The Gulf’s AI strategy is easy to misread if you look for it in software. Almost none of the money is going into chatbots.
It is going into land, substations, cooling, fibre and chips. The UAE, Saudi Arabia and Qatar have collectively committed sums in the hundreds of billions of dollars to data centres and computing capacity, and they are doing it as national industrial policy rather than as venture bets. The wager is not that a Gulf lab will out-build OpenAI. It is that whoever owns the compute gets paid regardless of who wins the model race.
Who is building what
In the UAE, G42 and MGX are the vehicles carrying most of the ambition. Abu Dhabi is putting up large AI campuses and next-generation computing facilities, with a planned multi-gigawatt campus that sits at a scale previously reserved for the largest US technology companies. That is a power-grid project as much as a technology one.
Saudi Arabia is running the same play through Vision 2030, funded by the Public Investment Fund. The Kingdom has been expanding cloud capacity, pursuing semiconductor collaborations, and backing hyperscale projects, with partnerships including Nvidia and Google Cloud. The framing from Riyadh has been consistent: the objective is to produce technology, not just buy it.
Qatar has taken a narrower path, building sovereign AI capability through selective international partnerships and investing in digital infrastructure aimed at long-term resilience rather than headline scale.
There is also a software layer, and it is more interesting than the compute story gets credit for. The UAE’s Falcon large language model and Saudi Arabia’s ALLaM project exist because models trained mostly on English internet text handle Arabic badly, and because regional industries have data and workflows that generic systems do not understand. Building locally is partly sovereignty and partly a practical fix.
Why the partnerships keep landing here
Microsoft, OpenAI, Nvidia and Google Cloud have all deepened their presence in the region. Three things explain most of that.
Capital that can wait. Hyperscale data centres and frontier training runs cost tens of billions and pay back over decades. Sovereign funds can underwrite that horizon in a way most private investors structurally cannot, because nobody is asking them for a quarterly result.
Geography. The Gulf sits between Europe, Asia and Africa, with major subsea fibre routes landing nearby. As more AI workloads move toward real-time inference rather than batch training, low latency across three continents stops being a nice line in a pitch deck and becomes an operational advantage.
Power. This is the underrated one. AI data centres are enormous electricity consumers, and the constraint slowing down projects in the US and Europe right now is not chips, it is grid connection. Gulf states have abundant gas alongside fast-growing solar capacity, which means they can offer power at a price and a timeline others cannot promise.
Permits and people
Speed is the other advantage, and it is administrative rather than technical. In many markets, land approval, grid connection and telecoms provisioning happen in sequence, each waiting on the last, and a data centre can lose two years to paperwork. Gulf governments tend to run those processes in parallel, which compresses build timelines significantly.
Talent has followed, drawn by tax treatment, new research institutions and the simple fact that the work is happening here. That produces the loop these strategies are designed around: money attracts engineers, engineers produce results, results justify the next round of money.
Where it could go wrong
Three risks are worth taking seriously.
Chips. Access to advanced AI accelerators is governed by export controls and shifting trade policy, not by willingness to pay. A campus rated in gigawatts is worth considerably less if the licences to fill it with hardware are withheld or delayed. Maintaining compute capacity is a diplomatic exercise as much as a procurement one, and that dependency does not disappear with scale.
Security. Data centres, cloud campuses, transmission infrastructure and subsea cable landing stations are all fixed, valuable and hard to move. Protecting them requires investment on both the cyber and physical sides, and cable landing stations in particular have become an obvious pressure point globally.
Regional stability. Investors and multinationals underwriting thirty-year assets price geopolitical risk into every decision. Sustained tension raises the cost of capital and slows the partnerships these plans depend on.
None of this is unique to the Gulf. Every jurisdiction chasing AI capacity is fighting over chips, power and permits. The difference is that Gulf governments are attempting to solve all of it through coordinated central planning, which either compounds the advantage or concentrates the failure.
The infrastructure-first logic
The argument underneath all of this is that infrastructure gets built before anyone knows exactly what it will be used for. Railways were laid before the trade volumes justified them. Electrical grids were financed decades before most households had anything worth plugging in.
Whether AI follows that pattern is a live question, not a settled one, and there is a real possibility of overbuilding at the top of a cycle. But the Gulf’s position is straightforward: the region generated extraordinary wealth from energy, that wealth is finite, and compute is the closest available substitute for a resource you can sell to everyone.
The countries doing this are not competing to build the cleverest model. They are competing to be the place the models run.
