Scarcity is driving AI innovation outside Silicon Valley

For years, the assumption around artificial intelligence infrastructure was that serious compute would be built in places where hyperscale cloud, developer density, and capital were already concentrated: Silicon Valley, Seattle, London, and a small number of other technology hubs. There was a practical reason for these geographies. Training and deploying AI at scale requires data centers, compute, networking capacity, energy, and advanced infrastructure. Over time, that dependence has hardened into market concentration. Amazon, Microsoft, and Google together account for nearly two-thirds of global enterprise cloud infrastructure spending. That earlier logic no longer holds. Compute is becoming more expensive, more power-intensive, and harder to access outside a small group of dominant providers. Builders are starting to confront questions like: Where will the power come from? Can chips be shipped to this jurisdiction? Whose laws apply to the data once it moves? Those questions are increasingly being answered outside Silicon Valley. What scarcity teaches In established cloud markets, the default answer to rising AI demand is to add more capacity through larger cloud contracts, denser data center build-out, and deeper dependence on the same centralized stack. Data centers consumed about 1.5% of the world’s electricity in 2024, enough to make energy one of the pressure points in AI infrastructure. That share is forecast to rise to just under 3% by 2030, making compute harder to treat as a hidden layer behind AI products. In much of the developing world, that pressure is the starting point. Builders have rarely had the option of treating compute access, power, and distribution as someone else’s problem; they have had to design for it. The result is that serious AI infrastructure is now being built in places where scarcity is treated as a design problem, not as an afterthought. The pattern is most visible in four places. In India, Yotta Data Services runs Shakti Cloud on more than 16,000 Nvidia H100 graphics processing units, with plans to add more. Over half the compute behind the IndiaAI Mission — the government program to build indigenous foundation models — sits on Yotta’s hardware. Earlier this year, the multilingual platform Bhashini moved off foreign hyperscalers and onto Shakti Cloud. Bhashini runs real-time translation across 11 Indian languages, and a decision was made that infrastructure they could not govern was not acceptable. In Africa, Cassava Technologies, founded by Zimbabwean entrepreneur Strive Masiyiwa, is deploying 12,000 Nvidia GPUs across data centers in South Africa, Egypt, Kenya, Morocco, and Nigeria. Cassava is the first Nvidia cloud partner on the continent; before this build-out, Nvidia estimated that roughly 80 of its GPUs were installed across the African continent. The constraint was not only pricing, it was the absence of advanced silicon. Cassava’s response is a pan-African network running on its own fiber backbone, so that African startups, researchers, and governments do not have to route through Europe or the U.S. to train and deploy AI. In Brazil, the government’s SoberanIA project reserves 500 megawatts for a sovereign AI factory in Piauí, powered entirely by renewable energy. Scala Data Centers is its lead infrastructure partner. Brazil has committed to attracting up to $370 billion in data center investment over the next decade, tied to tax incentives for projects sourcing 100% renewable power. Roughly 65% of Brazilian data is still stored abroad. The wager is that abundant hydroelectric and solar power will give Brazil a cleaner starting point for sovereign compute than markets where data-center power still leans heavily on gas or coal. The United Arab Emirates is taking the most expensive route. Core42, part of the G42 group, sells inference capacity on a mix of Nvidia and Qualcomm chips, and the country has committed, with the U.S., to a 10-square-mile, 5-gigawatt AI campus that should be partially operational by the end of the decade. The Emirati pitch is straightforward: Countries that want sovereign AI but cannot build the underlying stack themselves can rent one from a friendly government. It is a deliberate strategy of vertical integration — owning the chips, the power, the data centers, and the foreign relationships in one go. The projects all share the starting assumption that compute access, power, land, and chip supply are first-order design problems rather than externalities. That assumption produces different infrastructure. Why inference changes the map Training large models still rewards dense clusters, large capital resources, and access to advanced chips. That work is unlikely to leave the largest hyperscale facilities soon. Inference is different. Models are used continuously, by customers, devices, agents, and enterprise systems. McKinsey expects inference to overtake training in AI data centers by 2030, accounting for more than half of AI compute, and roughly 30%–40% of data center demand. For inference, the questions are where compute should sit, how fast it can respond, how reliably workloads can be routed, and whose laws govern the data. Those questions have geographic implications that hyperscale concentration cannot handle well, especially for the billions of people who do not live within easy latency of a U.S. or European data center. The compute fabric that inference demand requires is broader than hyperscale cloud alone can provide. Distributed GPU capacity, regional inference clusters, sovereign clouds, and emerging neoclouds in places such as Mumbai, Nairobi, São Paulo, and Abu Dhabi are not substitutes for hyperscale — they are the layer hyperscale cannot serve on its own. The old map of AI infrastructure was drawn around places where cloud capacity was already concentrated. That map made sense when compute was cheap and abundant. The new map will look different. It will be drawn around places that have learned to build when compute is costly and strategic, and where the question of who controls the stack is more consequential. The companies and governments drawing the new map are not catching up to Silicon Valley; they arrived at the problem first because they had to.