(Shutterstock AI Image)
We started the Powering Data in the Age of AI series with an examination of what might be AI’s most ignored constraint: energy. Not a shortage of models, but simply not enough power to run them as things stand at the moment. Part 2 followed that trail into fusion labs and nuclear reactors, where energy supply meets AI ambition. And in Part 3, the lens zoomed in as we moved from global power plays to immersion tanks, retooled racks, and the very physical redesign of AI data center infrastructure built for a whole new scale.
This fourth and final chapter explores the dynamics from a different angle: now that the infrastructure is here, who actually controls the levers? The battle for AI has shifted into a new phase. It’s no longer just about who writes the smartest algorithms or makes the fastest chips, but who controls the places where compute lives and perhaps even more importantly, who gets access to it. It comes down to jurisdiction, geography, and the right to plug in. This is where power gets traded, alliances get rewritten, and the real cost of falling behind starts to show.
The Global Map Is Being Redrawn in Compute
Global compute infrastructure has turned into a battlefield at the strategic level. The nesting of cloud regions, data corridors and undersea cables is beginning to define zones of influence in a way oil routes and trade chokepoints once did. Where data is located, where models are trained and where compute is purchased affect everything from jurisdiction to economic leverage.
Amazon’s push into Saudi Arabia is a case in point. The Kingdom is providing land, energy certainty, and political support in exchange for regional cloud infrastructure. It’s part of a broader effort to redefine Saudi Arabia’s geopolitical role in the world, from an oil provider to a digital power broker in the Middle East. Owning the compute lends it long-term leverage.
Global AI Data Center Locations (Source: NY Times)
Google’s €1 billion plan to expand its data center campus in Finland is about a lot more than access to renewables and efficient cooling. By building its infrastructure within the EU, the company ensures that its services are compliant with European data laws and can stay insulated from regulatory disputes going forward. For the EU, with critical compute on European soil, it solidifies sovereignty in a digital economy that has at times relied too much on providers outside its borders.
Singapore has been the infrastructure hub of Southeast Asia for many years. But tight land supplies and waning energy capacity have begun to cap its expansion. That has prompted nearby countries like Malaysia and Indonesia to offer incentives, available land and long-term energy contracts in a bid to attract hyperscale deployments. These aren’t overflow options. They are working aggressively to redraw the regional map in a manner more favorable to them.
These moves are not only about technical factors like user proximity and latency. It is about control. The nations that attract and host the infrastructure of the AI era will wield significant influence over all aspects of a computationally driven global society. Only now the map is being redrawn with server farms, cables and megawatt commitments rather than flags and borders. These new borders are not metaphorical. They are physical, potentially lucrative, tangled in geopolitics, and they will determine who has leverage for years to come.
Jurisdiction as a Defense Perimeter
Where AI infrastructure resides determines which government is in charge of what gets built, what gets trained and who has access. If servers or data centers are located in a country, that country typically has regulatory authority over data flows, access, and operations. So where the infrastructure lives often dictates which rules apply. So the location of compute in physical terms is something that can either open doors or slam them shut.
This kind of gatekeeping mentality is beginning to guide national strategies in a more profound sense. Countries are increasingly viewing data centers as muscle not only behind compute, but as a foundation beneath it.
(Gorodenkoff/Shutterstock)
As the World Economic Forum recently observed, these giant facilities are “the digital age’s equivalent of power plants or ports.” That change in perspective is animating industrial policy around the world as governments provide tax breaks, fast permits and regulatory carve-outs to bring compute within reach. Now the focus is less on turbocharging economic growth than on gaining leverage in a world system where access to compute power and software is gradually coming to define global influence.
In Europe, there’s a move to prevent critical workloads from leaving the European Union. This is not just about storing data anymore. Localizing processing, training and deployment is gradually becoming a priority. The Microsoft sovereign cloud service in Germany was created exactly for that reason. It preserves operations in the country, restricts who can control the systems and conforms with local laws from Day 1. This is infrastructure built to accommodate AI policy.
In the United States, enforcement is growing more aggressive. Cloud providers are now being called upon to watch their compute. It can have consequences if bad actors gain entry, even indirectly. It matters to regulators where the work is done, not just who owns the machines performing it.
This makes jurisdiction a type of power. It determines who builds, who trains, and who stays out. The law is the gatekeeper once infrastructure is established. In the age of AI, that gate is closing quickly for a few.
The Great AI Wall Is Already Under Construction
With every new data center breaking ground, the tension is rising. Infrastructure is growing fast, but so are the digital fences. Countries are quietly but decisively redrawing the outlines of their online borders — sometimes through new laws, sometimes by choosing exactly where their data centers are constructed. That familiar, now-defunct idea of a shared, borderless tech stack? It is vanishing. What we are witnessing is a fragmented world, divided by access, politics, and control.
(Source:DigitalInformationWorld)
China did not wait, of course. It built its own — a stack from top to bottom. Domestic chips, domestic clouds, homegrown language models. Even the operating systems have been rerouted. After U.S. sanctions were imposed, Huawei pushed HarmonyOS across phones, tablets, smart displays — everything. Not a mere patch, but a declaration: we will create the full digital chain ourselves, in our own way.
And others are moving fast too. The U.S. is tightening export controls, instructing cloud vendors to monitor usage, and treating AI computation as a matter of national interest. In Europe, “data sovereignty” isn’t a catchphrase. It’s the starting point for designing systems — from day one. These aren’t just regulatory patterns — they’re etched into hardware, baked into software, and embedded in every cloud contract.
Global compatibility isn’t what it used to be. Systems are still fast, still pushing boundaries, but they’re built around regional rules. Their own chips. Their own data laws. Interoperability now depends on whether two regions even want to talk. More and more often, they don’t.
For those without infrastructure of their own, the new rules are life-altering. They don’t get to dictate terms. They take what they’re given. Just look at the thousands of startups across Africa that rely on AI compute delivered from distant servers and uncertain cloud bills. The lines are drawn — through cables, chips, permissions — and they’re already deciding who stays in the race and who gets left behind.
The New Divide: Who Has Compute, and Who Rents It
The world has a quiet line running through it now — a divide that’s becoming harder to ignore. On one side are the nations and companies that own the compute. They’ve got the space, the money, and the political will to construct the massive data centers it takes to train today’s largest models. The other side, meanwhile, is those who rent, buying time on someone else’s platform, always a policy change away from disruption. The split is deepening.
The Google Cloud data center near Delfzijl in the Netherlands (Rudmer Zwerver/Shutterstock)
These buildings are today’s power plants — concrete, steel, racks of GPUs. They don’t just store data — they fuel what’s next. They decide who gets to innovate. In the U.S., the way NVIDIA’s GPUs and the cloud giants have combined has made this the easiest home on earth for today’s most important technology. France has responded with a €500 million push into national AI infrastructure, from shared GPU clusters to model training hubs. Japan’s economic ministry has earmarked H100 clusters just for its domestic AI builders — not to keep others out, but to keep their own moving.
Saudi Arabia isn’t waiting, spending the cash to ensure those zones get built within its borders. They’re trading land, power, and favorable rules for digital autonomy — the message is clear: they won’t just use AI; they want to be in charge.
Then there are the others stuck watching. In Kenya, startups train models on borrowed compute — Google Colab, some European GPU resellers, or whatever’s easy to access. In Bangladesh, researchers rent their compute off Singaporean clouds, watch its price swing, and spend days in the queue. These aren’t clever workarounds. They’re ceilings. To own compute is to hold leverage. To rent it is to remain dependent. As the race accelerates, it’s no longer just about who moves fast — it’s about who gets to move at all.
Related Items
The Quiet Rise of AI’s Real Enablers
Science Loses 90% of Its Data. A New AI Approach Could Change That
Goldman Sachs Chief Data Officer Warns AI Has Already Run Out of Data
