The numbers are extraordinary. Dell’Oro projects data centre capital expenditure will reach $1.7 trillion by 2030. Meta just broke ground on a $10 billion campus in Indiana. Google secured approximately $32 billion in debt financing for its DC buildout. Quarterly cloud revenues crossed $119 billion. That’s all from a one week period between 5 February 2026 and 11 February 2026! If you squint at the headlines, you would think the only skill required to participate in this market is the ability to pour concrete and plug in servers.

It is not.

The gap between “we are building a data centre” and “institutional capital is flowing into our data centre” is not a construction gap. It is a structuring gap. And in the UAE and broader MENA region - where sovereign capital, cheap energy and available land create a compelling surface-level pitch - that gap is wider than most operators realise.

Here is what I mean.

Look at how the most sophisticated capital is actually being deployed globally. Nscale just closed a $1.4 billion GPU-backed loan - debt secured not against land or buildings, but against the compute hardware itself. CyrusOne partnered with Constellation Energy to anchor a 760MW campus next to dedicated generation capacity at the Freestone Energy Center in Texas, effectively turning power purchase agreements into infrastructure collateral. Ares Management arranged $2.4 billion for Vantage Data Centers. Firmus secured $10 billion in financing from Blackstone and Coatue.

None of these are simple equity cheques. Each one represents a specific structuring decision about which layer of the data centre stack the capital is being deployed against, what the risk profile of that layer looks like and what kind of investor the structure is designed to attract.

This is the part most MENA operators are not writing about, because most are not thinking about it with sufficient precision - and most regional investors are not yet demanding it.

A data centre is not a single asset. It is at least three distinct investment layers stacked on top of each other: infrastructure (the shell, the land, the cooling systems), compute (the GPUs, the servers - depreciating equipment with a three-to-five year useful life) and energy (the power purchase agreements, the grid connections, the generation capacity). Infrastructure bankers already call these layers PropCo, ComputeCo and PowerCo - and the most significant deals of the past twelve months have been structured accordingly. Each layer has a fundamentally different risk profile, different return characteristics and attracts a different type of capital.

Infrastructure is a real asset play. Pension funds understand it. Insurance companies understand it. It looks like a warehouse with better economics. Compute is equipment finance - closer to aircraft leasing than to real estate, with technology obsolescence risk baked in. Energy sits somewhere between utility-scale project finance and regulated infrastructure, depending on how the off-take is structured.

When an institutional allocator looks at a data centre investment, the first question is not “how many megawatts?” It is “which layer am I buying and what is the structural protection around my capital?” A 100MW campus with a clearly separated infrastructure holdco, a compute financing vehicle and a PPA-backed energy structure is investable across three different pools of institutional capital simultaneously. The same 100MW campus structured as a single opco with everything commingled looks like a venture bet to anyone outside the founding investor’s network.

I saw this play out recently with a deal that crossed my desk - a 20MW facility in Quebec with potential access to 200MW of power. The pitch was framed as an opportunity to invest in GPUs. We passed on the GPU element immediately. But the infrastructure layer - the physical facility, the power access, the jurisdictional advantages of Quebec’s energy market - that was genuinely interesting as a joint venture. Same asset. Completely different conversation depending on how the investment was structured.

This is where the Gulf has both its greatest advantage and its most significant blind spot.

The advantage is obvious. The UAE, Saudi Arabia and other Gulf states have three things that most data centre markets would kill for: abundant energy, available land and sovereign capital with long time horizons. Microsoft just confirmed a Saudi Arabia data centre region for Q4 2026. G42 is leading a $1 billion data centre project in Vietnam, deploying Gulf capital into Asian infrastructure. The regional ambition is real and it is backed by serious money.

But ambition and capital are necessary conditions, not sufficient ones. The blind spot is the structuring layer between the sovereign commitment and the institutional co-investment that would allow these projects to scale beyond what sovereign balance sheets alone can support.

Consider what is happening elsewhere. In New York, proposed moratoriums on data centre construction are being driven by energy consumption concerns. In Spain, Merlin Properties abandoned a planned data centre campus in Andalusia because guaranteed power supply could not be secured. Across Europe, grid constraints are becoming the primary bottleneck for new capacity.

For the Gulf, these constraints elsewhere are a structural tailwind. But only if the investment vehicles are built in a way that institutional allocators outside the sovereign wealth ecosystem can actually deploy into. That means governance frameworks that meet OECD standards. It means audited SPV structures with clear waterfall provisions. It means separating the infrastructure, compute and energy layers so that each can be financed by the type of capital best suited to its risk profile. And for the compute layer specifically, it means accounting for U.S. export control frameworks that constrain advanced AI chip supply to the region - a risk that any serious institutional investor will price in before committing capital.

The operators who are getting this right globally are not the ones with the biggest campuses. They are the ones whose financing structures look familiar to an institutional allocator’s investment committee. Equinix is forecasting 10-11% revenue growth not solely because of its operational footprint and interconnection density, but because its financial architecture - REITs, predictable revenue models, transparent reporting - fits cleanly into how institutional portfolios are constructed.

Meanwhile, some of the most interesting structural innovation is happening at the margins. Anthropic recently agreed to cover the electricity costs of grid upgrades needed for its data centre buildout - essentially subsidising public infrastructure to unlock private capacity. Google is backing TeraWulf’s pivot from Bitcoin mining to AI data centres, repurposing stranded energy assets. NVIDIA invested $2 billion directly in CoreWeave, signalling that even hardware manufacturers now see the compute layer as a distinct asset class worth backing separately. These are not just operational decisions. They are structuring decisions about how to make data centre investments attractive to specific types of capital.

The MENA region has every ingredient to become a global data centre hub. Power. Land. Capital. Strategic geography between European and Asian markets. What it does not yet have, in most cases, is the structuring sophistication to make these assets investable beyond the sovereign and family office capital that has funded the first wave. 

There is also a compounding cost to late entry that goes beyond market pricing. At every layer, the operators who structure the vehicle capture a premium - the PropCo margin, the ComputeCo financing spread, the PowerCo off-take terms. Investors who arrive after the structure is built pay those premiums cumulatively. I intend to explore this in a future piece, but the short version is straightforward: in this market, the structurer sets the terms and the late capital pays the freight.

The next phase of value creation in AI data centres will not be won by whoever builds the biggest facility. It will be won by whoever structures the most investable one.

Arash Saidi is General Partner and Chief Legal Officer at Storm Group and Cypher Capital.