Artificial Intelligence, AI is no longer just a frontier of innovation but is fast emerging as a fault line where a new kind of digital divide is taking shape. One that splits the world between those with access to immense AI compute power, and those left watching from the sidelines.

Make no mistake, this is not your typical disparity in broadband or smartphone access. This is a structural, high costing divide, one that’s influencing geopolitics, reshaping economies, and quietly redrawing the global map of technological sovereignty.

Take for example – just last month, OpenAI CEO Sam Altman stood at the heart of a transformation, at the site of his company’s upcoming $60 billion data center in Texas. When complete, the facility, which spans more ground than New York’s Central Park and has its own natural gas plant, will become one of the world’s most powerful AI infrastructure projects.

At almost the same time in Argentina, computer science professor Nicolás Wolovick was running one of the country’s most advanced AI research labs (out of a modest university room with aging hardware and tangled wires) “We are losing,” he admitted.

He’s not wrong. According to Oxford University researchers, only 32 countries, roughly 16% of the world, possess the kind of large-scale data centers needed to build cutting-edge AI systems. And more than 150 nations have virtually zero compute capacity. Not surprisingly, the United States, China, and the European Union dominate the space, hosting more than half of all advanced AI compute hubs. U.S. and Chinese firms alone run over 90% of the world’s most critical AI data centers.

Furthermore, compute power is not just a tech metric anymore; it is a new form of leverage. The chips at the heart of these systems, mainly from Nvidia, have become strategic assets. Access to them is shaping trade deals, foreign policy, and national ambitions. AI is now where microchips meet realpolitik.

Take languages. The most widely used AI systems today, think ChatGPT and its peers, are naturally more fluent in English and Chinese, the tongues of the countries that own the most compute capacity, it’s a byproduct of the infrastructure and it reinforces who gets to shape AI’s global story.

And this divide is already taking a toll. Countries without AI muscle are facing obstacles in scientific research, startup growth, and talent retention. And the concern is not limited to economic lag but rather about being locked into dependency.

As Vili Lehdonvirta, an Oxford professor, aptly put it: in a future defined by AI, compute power could become what oil was in the last century, a core source of geopolitical clout.

Meanwhile, governments are waking up. From India to Brazil, and parts of Africa to the EU, public funds are beginning to flow into AI infrastructure in a bid to reclaim digital sovereignty. But the reality is, building sovereign AI requires more than ambition – it needs chips, power, talent, and time.

The Digital Divide Never Really Closed
Over the last decade, it looked like the global tech gap was narrowing. Smartphones became cheap. Internet access spread rapidly. App economies boomed across continents. From ride-hailing in Southeast Asia to mobile payments in Africa, it seemed the digital revolution was finally going global.

But that sense of optimism may have been premature. In April, the United Nations cautioned that the next digital divide, powered by artificial intelligence, could be far worse. Today, just 100 companies, overwhelmingly based in the U.S. and China, are responsible for nearly 40% of all global AI investment. In a world increasingly run on algorithms, that kind of concentration spells trouble.

The Chip Bottleneck
At the core of the issue is one hyper-competitive commodity: the GPU. Graphics processing units are the silicon engines that drive generative AI, and they’re not easy to come by. Manufactured in billion-dollar fabs and hoarded by the largest players, these chips (mainly made by Nvidia) are the lifeblood of any serious AI initiative.

Prices are skyrocketing, waitlists are growing, and even getting in the queue often requires geopolitical leverage. And once you have the chips, you still need massive data centers to house them, power-hungry, resource-intensive complexes that many countries simply can’t afford to build.

The Cost of Distance, Why Renting Compute Isn’t a Solution
For the global South, renting compute power from the tech giants sounds like a workaround. But it’s far from ideal. The costs are high, connections can be slow, and compliance with U.S. or Chinese tech governance can be murky. What’s worse, you are at the mercy of providers who may change terms, restrict access, or simply prioritize their own customers first.

Take Qhala, a Kenyan AI startup led by a former Google engineer. They’re trying to build a large language model tailored to African languages. Without any local computing infrastructure, to stay ahead of the curve, their team works around American downtime hours just to access faster transfer speeds.

“Proximity is essential,” says founder Shikoh Gitau. And she’s right.

The Infrastructure Arms Race, Billions for the Privileged Few
While companies in Kenya and Argentina scramble for scraps of compute, American tech giants are on an infrastructure binge. Amazon, Microsoft, Google, Meta, and OpenAI have pledged over $300 billion, yes, billion, with much of it earmarked for AI infrastructure.

To put that in perspective, it’s roughly equivalent to Canada’s entire national budget.

Harvard’s Kempner Institute alone boasts more AI compute power than the entire African continent’s locally owned facilities combined. It’s a staggering imbalance that shows just how steep the climb is for most of the world.

Brain Drain 2.0

The implications of this divide go beyond business, they’re deeply human. In Argentina, computer science professor Nicolás Wolovick is fighting a losing battle to keep his best students from leaving. The talent is there. The ambition is there. But the hardware? Not so much –  I don’t give up. I keep talking to people and saying: ‘I need more GPUs. I need more GPUs.’”

To some, this is more than a tech supply chain issue, it’s a question of digital sovereignty. Microsoft President Brad Smith put it bluntly: without the right infrastructure, entire regions risk being left behind in the AI era.

Microsoft is now building a data center in Kenya with UAE’s G42, but the company’s site selection still depends heavily on stable electricity, reliable markets, and skilled labor. Meanwhile, Nvidia is working with governments around the world to build capacity but even they admit, “it is absolutely a challenge.”

OpenAI has launched initiatives to localize its products and adapt them to underserved regions and languages. But as Chris Lehane, its VP of Global Affairs, warned, the risk is real: if AI benefits aren’t shared equitably, “they don’t get democratized.”

So far, many tech giants like Tencent, Alibaba, Google, Amazon have declined to comment. But the silence speaks volumes.

Two Tech Superpowers, One Global Tug-of-War

The global AI race is no longer just about innovation, it’s about alignment. And right now, nations around the world are being pulled into two distinct spheres of influence – those leaning on U.S. infrastructure and those turning to China.

Between them, the U.S. and China dominate the AI compute scene and are building more data centers than any other countries. But they’re not just building, they’re leveraging – through trade restrictions, strategic investments, and geopolitical bargaining, both powers are using compute access as a new-age tool of diplomacy and control.

In the 2010s, China made deep inroads into Middle Eastern tech ecosystems with generous loans and aggressive infrastructure deals. But the U.S. has since pushed back. A recent Biden-era deal with an Emirati firm involved a commitment to sideline Chinese tech – Nvidia chips and Microsoft cloud access came as the incentive. In May, President Trump went further, signing off on additional semiconductor access for Saudi Arabia and the UAE.

The AI Battlefield Expands
The jockeying isn’t limited to the Gulf. In Southeast Asia, cloud titans like Amazon, Alibaba, Google, ByteDance, and Nvidia are in a full-on sprint to build data centers in Singapore and Malaysia. It’s a land-and-chip grab in real time.

Globally, the U.S. has a clear lead – American companies have built 63 AI hubs outside the country, versus China’s 19. And despite Beijing’s push for homegrown chips, most Chinese-operated data centers abroad still rely on Nvidia’s technology, bought before Washington tightened the screws.

Ironically, even U.S.-friendly countries are getting edged out. Kenya’s President William Ruto dined at the White House last year, but months later, the country was excluded from a list of nations with open access to U.S. semiconductors. China, sensing the gap, is now in talks to retrofit African data centers with Huawei’s AI chips, even if they’re not quite on par with Nvidia’s.

As Lacina Koné of Smart Africa summed it up: “Africa will strike a deal with whoever can give access to GPUs.”

The Sovereignty Shuffle
Alarmed by this growing imbalance, several nations are taking bold steps to close the gap. Governments are offering cheap land, energy subsidies, and fast-track clearances to attract compute infrastructure investment. The idea is simple but ambitious: build “sovereign AI” infrastructure that local companies and institutions can own, control, and rely on.

India is already backing compute-heavy projects and developing an AI model fluent in regional languages. Africa is exploring regional data-sharing pacts. Brazil has pledged $4 billion toward AI investments. Even the European Union, long dependent on U.S. cloud firms, is planning a €200 billion push to build data centers and reclaim strategic autonomy.

But as Mathias Nobauer, CEO of Swiss cloud provider Exoscale, points out: “This doesn’t happen overnight.”

In Africa, one of the boldest moves on the continent is coming from Cassava Technologies, the Zimbabwean-founded company spearheading a $500 million initiative to build five data centers across Africa. The first of these, set to open this year, was greenlit after a pivotal meeting between Cassava executives and Nvidia CEO Jensen Huang in California. Google is among the firm’s early investors.

Still, even Cassava’s state-of-the-art facility will meet only 10–20% of projected regional demand. Over 3,000 startups have already expressed interest in tapping into the system. It’s a drop in the bucket but it’s a start.

SoftBank’s India Pivot
Meanwhile, India may not yet be in the league of nations with foundational AI models, but that hasn’t deterred Masayoshi Son’s SoftBank from spotting an opportunity in the gap. In a strategic pivot from passive funding to active ownership, the Japanese investment giant is now exploring direct acquisitions in the country’s outsourcing sector.

SoftBank has already held talks with AGS Health in what was shaping up to be a potential $1 billion deal. While Blackstone ultimately took the lead, the conversation signaled a shift. SoftBank is also said to be in discussions with other mid-sized players like WNS Global, with one clear goal in mind – marrying AI infrastructure with India’s globally respected IT services engine.

The rationale is simple, India doesn’t yet have a homegrown equivalent of OpenAI or Anthropic. But what it does have is deep domain expertise in sectors like healthcare, finance, and legal processing and millions of workers who power the global back office. SoftBank wants to embed AI into these workflows, turning traditional BPOs and KPOs into smarter, leaner, and more scalable tech-driven outfits.

As one insider put it, “They’re evaluating a whole range of small-to-mid-sized firms. SoftBank believes it can inject the capital and capability needed to deploy AI at scale.”

The Stargate Strategy and Crystal Clarity
This India-focused push is only one flank of SoftBank’s broader AI war plan. The company has unveiled a $500 billion moonshot, dubbed Project Stargate, aimed at building AI infrastructure in the United States to support OpenAI. The project includes massive data centers and next-gen AI systems that could redefine enterprise automation.

Simultaneously, SoftBank has teamed up with OpenAI to develop Cristal Intelligence, an enterprise-grade AI platform designed to serve the diverse needs of businesses, from process automation to advanced analytics. The goal: to embed intelligent agents that can learn, adapt, and optimize across industries.

OpenAI will provide the core tech and R&D muscle for Cristal Intelligence through a joint venture called SB OpenAI Japan, while SoftBank will scale deployments using its existing portfolio and workforce.

“Cristal Intelligence is not just a product,” Son said at the Tokyo announcement earlier this year. “It’s a system that adapts to how enterprises live and breathe.”

Crystal Land, Arizona’s Answer to Shenzhen?
The boldest chapter in Son’s vision, however, may play out in Arizona. According to reports, SoftBank is eyeing a $1 trillion AI and robotics zone modeled after Shenzhen; a high-tech manufacturing hub that could become ground zero for artificial general intelligence (AGI), semiconductor fabrication, and smart infrastructure.

Codenamed Crystal Land, the mega-zone would host R&D labs, chip fabrication units, housing for tech talent, and a smart grid backbone – all tailored for the AI age. SoftBank is actively lobbying U.S. state governments and the Department of Commerce for tax incentives to make the plan feasible.

If realized, Crystal Land could be one of the most ambitious bets on AI industrialization since the internet boom.

The Last Bit, Future Isn’t Written in Code but Etched in Silicon

Without doubt, we are standing at the edge of a new digital order – one not just defined by who codes the algorithms, but by who owns the infrastructure that runs them.

The AI age, for all its promise, is already creating a deep and potentially irreversible fault line between the digitally sovereign and the digitally dependent. What oil was to the 20th century, compute power is to this one – a tool of influence, leverage, and control.

From the deserts of Arizona to the data corridors of Singapore, the race is on to secure silicon dominance. And while countries like India, Brazil, and those across Africa scramble to build sovereign AI futures, they’re also racing against a clock set by giants who have a head start, deeper pockets, and a tighter grip on the chips.

This is Digital Colonialism 2.0 and its not about cables and servers but  about control. And in this new world, being plugged in isn’t just about access; it’s about autonomy. It’s about survival. Because the uncomfortable truth is, in the AI age, you either shape the code or you become subject to it.

As Lacina Koné of Smart Africa warned, “It’s not merely a hardware problem. It’s the sovereignty of our digital future.”

In other words, the AI revolution is here. Whether a nation thrives in it or gets digitally colonized depends on what’s inside its data centers and who controls the chips!