THE CORRIDOR · Infrastructure
The UAE spent two years winning access to American AI chips. Now it is renting them out.
For a Gulf bank, the question was never whether sovereign AI was a good idea. It was what it costs, and who has to file the export paperwork. There is finally an answer to both.
Chandni Melwani
Founder & Editor
Jul 22, 2026 · 3 MIN READ
The News
On July 20, 2026, e& UAE and Core42 announced a partnership to sell Sovereign AI Compute: paid access to AI chips housed in UAE data centres, so companies can build and train AI systems without their data ever leaving the country. Core42 supplies the computing power and implementation services; e& UAE supplies national-scale connectivity, local customer relationships and professional services. The companies advertise no capital expenditure requirement and zero egress fees. It is the first joint offering under what the two describe as a broader global partnership.
Know More
- — Who they are: Core42 is an Emirati AI infrastructure company, part of the G42 group. e& UAE is the country's largest telecom operator, formerly Etisalat.
- — Bought at the top, rented at the bottom: American companies sell the chips outright and the buyer owns them. Core42 owns the hardware here and rents access to it. Nobody is renting anything from Washington.
- — What changed in July: a US rule effective July 10 moved the UAE into Country Group A:5 and allows advanced AI chips to be bought licence-free — but only by a named list of approved entities. That list is UAE government bodies, the UAE arms of eight American technology companies, and two Emirati firms: G42 and Core42. Every other UAE buyer still needs a licence.
- — There is an expiry on it. The rule states that the authorisation for G42 and Core42 'shall automatically expire on April 6, 2027' — and that if the two do not become US companies by then, they must apply for authorisation like anyone else.
- — The G42 connection: G42 is the group that removed Chinese equipment and accepted US security conditions during the vetting that preceded that decision.
- — Also launched: GPU Connect, which sells the chips together with the high-speed network connection needed to reach them, in one package rather than two contracts.
- — The wider partnership covers keeping the country's AI services running through outages, cyberattacks or other disruption — what the companies call national digital continuity.
- — Not published: the price, how much total capacity exists, which generation of chips is available, or the name of a single customer.
A bank in Dubai that wants to train an AI model on its own customer records has, until recently, had three options and no good one. Put the work on a global cloud, and explain to a regulator why customer data now sits on infrastructure run by an American company. Buy chips and build somewhere to keep them, which meant a capital project and a US export licence, granted case by case, for hardware Washington had spent two years deciding whether to let into the country at all. Or shelve it.
On July 20, e& UAE and Core42 put a fourth option on the market: the chips that AI runs on, housed in the UAE, rented rather than bought. Core42 owns the hardware and sells access to it. The licence was somebody else’s problem, the data never leaves the country, and there is no capital outlay to begin.
Two of the terms are worth slowing down for. The companies promise no capital expenditure requirement, so the spend starts small and scales with use instead of arriving as a multimillion-dollar purchase order. And they advertise zero egress fees, the charge a cloud provider normally applies when you move your data off its platform, priced per gigabyte, trivial at first and painful once you have years of it. Egress fees are a large part of why companies stay put even when something better appears, and waiving them is a provider betting it can win on merit rather than on the cost of leaving.
Set that against the last two years of sovereign-AI news, which was almost entirely governments arguing with each other: export controls, security undertakings, the vetting the UAE submitted to before Washington eased the licence rule this month. That easing is narrower than the headlines made it sound: licence-free buying is confined to a named list of approved entities, which includes G42 and Core42 and does not include a bank in Dubai. All of that answered one question: whether the country could have the hardware. It never answered the question a company acts on, which is whether it can buy the capability this quarter and what it will cost.
It helps to be concrete about what the newest chips buy that older ones did not. Ordinary machine learning on private data has been possible in the Gulf for years, on hardware nobody needed a licence for: a fraud model over claims history, a classifier on scans. What advanced chips add is the ability to run and adapt frontier-scale models, the general-purpose systems a Gulf institution could previously reach only by calling an American API and sending its data along with the request. Renting this capacity is what puts the model and the data in the same building.
Which is why the scarce thing is about to change. Getting hold of chips was the hard part, and it rewarded whoever could work export policy. Now that anyone can rent them, the advantage moves to whoever knows which of those jobs is worth doing at all, and that is a question about how a Gulf bank or hospital actually runs rather than about hardware.
Related
Frequently Asked Questions
What does "sovereign AI compute" actually mean?
The computers that run AI sit physically inside the country and are operated by a company under that country's laws, so data never crosses a border. The reason it matters is legal rather than technical. Banks, hospitals and government departments frequently cannot put customer records on infrastructure a foreign government could compel access to, which has ruled out most global cloud providers for their most sensitive work.
Is the US selling these chips to the UAE, or renting them?
Sold, and the chips are owned outright by the buyer. What the US controls is permission to export them. Advanced AI chips could only reach the UAE under an American export licence, granted case by case and carrying volume caps and security conditions; a rule effective July 10, 2026 lets a short list of approved entities buy licence-free, and Core42 is on that list. Core42 buys and owns the hardware, then rents access to it. The renting happens between an Emirati company and its customers, not between Washington and Abu Dhabi.
If there is no upfront cost, what do customers actually pay?
Neither company published rates. In this market the usual model is pay-as-you-go by the hour for each chip you use, with cheaper rates if you commit to a fixed block of capacity for a year or more. The alternative it displaces is buying hardware outright, which is a capital purchase running to millions before a single model is trained. What the companies promise is "no capital expenditure requirement" — you are renting rather than buying, so the spend starts small and scales with use. That is a claim about capital cost, not a promise that nothing is payable up front.
Why does "no egress fee" matter?
An egress fee is what a cloud provider charges to move your data off its platform. It is small per gigabyte and very large once you have accumulated years of it, and it is one of the main reasons companies stay where they are even when a better option appears. Charging nothing removes the financial penalty for leaving.
Sources
- Core42, "e& UAE and Core42 partner to deliver sovereign AI infrastructure at scale"
- e&, "e& UAE and Core42 partner to deliver sovereign AI infrastructure" (July 20, 2026)
- Compare the Cloud, "Core42 and e& launch UAE's first sovereign GPU compute platform for enterprise AI"
- Data Centre Magazine, "Inside e& UAE and Core42's Sovereign AI Cloud Strategy"
- Bureau of Industry and Security, "Enhanced Favorable Treatment for the United Arab Emirates Under the Export Administration Regulations," 91 FR 43034 (effective July 10, 2026)
Chandni Melwani
Chandni Melwani is the founder and editor of New in AI, covering AI agents, M&A, and enterprise adoption. She holds a Master's in Management of Artificial Intelligence from Queen's University and brings a practitioner's perspective from her work in Data and AI leadership.
Follow ↗The Corridor Briefing
Get the two-beat briefing before the market opens.
One email. Corridor capital moves and Commerce platform shifts, decoded daily.
No spam. Unsubscribe anytime.
More from The Corridor
Infrastructure
Nvidia is helping pay for the data centre that will buy its chips
A Korean internet company, an American chipmaker and a Canadian asset manager are building Korea's national AI capacity together. Look at who is funding whom.
Chandni Melwani · East Asia
Policy
The White House says Moonshot AI trained its model on Anthropic's, and accessed restricted Nvidia chips in Thailand
Washington built export controls to keep advanced chips out of Chinese hands. The allegation is that a Chinese lab reached them in Thailand instead, and that the model capability never needed to cross a border at all.
Chandni Melwani · China
Infrastructure
Microsoft is paying to put Azure's European customers on compute it does not own
Europe is trying to build its own AI sovereignty and rent it back. A US hyperscaler is underwriting the capex, and structuring the cheque to stay clear of Brussels.
Chandni Melwani · Europe