Digital transformation has run out of road, and AI is now forcing a far more fundamental redesign of how enterprises think about infrastructure. The real question is no longer whether organisations adopt AI, but whether their data architecture can survive it.
Digital Realty’s Rewire for Data and AI Playbook makes an uncomfortable point that many boardrooms are still avoiding. AI is not another application layer to be absorbed into existing IT estates, it is a structural force that reshapes how data behaves, where workloads must live, and how organisations need to organise themselves around infrastructure. The implication is not incremental change, but architectural rewiring, where physical systems, organisational design and strategic ambition become inseparable.
For years, digital transformation was framed as a software problem. Enterprises invested in cloud platforms, ERP systems, analytics tools and collaboration layers, assuming that intelligence would naturally emerge from digital maturity. AI breaks that assumption. Intelligence now depends less on interfaces and more on how data is physically created, moved, stored and processed. Infrastructure is no longer the plumbing beneath strategy; it is part of the strategy itself.
What emerges from Digital Realty’s research is a picture of enterprises caught between intent and capability. Most organisations now recognise data as a strategic asset, with formal data strategies guiding infrastructure investments and increasing budgets allocated to analytics platforms. Yet only a minority can translate this into measurable AI-driven outcomes. The constraint is not imagination or executive sponsorship; it is the physical and organisational limits of legacy infrastructure.
Data gravity is reshaping enterprise behaviour
One of the most important concepts running through the playbook is Data Gravity. As data volumes grow to feed machine learning, generative and agentic models, that data attracts applications, services and workflows into ever tighter proximity. Over time, data becomes harder to move, more expensive to replicate, and more difficult to govern. This creates a gravitational effect where infrastructure choices made early in an organisation’s digital journey quietly determine what becomes possible years later.
In practical terms, Data Gravity means that enterprises can no longer treat data location as a technical detail. Where data sits increasingly defines how quickly models can be trained, how reliably insights can be generated, and how effectively regulatory obligations can be met. The more data accumulates in each environment, the more systems are forced to orbit around it, creating architectural lock-in that is difficult to unwind.
This has profound implications for AI deployment. Models rely on proximity to large, high-quality datasets, and moving those datasets across fragmented environments introduces latency, cost and security risk. What appears on paper as a flexible cloud strategy often becomes a patchwork of loosely connected systems that slow down every stage of the AI lifecycle.
The strategic consequence is that infrastructure stops being a support function and becomes a primary business constraint. Decisions about colocation, edge deployment, regional data centres and private interconnection are no longer operational optimisations, they are determinants of competitive capability. Organisations that fail to address Data Gravity early find that their AI ambitions are quietly constrained by architectural decisions made years before.
The cloud is not enough
For more than a decade, cloud computing has been treated as the default answer to almost every infrastructure problem. Elastic capacity, global reach and consumption-based pricing created a compelling narrative of infinite scalability. AI exposes the limits of that narrative.
Training and deploying modern AI systems requires specialised hardware, particularly GPUs optimised for parallel processing. These workloads demand high-density environments, predictable performance and low-latency access to large datasets. General-purpose cloud environments struggle to meet these requirements at scale, especially when workloads must run continuously rather than intermittently.
Latency becomes a business constraint rather than a technical metric. Many AI applications require real-time or near-real-time responses, which are difficult to guarantee when data must travel across long network paths. At the same time, regulatory frameworks increasingly restrict where sensitive data can be processed, limiting the extent to which cloud centralisation remains viable.
Cost also re-enters the conversation in uncomfortable ways. Running high-performance AI workloads continuously in the cloud can become prohibitively expensive, particularly when data egress fees and specialised hardware premiums are factored in. What initially looks like operational simplicity often becomes long-term financial rigidity.
The result is a shift away from pure cloud models towards hybrid, distributed and colocated architectures. Enterprises are being forced to bring compute closer to data, using edge deployments, regional data centres and private interconnection to regain control over performance, cost and governance. Infrastructure becomes less about centralisation and more about intelligent distribution.
Legacy infrastructure as a silent failure point
One of the most damaging assumptions identified in Digital Realty’s playbook is the belief that existing infrastructure will continue to work simply because it always has. Many organisations still treat AI as either a modelling problem or an extension of cloud capacity, leaving underlying data architectures largely untouched. This creates silent failure points that only become visible once AI initiatives begin to scale.
Latency accumulates as data is moved between fragmented systems. Data sprawl undermines governance as information is duplicated across multiple platforms without clear ownership or accountability. Siloed teams pursue disconnected strategies, with AI roadmaps divorced from infrastructure planning. Over time, these structural weaknesses transform AI programmes into expensive experiments rather than sustainable capabilities.
The irony is that many of these organisations appear highly advanced on the surface. They have data scientists, cloud partnerships, pilot projects and executive sponsorship. Yet beneath that veneer, infrastructure remains brittle, slow and fragmented. AI becomes a layer of complexity added on top of fragile foundations.
Digital Realty’s framing of AI maturity is therefore deliberately uncompromising. True maturity is not defined by the number of models deployed or dashboards produced, but by whether AI has been embedded into the business model through a distributed data strategy. Only a small proportion of enterprises currently meet that standard.
The risk is not missing the AI wave altogether but mistiming it. Organisations that delay infrastructure rewiring may find that by the time their AI use cases are ready for scale, their ability to deploy them competitively has already been undermined by technical debt.
Six imperatives for rewired infrastructure
The playbook structures rewiring around six imperatives that reveal how deeply AI reshapes infrastructure logic. These imperatives move beyond technology choices and expose the systemic changes required across architecture, governance and organisational design.
Capacity becomes dynamic rather than forecastable. AI workloads are unpredictable, with sudden spikes in demand and rapidly evolving requirements. Infrastructure must scale up and down efficiently without chronic over-provisioning or performance bottlenecks.
Complexity shifts from a technical inconvenience to a strategic liability. As systems become more interconnected, brittle integration layers multiply risk and slow innovation. Pre-integrated, modular architectures reduce the burden of stitching together disparate components and allow organisations to evolve without constant reengineering.
Latency becomes a first-order business constraint. Bringing applications to data rather than data to applications minimises response times and improves model performance. Edge computing, regional data centres and colocated resources become essential parts of the AI stack rather than optional enhancements.
Compliance evolves from a legal function into an architectural principle. Data residency, privacy and regulatory requirements must be built into infrastructure design rather than retrofitted through policy. Distributed architectures and in-region processing become critical to maintaining operational freedom.
Risk becomes systemic rather than localised. Infrastructure failures no longer affect isolated systems; they disrupt entire AI ecosystems. Validated platforms, robust governance controls and auditable processes become prerequisites for sustainable deployment.
Finally, Sustainability moves from corporate messaging into core design criteria. AI’s energy demands make efficiency a strategic concern rather than a reputational one. Energy-optimised hardware, advanced cooling and green data centres become essential components of responsible AI infrastructure.
Together, these imperatives describe a world where infrastructure behaves more like a living system than a static platform. AI-ready environments must adapt continuously, absorb uncertainty and evolve alongside business strategy.
One enterprise, one mindset
Perhaps the most overlooked dimension of rewiring is cultural rather than technical. AI transformation cannot succeed if different functions pursue independent strategies. IT, data, infrastructure, security and business leadership must converge around a single architectural vision.
Digital Realty frames this as a shift towards one enterprise, one mindset, one AI strategy. This is more than governance rhetoric. It reflects the reality that AI systems cut across every organisational boundary, from customer interaction to supply chain optimisation to regulatory compliance.
When teams operate in silos, infrastructure decisions become fragmented. AI initiatives compete for resources rather than reinforcing each other. Technical debt accumulates as each function optimises for its own priorities. The organisation becomes structurally incapable of learning from itself.
Rewiring therefore requires breaking down institutional habits as much as technical ones. Strategy must be shared. Decision-making becomes collective. Architecture stops being something designed by specialists and becomes something owned by the enterprise. This cultural shift is often more difficult than any technology upgrade. It challenges power structures, budget allocations and long-established roles. Yet without it, infrastructure remains misaligned with business ambition, and AI becomes a series of disconnected projects rather than a coherent capability.
Infrastructure as competitive advantage
The most radical implication of the playbook is that competitive advantage in the AI era is shifting away from software differentiation and towards infrastructure positioning. Where data sits. How close compute can get to it. How quickly models can be trained, deployed and governed. How resilient systems are to regulatory change and geopolitical constraints.
These are not product features; they are architectural conditions. Two organisations may use similar models and tools, but the one with better infrastructure will deploy faster, scale more reliably and adapt more effectively to change. AI therefore becomes less about who has the smartest algorithms and more about who has built the least fragile foundation. This reframes investment priorities. Spending on infrastructure is no longer defensive or operational, it becomes directly tied to growth, innovation and strategic flexibility.
The deeper tension is that AI is evolving faster than organisational adaptation. Digital Realty observes that it is now seeing a year’s worth of AI innovation compressed into each quarter. Yet infrastructure investment cycles, regulatory frameworks and enterprise culture still move at industrial speed.
This creates a structural mismatch where ambition outpaces capability. Without deliberate rewiring, organisations accumulate complexity faster than they can absorb it. AI becomes a multiplier of fragility rather than a source of resilience.
The cost of not rewiring
What ultimately emerges from Rewire for Data and AI is not a technology roadmap but a warning. AI does not fail because models underperform. It fails because infrastructure cannot absorb the consequences of success.
Data grows too fast. Latency becomes visible. Compliance breaks. Costs spiral. Teams fragment. At that point, no amount of algorithmic innovation can compensate for architectural inertia.
Rewiring is therefore not an upgrade project; it is a recognition that AI changes the physics of the enterprise. Data becomes gravitational. Infrastructure becomes strategic. Organisational design becomes part of system performance.
Enterprises that treat architecture as an afterthought will discover that their AI future was already broken before it began. Those that take rewiring seriously will find that infrastructure itself becomes a source of strategic advantage, shaping not just what they can build, but how quickly they can learn, adapt and compete.
