Warning on the Gulf’s bet on data centres, Mo Joodah calls for smaller local agents
When the region talks about artificial intelligence, the size of the infrastructure can easily appear as sufficient proof of seriousness. But Mo Joodah, the former commercial lead at Google’s X lab and a business and government adviser, asks a different question in an interview with AGBI: does the large spending on computing turn into returns that justify its cost, or is it a bet that precedes real demand?
The interview does not deliver a definitive judgment on Gulf investments, nor does it predict what will happen with certainty. It is a warning from Joodah himself, at a time when AGBI says Gulf entities are pouring billions of dollars into artificial intelligence. He believes that costly, energy-intensive data centres may not recoup what has been spent on them, and that the race by giant firms to increase capacity may not translate into revenue to the extent that would justify the investment.
Size is not a usage planJoodah bases his view on a simple idea: building massive computing capacity does not by itself answer the usage question. He compares this to what he calls an American approach that assumes bigger is better, while urging Gulf investors to consider smaller, more efficient models and applications designed for local markets. In his view, value may lie in software that employs models and in smart agents designed to perform specific tasks, rather than assuming every benefit requires the largest available infrastructure.
That does not mean infrastructure is absent from the region’s plans. The interview notes that the UAE’s “StarGate” project, valued at $30 billion, is designed to provide up to 5 gigawatts of AI data-centre capacity, compared with 358 megawatts of the country’s total capacity in 2025 according to the source. It also notes that Gulf investors have committed billions of dollars to companies and funds in the field, and that the UAE is investing heavily in infrastructure.
The gap the warning focuses onJoodah does not deny the need for computing in every case. He links massive capacity to inference and serving millions of users, but he says the region does not have the same usage scenario at that scale. Therefore he proposes a smaller, more efficient architecture, with agents that understand regional markets. This is not a complete technical recipe in the interview, but an investment direction the speaker prefers, and its success remains contingent on local applications turning into actual use.
The interview places the warning within a broader competition. According to Joodah, open-weight, low-cost Chinese models could steer users toward cheaper alternatives if the global economy slows and companies cut spending. He believes this possibility calls on Gulf states to diversify their options, rather than tying their future to a single technological path. The interview does not demonstrate that such a shift will occur, but it conveys his assessment of the risks.
What matters to the worker and decision-maker in the regionThe practical point is not choosing a technological camp, but testing what the computing capacity will do within a specific market. In the UAE, for example, the interview says the goal is to shift half of the federal government’s sectors, services and operations to independent AI agents within two years. This makes the question of local application as important as the question of capacity: which services will use the agent, what task will it perform, and what cost can be justified?
AGBI also outlines a network of cross-investments among chip, model and cloud companies, and cites Joodah’s warning that this interconnection may inflate revenues and valuations. Therefore, dealing with his article does not require accepting his forecast of a bubble burst next year or the year after. What is needed is to separate opinion from evidence: the large investment is a fact mentioned in the interview, while its outcome remains contested. For the Gulf, this provides a tougher benchmark than announcing capacities, it is the ability to link every computing bet to a local need and a measurable return.