working out which environment makes sense for each workload.
For some workloads, hyperscale cloud is exactly the right answer. For others, factors such as latency, predictable performance, data volumes, cost, regulation or resilience may make infrastructure closer to users and operations more appropriate.
Regional edge data centres offer a way to host selected workloads closer to users and operations while retaining connections to public cloud services. Their role is not to replace hyperscale infrastructure, but to give organisations more choice over where individual workloads sit.
That feels like a much more useful question for technology leaders than simply asking whether they should be "in the cloud".
Location matters again
Physical location can sometimes disappear from the cloud conversation. But data still has to be stored, processed and transported somewhere.
As real-time applications, AI inference and data-intensive services become more common, distance can matter.
Bringing certain applications and data closer to where they are generated or consumed can reduce latency and unnecessary data movement. But proximity alone is not enough: network routes, bandwidth and reliable connections to cloud platforms determine whether a regional location delivers a practical advantage.
This does not replace the hyperscale data centre. It complements its role within a broader architecture.
Think of hyperscale facilities as providing enormous concentrations of capacity and cloud capability, while regional edge facilities bring elements of that digital infrastructure closer to businesses, users and data sources.
The most effective architecture can use both.
AI makes that distinction more important
AI adds another dimension.
There is understandably huge attention on the infrastructure required to train the largest models. But AI infrastructure is about more than building ever-larger concentrations of GPUs.
Some AI inference workloads benefit from being closer to users or data sources, particularly where response times or data movement matter. Others remain well suited to centralised cloud infrastructure. The choice depends on the application, alongside its requirements for connectivity, power density, cooling and resilience.
AI readiness is therefore not simply a question of installing GPUs. It means matching power, cooling and connectivity to the workload, while considering interoperability and sustainability from the outset.
A real-world example shows what this can look like. DCENT used nLighten's London facility as an additional European region for decentralised storage, AI and Web3 workloads, deploying more than 800 hard drives within weeks, alongside high-performance GPU and storage servers and redundant 100G connectivity.
The interesting point is not the equipment itself. It is that infrastructure can be distributed across locations while still operating as part of a wider digital ecosystem.
Sovereignty is increasingly about control
There is also a growing European conversation around digital sovereignty.
Data residency is important, but sovereignty is broader than knowing which country contains a server. Organisations are increasingly considering jurisdiction, infrastructure ownership, access, connectivity, encryption and operational control.
That is particularly relevant for regulated industries, government, healthcare, financial services and organisations supporting critical infrastructure.
Regional infrastructure can form part of an architecture that keeps sensitive workloads under defined controls while retaining access to global cloud services. But a regional address is not a sovereignty guarantee: organisations still need to assess jurisdiction, access rights, encryption-key control and operational dependencies.
Digital infrastructure can also belong to a place
The new Data Centre Covenant makes local benefit an explicit part of the infrastructure debate. Regional facilities have an opportunity to make that relationship tangible: connecting local businesses to wider cloud ecosystems, supporting skills and supply chains, and exploring heat reuse where suitable demand exists. The test is not simply whether a data centre is located in a region, but what useful connections it builds there.
They can support local businesses, technology partners and connectivity ecosystems. Their investment can contribute to regional skills and supply chains. Their energy strategy can increasingly consider interaction with local grids and the potential reuse of heat, rather than treating the data centre as an isolated building.
These benefits are neither automatic nor exclusive to regional facilities. Heat reuse, for example, depends on suitable nearby demand and supporting infrastructure. Local integration should be a practical design consideration, not simply a claim attached to a regional postcode.
As demand for digital infrastructure accelerates, we should therefore resist the temptation to assume that bigger and more centralised is automatically better.
The future is unlikely to be hyperscale or edge.
It will be hyperscale and edge, public and private, centralised and regional.
The organisations that get this right will be those that place each workload where it makes the most sense, balancing scale with proximity, innovation with control, and global digital capability with resilient local infrastructure.
And increasingly, that means the future of the cloud may be much closer than we think.
