The New Map of Compute: How the Gulf, India and Africa Are Building AI Infrastructure

Blaze Meridian Group link-share card headlined "The New Map of Compute — How the Gulf, India and Africa Are Building AI Infrastructure", with gold data centre tower icons on a navy background.

Global data centre capacity is set to nearly triple by 2030, growing from 82 gigawatts today to about 220 gigawatts, backed by roughly $1.5 trillion in spending. The US still leads the build-out, with China in second place. But a growing number of other countries are now building their own AI infrastructure too, driven by a mix of data control, speed of access and the desire to diversify their economies.

Getting a data centre built takes more than capital. It requires:

A project needs all five in place before it can actually be built.

The Gulf: cash, energy and land

The UAE and Saudi Arabia treat data centres as national infrastructure. Government-owned entities led the first wave of investment, funded through sovereign wealth, and private capital is expected to take on a larger share as the market matures. Both countries also bring some natural advantages to the table: historically low energy costs, abundant desert land, and a strategic location between Europe, Asia and Africa that gives them low-latency reach into all three markets. That reach lets operators use the Gulf as a base to serve customers well beyond its own borders. They’re backing this up with heavy investment in solar, nuclear and hydrogen power to meet growing demand.

That investment is already turning into real capacity, largely through state-backed entities. Microsoft has committed $15.2 billion to the UAE since 2023, including a $1.5 billion direct stake in G42, and a 200 megawatt data centre expansion delivered through Khazna, G42’s data centre subsidiary and the UAE’s largest data centre platform, due online before the end of 2026. Separately, Stargate UAE, OpenAI‘s partnership with G42, Oracle, Nvidia, Cisco and SoftBank, starts with a 1 gigawatt cluster in Abu Dhabi, with 200 megawatts due online in 2026. In Saudi Arabia, the Public Investment Fund has signed a $10 billion AI hub deal with Google Cloud to be operated through HUMAIN, the kingdom’s state AI vehicle, which has separately signed a $3 billion data centre partnership with Blackstone‘s AirTrunk. Oracle has committed $14 billion over ten years to the Kingdom’s own cloud and AI build-out. Qatar’s Qai signed its own $20 billion venture with Brookfield.

As the region scales up, two things are worth watching: water for cooling in a desert climate, and grid capacity, which will need to keep pace with ambition. Gulf governments are also tightening data rules, part of a broader push to build trust with customers and regulators.

India: demand pulls the market

India’s growth is powered largely by its own economy. Data localisation rules, fast-growing cloud and AI use, and a booming digital sector are all pulling demand higher. KPMG values the opportunity across the full data centre value chain at $90 billion by 2035, and on the ground, the Indian government puts investment already underway at nearly $70 billion, with a further $90 billion in announced projects layered on top.

Capacity is set to grow from between 1.5 and 1.9 gigawatts today to somewhere between 7 and 10 gigawatts by 2030, according to two separate studies: Deloitte (March 18, 2026) and KPMG (July 2026). That growth will require a lot more of everything: Deloitte puts the additional land need at 45 to 50 million square feet, and expects power demand to reach 57 terawatt-hours by 2030.

Efficiency is also a work in progress. Data centres are measured on a ratio called PUE: for every unit of power that goes to actually running the computers, how much extra gets burned on things like cooling and backup systems. A score of 1.0 would mean no waste at all. Deloitte puts India’s average at about 1.9, against 1.3 at the best facilities elsewhere, which means Indian data centres are using close to twice the power their servers actually need.

The reasons are specific to India. Heat and humidity make air conditioning work harder than it would in a cooler country. Power cuts are common enough that operators keep diesel generators and backup systems running just in case, and that itself eats electricity. Most states haven’t set any required efficiency number for operators to hit. And the newer cooling technology that could help, piping liquid straight onto the chips instead of cooling the whole room, is expensive and still hard to source locally, according to research from the Council on Energy, Environment and Water.

Some operators are already doing better than that national average. STT GDC India, one of the country’s largest, says its new facilities now hit 1.5, well below the 1.9 average, meaning less wasted power per unit of computing. The government is pushing operators further still: since 2024, any AI infrastructure it funds has had to hit a PUE below 1.35, a tougher bar than what most of the industry manages today.

Policy is helping too. A new tax holiday running to 2047 gives foreign cloud providers using Indian data centres two decades of certainty to plan around.

Domestic conglomerates are leading the spending, with hyperscalers building alongside them. Reliance has committed roughly $110 billion over seven years to a multi-gigawatt AI campus at Jamnagar, backed by its own 10 gigawatts of renewable capacity in Gujarat and Andhra Pradesh, with over 120 megawatts due live in the second half of 2026. Adani has pledged close to $100 billion to its own AI and data centre push. Google is putting $15 billion into an AI hub in Visakhapatnam through 2030, built with local partners AdaniConnex and Airtel. Amazon is putting $13 billion into expanding AWS data centre capacity in Mumbai and Hyderabad, part of a wider $48 billion commitment across its India businesses through 2030. HCLTech has both invested in and partnered with the government-backed AI startup Sarvam, together with the Odisha state government, on a ₹14,257 crore (about $1.7 billion) facility. Tata‘s TCS is building gigawatt-scale AI data centre capacity under its HyperVault brand, backed by up to $1 billion from TPG.

Africa: small base, real potential

Africa currently holds just 0.6% of global data centre capacity. The US, by comparison, hosts about 45% of the world’s data centre facilities. Even with a lot more under construction across the continent, Africa’s overall share of capacity is expected to stay roughly level, simply because the rest of the world is scaling at a similar pace.

Power remains the main bottleneck. Grids in parts of East and West Africa lose 18–25% of electricity in transmission, roughly three times the global average, and total grid capacity is often too limited to support large-scale data centres.

The upside is real, though. The continent has huge untapped renewable potential in solar, wind, hydro and geothermal power, and 15 countries now have national AI strategies in place. Foreign capital owns much of the capacity built so far: South Africa’s Teraco, the continent’s largest operator, is majority owned by US REIT Digital Realty, which took a controlling stake in a $3.5 billion deal that closed in August 2022 and is now raising its ownership to 77% under a deal announced in June 2026, and pan-African operator Raxio has raised more than $380 million, including $100 million in debt financing from the International Finance Corporation and further backing from Proparco. Development finance institutions are playing a role that domestic capital plays in the Gulf and India.

Summary Comparison

GulfIndiaAfrica
DriverDiversifying away from oil revenueData localisation rules and booming domestic cloud/AI demandSmall base, growing investor interest
Main constraintWater for cooling; grid capacity under strainLand (45–50 million sq ft needed) and power (57 terawatt-hours by 2030)Transmission losses of 18–25%, roughly three times the global average
Efficiency (PUE)Gulf operators are targeting below 1.3Averages about 1.9 versus 1.3 for best-in-class; new builds targeting 1.5 or below, AI tenders below 1.35Averages 1.67 versus a 1.58 global benchmark
Structural advantageCheap energy, desert land, and low-latency reach across Europe, Asia and AfricaBuilt-in demand and competitive costsUntapped renewables: solar, wind, hydro and geothermal
Policy supportData rules tightening to build trustTax holiday to 2047 for foreign cloud providers15 countries now have national AI strategies

Conclusion

Each region is really solving a different version of the same problem: securing enough power, equipment and trust to build at scale. The Gulf’s edge is cash, energy and land. India’s is built-in demand. Africa’s is renewable energy potential still waiting to be tapped. How quickly each one works through its own bottleneck will set the pace of the build-out ahead.

Sources:

Tagged , , , , , , , , ,

Leave a Reply

Your email address will not be published. Required fields are marked *