AI companies in Dubai: who does what, and how to choose one

A map of the market from a company that is on it

Dubai counts more than 800 AI companies. Very few will run a model on your data, inside the country, and sign for what it does.

A Dubai Centre for Artificial Intelligence study counted more than 800 AI companies in the city in late 2024; 72% of them were SMEs or start-ups and about half described themselves as AI software developers. By April 2025 more than 325 firms had applied for the Dubai AI Seal. So the shortage is not of companies. It is of companies that can take an AI project past the demo: onto your own data, inside the UAE, with someone accountable when the model is wrong.

This page maps the AI companies in Dubai and the wider UAE by what they actually do, explains how the market is put together (the government layer matters more here than anywhere else we work), gives you five questions that sort the field, and says where 10ⁿ Tech sits. We have written it the way we would want to read it if we were buying.

Five layers, not one market

“AI company” covers a national champion with a gigawatt of compute and a four-person chatbot shop. They do not compete with each other, and a buyer who confuses the layers ends up asking a hyperscaler for a pilot or a dev shop for a sovereign platform. Here is the map.

Most enterprise and government buyers in Dubai need the fourth layer and shop in the fifth, because the fifth advertises. The first three are worth knowing because your project will sit on top of them: the compute you rent, the model you host, the sovereign cloud your data must stay in.

How the UAE AI market works

Government sets the pace. Dubai’s Universal Blueprint for AI came in April 2024; 22 Chief AI Officers were appointed across government entities in June 2024 with a mandate to ship AI plans, not study them; the Dubai AI Seal grades providers from E to S on workforce, product maturity, local R&D and partnerships; DIFC opened an AI Campus in May 2024 (75 firms in phase one, 500 targeted by 2028) and in April 2026 declared itself an AI-native financial centre with a USD 3.5 bn contribution target and 25,000 jobs. Federal tenders for agentic AI platforms are now routine. If you sell to government, these are your buyers’ KPIs.

Residency is decided lower down than people think. Health data cannot leave the country under Federal Law No. 2 of 2019; personal data is governed by the PDPL (Federal Decree-Law 45 of 2021) and, in the free zones, by DIFC and ADGM regulations; Abu Dhabi has its own AI council since January 2024. On Azure, “UAE North” on the resource does not keep inference in the UAE; the deployment type does, and the in-region model list is short. We wrote that up in data residency on Azure UAE North depends on the deployment type.

Compute is the long pole. A GPU server is quoted at 10 to 16 weeks from order and a rack built for 6 kW will not take a 15 kW node. Cloud GPU capacity in the region is reserved, not guaranteed. Programmes that keep their dates split the work by data class: public content on cloud now, confidential workloads on the sovereign estate when it lands. Our cloud first, sovereign second piece walks through one.

The problems start when the project is real. A chatbot on public FAQs is a week’s work for anyone on this page. The work gets hard, and the field gets thin, when the agent has to read a citizen’s file, book a venue, submit a grant, or write to a system of record; when the model must run in-country on a data class the law names; when four vendors’ products have to behave as one; when the answer has to be signed by a human before it leaves. That is the fourth layer’s job, and it is where the price of a wrong choice is paid in a project that stays a pilot.

How to choose an AI company in Dubai

Five questions, asked in the first meeting. The answers place a firm on the map faster than its website does.

• Which of my data does the model see, and where does inference run? Expect a classification, not a reassurance. “In the UAE” should come with a deployment type, a region and a named fallback.
• Who signs when the agent is wrong? If the answer is “the model is 95% accurate”, the firm has not built for production. The right answer names an approver, a log and a way to reverse the action.
• Which of your people has shipped this to production, and will they be on my project? Not the company. The person.
• What do I own at the end? Agent definitions, prompts, evaluation sets and infrastructure code in your repository from week one, or a subscription to theirs.
• What is the smallest first release with a number on it? A public beta in six weeks with a routing-accuracy target is a plan. A twelve-month programme with a demo at month three is a hope.

AI companies in Dubai and the UAE worth knowing

Grouped by layer, with what each firm says about itself in September 2026 and what we have seen. None paid to be here. 10ⁿ Tech is listed on the same terms.

National infrastructure and models

• G42 and Core42 (Abu Dhabi). A group of ten companies building what it calls the Intelligence Grid across 30 countries. Core42 is the sovereign cloud and compute arm: NVIDIA, AMD, Cerebras and Qualcomm accelerators, a Sovereign Public Cloud built on Azure with UAE controls, and the Compass model platform. If your programme needs sovereign compute at national scale, this is the counterparty.
• Stargate UAE. A 1 GW AI campus in Abu Dhabi announced in May 2025 by OpenAI, G42, Oracle, NVIDIA, Cisco and SoftBank, with the first 200 MW due online in 2026. Not something you buy directly yet; it changes what will be rentable in-country.
• TII and MBZUAI. The Technology Innovation Institute publishes the Falcon models; MBZUAI’s Institute of Foundation Models released K2 as fully open weights. Both matter to a buyer who wants an Arabic-capable model they can host without an external API.
• AI71. Falcon commercialised: products for government (AskGOV), health (customers include the Ministry of Health and Prevention and Emirates Health Services) and agriculture, with advisory around them.

Applied national champions

• Presight (Abu Dhabi, ADX-listed). Big-data and AI analytics for public safety, cities and enterprise: AED 3.03 bn revenue in FY25, 20-plus flagship products, projects on five continents. The default bidder on large national analytics programmes.
• AIQ (Abu Dhabi). ADNOC and G42’s joint venture for machine learning and computer vision in energy. If you are an oil and gas operator, they are already in your building.
• Space42. Satellite communications and geospatial intelligence with AI on top; Thuraya and the Foresight constellation. The geospatial layer under smart-city work.
• e& enterprise. The telecom’s enterprise arm: AI in a box, SLM in a box, sovereign inference, an AI governance suite and video AI as a service, sold to large organisations with the connectivity and cloud around it.

Integrators and boutiques

• Aleria (Abu Dhabi). Builds and operates sovereign AI systems, including air-gapped deployments; names FAB, ADIA, Masdar, Aldar and the Abu Dhabi Cyber Security Council among its customers and backers. The strongest pure sovereign-AI boutique in the capital.
• TierOne (Dubai). GPU and HPC systems: NVIDIA B200 and B300 servers, Supermicro HGX, Dell and ASUS nodes, NetApp AIPod. The hardware half of an on-premises AI build.
• DataBorn (Dubai Internet City). 350-plus developers and consultants, twenty years in banking and telecom analytics: fraud, AML, churn, predictive maintenance, process mining. The delivery bench behind large agent programmes.
• Lumitech (UAE, engineering in Eastern Europe). Custom AI software, RAG and agents; 100-plus engineers, WhatsApp commerce and identity-verification builds for GCC clients.
• viAct and ForgeSentry. Video AI for worker safety: PPE detection, danger zones, fall protection. viAct is Hong Kong-based with UAE construction customers such as ALEC; ForgeSentry is Dubai-native.

Development shops

Master of Code, Code Brew Labs, Apptunix, Aleddo, TechGropse and a few hundred others. Good at a chatbot, an app with an AI feature, a proof of concept on a public API. Ask the five questions above before you give one of them a system of record.

Where 10ⁿ Tech sits

Fourth layer, Dubai, established 2016, around thirty people, a Dubai AI Seal holder since October 2026, with the infrastructure half and the software half under one roof. We design and build agentic AI platforms for government and enterprise on Azure UAE North or on the customer’s own GPUs, with the governance that makes an agent’s output usable: approval inboxes, hash-chained action logs, an agent registry, ISO/IEC 42001 and NIST AI RMF as the reference frames. Our work on sovereign estates runs to the rack: sixteen H200s on two 32 kW feeds, six environments on two nodes, disaster recovery decided before the hardware ships.

We have written the thinking down rather than claim it: AI outputs are advisory until someone signs, sixteen GPUs and two racks, a bilingual AI avatar that never leaves the UAE, digitising a library so an assistant can search it. Digital twins and video analytics we have delivered for Dubai Municipality and Abu Dhabi Housing Authority; the agentic platforms are under evaluation with federal entities, so we describe them and do not name them.

If your project is in the hard part of the map, contact us with the data classes involved and the date you have promised. We reply within a working day, and the first conversation is about whether the date is achievable, not about a demo.

Company descriptions are drawn from each firm’s public website and from Dubai Future Foundation, DIFC, Reuters and OpenAI announcements, September 2026. Photo credits: Robert Bock, Vishnu Mohanan, Fatemeh Rezvani and Winston Chen on Unsplash.