What UBS actually said
UBS scores cities on how far house prices have run from local incomes and rents, not on whether a crash is coming. Dubai's score moved to 1.16 from 1.09 and sits fourth of 23 cities, in what UBS calls elevated bubble-risk territory — alongside Miami and Seoul, and below Zurich and Tokyo, the only two cities it puts in the high-risk band.
The supporting points reported alongside it: real house prices have fallen back to mid-2025 levels, real rents are down about 4% over the year, and Dubai still has one of the most attractive price-to-rent ratios in the study at roughly 16 years of rent to buy — which is why UBS's own framing is that high rents keep favouring ownership.
So the index is not saying prices are about to fall. It is saying the gap between prices and fundamentals widened slightly while the market cooled. Those are different claims, and the headline merges them.
Figures as reported by The National and MEED on 25 September 2026 from the UBS Global Real Estate Bubble Index 2026.
Score 1.16, up from 1.09. Fourth of 23 cities. Zurich and Tokyo are the only cities UBS rates high risk.
Sources The National, 24 Sep 2026 · MEED, 25 Sep 2026
The two numbers, month by month
Every row is the middle of all recorded off-plan sales in that month — not an average, and never combined across months. The peak per square foot was April at AED 1,934. August sits 10% below that and 7% below January, while the median price paid fell 39% over the same stretch.
September is excluded: the snapshot was taken on 7 September, so that month is a week of data, not a month.
Sources Dubai Land Department open data
| Month | Recorded sales | Median price | Per sqft |
|---|---|---|---|
Why the two numbers disagree
A median price answers what a buyer spent. A median price per square foot answers what a buyer paid for the same amount of home. When the first falls four times faster than the second, buyers have not been handed a 39% discount - they have moved to smaller and cheaper stock.
That is what the register shows: volume held in the nine to eleven thousand range for most of the year while the ticket size collapsed. The market did not lose a third of its value. Its mix changed.
Anyone quoting the 39% as a price crash, or the 7% as proof nothing happened, is reading one column and ignoring the other.
So has growth stalled?
On the measure that controls for size, yes - and it stalled in spring, not this month. Price per square foot rose into April, dropped 12% between April and May, and has moved sideways since: 1,700, 1,725, 1,740, 1,736. Four months inside a 40-dirham band is a plateau, not a slide.
What changed underneath is what sells. The launches that cleared in the second half were cheaper, smaller units, which is a supply decision by developers as much as a demand signal from buyers.
A plateau is not a bubble bursting. It is also not the double-digit growth the marketing still quotes, and anyone buying off-plan today on the assumption of 2024-style appreciation is pricing in something the register stopped showing five months ago.
What to check before you accept either headline
Ask which measure a claim uses. Price per square foot, median price and average price answer different questions, and the difference between the first two in Dubai this year is 32 percentage points.
Ask what period it covers. A figure measured from January flatters; one measured from the April peak does the opposite.
Ask whether it is asking prices or recorded prices. Portal indices track what sellers ask. The DLD register records what buyers paid, which is the only number that has already happened.
And check the area, not the city. A city-wide median mixes Palm Jumeirah with Dubai South, and the two have nothing to say about each other.
Where the register agrees with UBS, and where it does not
Agrees: the boom stopped. Price per square foot peaked in April and has been flat since, and UBS describing real prices as back at mid-2025 levels is consistent with a market that stopped climbing in spring.
Does not settle: whether that is dangerous. The register shows volume still running at eight to ten thousand recorded off-plan sales a month through the slowdown. A market with a valuation gap and steady volume behaves differently from one where transactions dry up, and the index does not measure transactions.
Missing from both: the mix. UBS tracks a price level; the register shows what actually sold, and what sold got smaller and cheaper through the year. Anyone reading a falling median as falling values has the story backwards.