The Other Funded City

Every founder carries a private map of where money lives. San Francisco, Boston, London, Berlin, Singapore. The map is not wrong. It is just older than the money. Dubai has been running a public programme layer for years, and it is documented the way every funding regime is documented: host, structure, amount, date. The Dubai AI Campus inside the DIFC Innovation Hub is the largest dedicated AI cluster in the region. Phase one holds more than 75 businesses. The published target is more than 500 by 2028. Tenants get innovation and AI licences, coworking and offices, and access to an NVIDIA DGX supercomputer. Applications run all year. That is not a pitch. That is a building with a capacity figure and a portal. I have built companies across twelve countries, and the same mistake repeats in each one. People read a headline about a city, decide it is not their industry, and never open the actual programme list. The list is public. The list has numbers on it. The capital moved. The map did not.

The programme list behind this piece was compiled on 21 July 2026 from the operator portals and press releases. That date matters more than any figure in it. Read the strongest line in the document. Dubai Future Foundation runs an accelerator for AI in government services, alongside more than 20 government entities. Eight weeks, on site, no equity, accommodation and travel sponsored. On terms it is the best entry in the file. The second cohort closed in August 2025. The third had not been announced when the tracker was compiled, eleven months later. So the honest reading is this. The programme is real, the terms are published, and the door has been shut since August 2025 with no date on the next one. I have restructured a EUR 75 million industrial group, and every capital decision I have been inside of came down to reading the dates before the terms. A standing offer with no closing date is marketing. A closed intake with a published reopening is a calendar entry. The difference between those two things is the difference between a document and a brochure.

Most founder money is priced in ownership. One round, 15 per cent of the company, permanently gone, and every later win is taxed by the people who arrived before the win. The programme layer prices differently. The DIFC cluster takes no equity. Neither does the Dubai Future Foundation accelerator, whose entry in the tracker lists equity as none. Dtec runs a twelve-month hub with coworking and free zone setup and takes no share. in5, run by TECOM Group out of Dubai Internet City, incubates across technology, media and design on rolling applications. What these arrangements actually sell is speed and standing. A licence in a recognised free zone, an address that banks accept, a working space, and a corridor full of other people solving the same problems in the same quarter. None of it is free. Rent is rent, administration has a price, and the licence renews on a schedule. The trade is simply not paid in equity, which means the upside stays where it was earned. The probability that a founder can name three programme routes their company could enter today without giving up a share is low. That is not a shortage of options. It is a shortage of reading.

In February 2026 Dubai Silicon Oasis Authority announced a second AI district. The published numbers are unusually specific for this kind of announcement. AED 5 billion of investment across 2 million square feet. More than 100 AI companies targeted. Five research labs, with partnerships named as MIT, Stanford, MBZUAI and Carnegie Mellon. 50,000 professionals trained by 2030. Then the incentive list. Zero per cent corporate tax for 50 years. A 50 per cent rent reduction for three years. AED 100,000 of GPU compute credits for start-ups. Research and development grants up to AED 5 million. Licensing in 48 hours. Phase one occupancy is stated for the fourth quarter of 2026, and the incubator holds 20 places in its first cohort. Read that as a price list rather than a gift and it becomes useful. Zero per cent corporate tax for 50 years is a statement about what the state wants to attract. A research grant ceiling is a statement about which activity it wants visible on its soil. The 48-hour licence is a statement about who it is competing with. A government publishes exactly what it is willing to pay for. That document is the map.

A tracker is a photograph, not a promise.

The most interesting entry in the DSOA package is not the tax rate. It is the GPU credits. A start-up used to be rationed by salary, then by cloud billing. The subsidy layer has noticed. Access to an NVIDIA DGX supercomputer is now a headline benefit of the DIFC cluster, and 100,000 dirhams of compute credits sits beside the rent discount in the Silicon Oasis list. Compute has moved from a variable cost into an instrument of industrial policy. Understand what that means. The state is buying capacity on behalf of companies that cannot buy it themselves, and in doing so it is choosing which sectors get to train models at all. Credits are a meter. They are consumed at a rate set by someone else, and the rate is not inside the credit. A grant that pays salaries is a floor under a business. A grant that pays compute is a door into one. Both expire. Neither arrives without an application.

Sort the whole tracker by one field and it becomes a different document. Open all year: the DIFC cluster, in5, Dtec. Rolling: TURN8, Intelak and the development partnerships run through donor agencies, which accept applications continuously. Periodic cycles: Dubai Future Accelerators, running nine-week pairings with government entities. Closed with no successor announced: the government services accelerator, whose second cohort shut in August 2025. That sort is the whole analysis. Anything with a closing date is operable. Anything described as ongoing needs a second question, which is what the current intake looks like and how many places it holds. Anything closed is history until a new date appears. Founders lose years to this. I have watched a team spend four months positioning itself for a programme whose previous intake had closed before they started writing. The work was not wasted because the programme was bad. It was wasted because nobody checked the date. The cheapest diligence in the world is reading when an application window opens and closes. It costs one visit to the operator page. The tracker is a photograph of that page. The page is the truth.

The point is not that one city is better. It is that the programme layer is now global, and the map most founders carry is regional and about ten years out of date. The mechanism is identical everywhere. A state publishes what it will pay for, what it will not take in return, and the date the door closes. The DIFC cluster takes no equity. The Silicon Oasis district publishes a 50-year tax position and a 48-hour licence. The government services accelerator sponsors travel and takes no share, and its door has been shut since August 2025. Run the same pass over Berlin, Madrid or Brussels and the structure does not change. Envelope, condition, deadline. The names change. The arithmetic does not. So the useful habit is not enthusiasm about any one destination. It is a monthly hour on the raw programme pages, with a spreadsheet open, filtered on dates that are still live. The capital moved. Nobody updated the map. The map is still the thing that decides who gets funded. I took a company through a Frankfurt listing, and the money that took the longest was never the money with the best terms. It was the money nobody had gone and asked for. What this piece describes are published programme facts as compiled on 21 July 2026 from the operator portals and press releases. It is not advice, terms change without notice, and the only authority on any programme is the operator current page.