Neel Khokhani, Epochal Corporation, writes that for the past decade, the infrastructure investment community has operated on a seductive premise: that the future of computing belongs to whoever builds the biggest facilities.
Concentrate computation, reduce unit cost, compound efficiency at scale. In this view, scale itself becomes the strategy.
The Gulf states took this thesis to its logical extreme, and the numbers were extraordinary — the combined Saudi and UAE data centre market is projected to grow from USD35 billion this year to USD115 billion by the end of 2027, a rate of 62.5 per cent annually. Global technology companies raced to anchor themselves in the region’s infrastructure boom. Amazon committed USD5 billion to an AI hub in Saudi Arabia. Microsoft pledged USD7.9 billion in UAE investment by 2029. The Stargate UAE campus in Abu Dhabi, under construction as the largest AI facility outside the United States in collaboration with the US government, was described as capable of eventually serving half the world’s population.
For investors, the Gulf became the purest expression of a model the industry had largely accepted as settled. Then a simple question emerged: what happens when the world around these concentrated systems becomes less predictable?
On 1 March, Iranian drones struck three Amazon Web Services facilities in the UAE and Bahrain. The attacks disabled two of three availability zones in AWS’s UAE cloud region, with 11 million people across the UAE affected for days on end. For the first time, geopolitical risk collided directly with hyperscale digital infrastructure. The settled logic had a flaw, and the flaw had an address.
The flaw is this: concentration is efficient, but under the wrong conditions, it becomes a liability. Large data centres are enormous, above-ground structures with exposed cooling systems and significant heat signatures. They are not hardened in the way that military infrastructure is. And because the Gulf’s investment model placed enormous compute capacity into a small number of clustered facilities, two successful strikes were sufficient to take down a regional system that millions of people and businesses depended upon.
The infrastructure had been optimised for cost. It had not been optimised for resilience.
That vulnerability helps explain why a different model of digital infrastructure should be gaining momentum with investors all over the world: edge computing. Rather than concentrating computation in a handful of massive campuses, edge infrastructure distributes it across many smaller sites located closer to where data is generated and decisions must be made.
Three structural forces are now pushing investors in that direction. The first is the changing economics of artificial intelligence. Training large models still rewards hyperscale aggregation. But the next phase of AI operates under different constraints. Autonomous vehicles, industrial robotics, real-time medical diagnostics and AI-driven logistics require decisions in milliseconds. Routing those decisions to distant hyperscale facilities introduces delays that many real-world systems simply cannot tolerate.
Global demand for data centre capacity is projected by McKinsey to more than triple by 2030, driven overwhelmingly by inference workloads. The infrastructure supporting those workloads increasingly needs to sit closer to factories, hospitals, logistics hubs and telecom networks, not hundreds or thousands of kilometres away.
The second force is energy. The Middle East holds 4.5 gigawatts of existing data centre capacity with a further 1.7 gigawatts in the pipeline: sustaining that expansion requires concentrated megawatts at a moment when grid infrastructure globally is struggling to keep pace with AI-driven demand. Edge deployments, by distributing compute across multiple smaller sites aligned with local energy availability, offer a structural flexibility that singular concentration cannot match.
In the place of power price, power availability is increasingly becoming the binding constraint on which infrastructure assets can actually scale.
The third force, which the Gulf has now demonstrated with unusual clarity, is resilience. A distributed architecture does not offer a single point of failure. When compute is spread across multiple nodes, the loss of one facility does not cascade into a regional outage. Saudi Arabia’s HUMAIN programme is targeting 1.9 gigawatts of AI compute capacity by 2030, and that ambition is unlikely to disappear. But the architecture through which that ambition is delivered may look very different from the hyperscale campuses investors initially imagined.
Edge computing was already emerging as a distinct infrastructure asset class before this year. Its demand drivers — AI inference, energy constraints and the growing need for resilient digital systems — are structural and accelerating.
Training the world’s largest AI models will still require massive, concentrated facilities. But the next phase of digital infrastructure may be defined less by monuments of computation and more by networks of smaller, distributed assets.
The future of computing may not simply become bigger. It may become closer, and thus more resilient.
