GUIDES & INSIGHTS · LAND X 001

Two ways to buy the land. Two different markups on it.

A developer gets residential land in Singapore two ways: it wins a state tender, or it buys a whole estate from its owners. We have launch prices for sixty-one projects across both routes, and the markup applied to the land is not the same. State-tendered land comes out at 2.23× what was paid for it. Freehold land bought en bloc comes out at 1.96×. That difference is real, it is smaller than it first looks, and the reason it exists is not the one most people reach for.

Farhan Adenan · 3 Sep 2026 · 12 min read

Multiples are launch average PSF divided by land PSF per plot ratio, from the POV land-bid tracker as at 7 Aug 2026, rebuilt 3 Sep 2026. Land figures are inclusive of land betterment charges and any lease upgrading premium where the source states it. Every figure is arithmetic on those inputs and is shown so you can reproduce it. Projections are indicative, not forecasts.

Almost every new condominium in Singapore sits on land a developer acquired in one of two ways, and the difference between them is more than administrative.

A Government Land Sales tender is a sealed-bid auction. The state puts up a 99-year leasehold parcel, developers submit one number each, and the highest wins. Nobody sees anybody else's bid. The winner finds out how much they overpaid only when the results are published, and the gap between first and second is routinely double digits.

A collective sale is the opposite in almost every respect. An existing estate's owners agree among themselves to sell, set a reserve price, and go to market. The developer negotiates against a published floor, with time to underwrite, and usually against a small field. Where the estate is freehold, the land it buys is freehold too.

Those are different games, and there is no particular reason to expect a developer to apply the same markup to land won in a blind auction as to land bought across a table. So we checked.

Nine is not many, and an average of nine is easy to hide behind. Every record is below, so you can see which ones are carrying the result and decide for yourself whether you believe it.

Exhibit 2. The complete freehold collective-sale set. Three are marked †: The Hillshore, Ardor Residence and Terra Hill are boutique projects with no publicly reported launch-weekend average, so the figure is the tracker's own and we could not corroborate it against a named report. Note which way that cuts — all three sit at the top of the range (2.04×, 2.05×, 2.21×). Drop them and the en-bloc median falls to 1.95×, widening the gap rather than closing it. The finding is conservative as it stands. Source: POV land-bid tracker; launch prices cross-checked against EdgeProp, 99.co and PropertyReview reports where available.

Here is the correction, and it is the part that separates an observation from a finding.

Collective sales are not spread evenly across Singapore. Seven of the nine in our set are city-fringe — the Continuum, Terra Hill, Meyer Blue, Arina East, the Arcady, Ardor Residence, the Hillshore. One is core central, one suburban. That is not an accident: an estate worth assembling is an older, larger, well-located block, and those cluster on the fringe.

Now note where state-tendered multiples run highest. They are lowest on the city fringe — a median of 2.05× there against 2.24× suburban and 2.30× core central. Land is cheapest in the suburbs and construction cost is a fixed floor underneath it, so the multiple has to stretch furthest there to clear the same margin.

So the comparison in Exhibit 1 puts a group that is seven-ninths city-fringe against a group spread across all three regions, and then reports the difference as though tenure caused it. It did not. Most of it is a map.

Two-thirds of the difference was a map, not a tenure.

We would rather publish the 4.8% than the 14%, because the 4.8% is the one that survives the obvious objection. It is also small enough that we would not build a decision on it alone.

The robust difference is not the average at all. It is the range.

State-tender multiples run from 1.68× to 2.99×. En-bloc multiples run from 1.84× to 2.25×. Measured properly, the tender spread is a little over twice as wide, and on the city fringe, where the samples are most comparable, it is two and a half times as wide.

The mechanism is not mysterious once stated. A sealed-bid tender rewards the most optimistic bidder in the room. Whoever holds the highest view of the site wins it by construction, and that winner then has to price a launch off a number nobody else was willing to pay. Sometimes that is vindicated. Sometimes it produces a 2.99×.

A collective sale has a floor and a conversation. The reserve price is public, the developer underwrites the site before committing, and the freehold product it will eventually sell has a price ceiling it can observe in the resale market next door. The result clusters.

For a buyer, that is the practically useful sentence in this article: launch pricing on freehold en-bloc land is more predictable, not cheaper. If you are trying to guess what an unlaunched project will ask, you will be closer more often on an en-bloc site — and if the site came from a state tender, you need the band, not the midpoint.

It is tempting to read a 1.96× against a 2.23× as the freehold buyer getting a better deal. Run the arithmetic and it disappears.

On the city fringe, where the comparison is fair, the land itself costs more when it is bought en bloc: a median of $1,318 psf per plot ratio against $1,235 for state-tendered land, about 7% dearer. Apply each route's own multiple and the launch prices land at a median of $2,692 psf for en-bloc freehold stock against $2,487 for tendered leasehold — roughly 8% more.

So the freehold premium at the launch counter is about the same size as the freehold premium in the land. The thinner multiple does not hand the buyer a discount; it very nearly offsets the dearer land, and the two together leave you paying more for the freehold than for the leasehold — which is what you would expect, and is not what a 1.96× against a 2.23× suggests on first reading.

None of which is an argument against freehold. It is an argument against treating the multiple as a measure of value. It measures what the developer added, not what you got.

The nine above are all freehold, because that is what the collective-sale market produced in 2023 to 2025. The current cycle is producing something else, and it fits neither bucket.

Thomson View sold collectively for $810 million in November 2024 at $1,178 psf ppr — a figure that already includes the land betterment charge and the lease upgrading premium — and comes back to market on a fresh 99-year lease as Thomson Reserve, 1,268 units, guided for the fourth quarter of this year. Loyang Valley followed in April 2026: $880 million, $940 psf ppr inclusive of a $226 million betterment charge and a $246 million lease upgrading premium, around 1,249 units to come.

Both were bought the en-bloc way. Both will sell a leasehold product. The freehold multiple was fitted on freehold sites and the tender multiple on land won at state tender, and there is no honest basis for asserting which of the two a leasehold collective sale should follow.

We have exactly one launched example — Chuan Park, bought collectively in 2022 on a 99-year lease running from 1980, which launched at 2.05× its land. One observation is an anecdote. So rather than pick a multiple and present the result with a band around it, we have given this category its own bucket and published no projection for anything in it, Thomson Reserve included. When there are enough of them to fit a multiple, there will be one.

That is a deliberate subtraction. Until last week our own model ran Thomson Reserve through the city-fringe tender multiple and printed an implied launch price. It was a plausible-looking number built on land bought a different way, and a plausible-looking number is worse than no number, because you cannot tell by looking that it is unsupported.

Everything below sits on state-tendered land, so the tender multiple is the right one to apply. The band is not decoration: it comes from a leave-one-out backtest across the launched set, where the typical miss is 8.9% and the band catches the eventual price about 80% of the time. Use the range. The midpoint is a midpoint.

Exhibit 5. Implied launch price is the region's median-adjusted mean multiple applied to the land figure, with the band from the backtest. Launch quarters are from the Huttons pipeline deck of 31 Aug 2026 and are indicative — a site with no quarter shown has not been guided. Executive condominium sites are excluded entirely: they are price-capped and sold under an income ceiling, so a private multiple does not describe them. Thomson Reserve is excluded for the reason in section 06. Source: POV land-bid tracker, GLS records, Huttons pipeline deck.

Two honest limits on this table. The land bid is point-in-time — a site tendered in 2024 carries two years of market drift that the multiple does not know about. And a projection is not a price list: the developer sets the launch price, and it will sit where it sits regardless of what our arithmetic says. What the table is good for is spotting the sites where the implied number and the neighbourhood's resale market are far apart, and asking why before the showflat opens.

If you are buying a launch

Ask how the developer got the land before you judge the asking price

If the site came from a state tender, the multiple applied to it could reasonably be anywhere from 1.68× to 2.99×, and where it lands tells you something about how hard the developer had to bid. A high implied multiple is not proof of overpricing, but it is the right question to put to the agent.

If it came from a freehold collective sale, expect a tighter, more predictable number — and expect the land underneath it to have cost more, which is already in the price you are being quoted.

If you own in an estate exploring a sale

The multiple is the developer’s margin, not your reserve price

Knowing that launch prices land around twice the land rate tells you what the site has to be worth to a developer for the numbers to work. It does not tell you what your unit is worth, which depends on the reserve, the share value apportionment and how many owners sign.

What it does tell you is that a developer buying en bloc has historically worked to a tighter, lower markup than one bidding at tender — so the room to push a reserve is real but narrower than the headline multiple suggests.

MARKET & POLICIES · INSIGHT 001

How developers price your condo years before it exists

The rule itself, the evidence behind it, and the pipeline it implies.

GUIDES & INSIGHTS · MYTH BUSTER

The freehold premium, argued rather than asserted

What freehold actually buys, and where the premium stops paying for itself.

Dataset — POV land-bid tracker as at 7 August 2026, rebuilt 3 September 2026. 61 launched records with both a land rate and a launch average: 52 state tender, 9 freehold collective. Pipeline timing from the Huttons “Next Housing Cycle” deck, 31 August 2026.