American builders finished 5.7 million square feet of retail space in the second quarter, the lowest quarterly and rolling four-quarter totals CBRE has recorded, as Modern Retail reported this week. Every 2026 store programme in fashion retail was written above that line. The plans were underwritten on capital, on consumer demand, and on a shop-floor technology stack priced against a fleet that grows. None of those three is the binding constraint this year. The buildings are.

The rolling four-quarter number is the one that carries the argument. Single quarters bounce, because deliveries slip across period ends and get restated once late completions are counted. A trailing-year total sitting at the floor of the series says what one quarter cannot: this is a level, not a scheduling gap. CBRE has tracked the metric since 2005, and its peak was more than 25 million square feet delivered in a single quarter at the end of 2015, as Bisnow reported in May. Completions now run at under a quarter of that, and have done so long enough for the twelve-month average to catch down.

Demand is not what is missing, which is why the supply figure deserves a second reading. Retail availability held flat at 4.9% in the second quarter, on the back of a full year of positive net absorption. That is a market in which tenants keep taking space and nothing accumulates behind them. CBRE ties the 2.4% year-on-year rise in average asking rent to low completions and a year of sustained absorption, in its own Q2 figures. Rent that climbs because nothing is being built is a supply price — and supply prices do not answer to better demand forecasting.

Nothing is being built because of a spread, not a shortage of willing money. James Cook, JLL’s global head of retail research, told Modern Retail that rents are growing in most US markets “but they haven’t grown fast enough or high enough to justify it making economic sense,” with land, labour and materials all expensive. CBRE names rising construction costs and labour shortages as the direct cause of the record-low deliveries. A developer who cannot clear the hurdle rate does not build, whatever the financing costs. That is why the pipeline has not responded to three years of visible scarcity, and why it would not respond quickly even if the spread reopened tomorrow.

Which means every store-technology business case that multiplies a per-location return by a planned store count has quietly lost control of its multiplier.

Fashion retail spent this year buying operating systems for a shop floor it assumed would replicate. Clienteling platforms, RFID inventory decisioning, computer-vision shrink control and AI labour scheduling are licensed per site against a fleet expected to expand, a convergence we traced in March. When the fleet does not expand on schedule, the return has to come out of the stores already trading. That converts the pitch from coverage across more square feet to productivity within the same ones, which is a harder sale on a slower payback. Nothing in the vendor deck needs to change for the economics underneath it to invert.

Scarcity of this kind sorts tenants by balance sheet, not by taste. Uniqlo is adding eleven US stores across spring and summer — enough, absent closures, to lift its American count from 78 to roughly 89 — with flagships in Chicago and San Francisco and four more in New York, as FashionUnited reported. Executing that against near-zero availability is not a merchandising win. In a market this tight the landlord picks the covenant it can underwrite for a decade, and a balance sheet like Fast Retailing’s clears that test in a way a growing specialty brand’s does not. CBRE’s Ebere Anokute, once a real estate manager at Warby Parker, describes “a major flight to quality” in the industry, which is the same fact from the landlord’s side of the table. If completions hold near this level through 2027, then a mid-sized brand’s expansion plan stops being a decision and becomes a queue position.

The strongest objection is that completions are the wrong number to watch. Expansion in 2026 runs through backfill rather than new build, and closures keep releasing inventory; Cushman & Wakefield’s James Bohnaker says demand to fill vacated space is “tremendous right now,” with survivors holding “the conviction right now to open brick-and-mortar stores.” Chris Ressa of DLC Management describes landlords and tenants finding “a lot more creativity and flexibility” to get stores open. JLL, counting separately, put net absorption at 10.2 million square feet in the second quarter, its second-strongest in two years, crediting malls and neighbourhood centres with returning to positive territory. Absorption running at roughly double the completions figure is not the behaviour of a market that cannot find room. For the pipeline to bind, released space would have to differ from sought space in format and in location.

It differs, though the trackers do not agree on where. CBRE’s own segment data, cut by Chain Store Age, shows power centres leading absorption in the second quarter with close to 1.6 million square feet while neighbourhood, community and strip centres turned negative for the first time since mid-2025 — the opposite read JLL gives the same category. The two firms draw their format lines differently, but CBRE’s own detail explains the split: the stated cause of the strip-centre contraction was apparel boutique closures and shrinking bank branches, even as grocery and apparel retailers were named as the lead driver of absorption in Boston and Los Angeles’ downtown cores. Apparel is not retreating from retail real estate; it is retreating from suburban strip centres and pushing into center-city footprints, and a carved big box beside a power-centre parking field does not substitute for a 3,000-square-foot unit on a corridor a brand has courted for five seasons. Backfill is a fact about square footage; expansion is a decision about specific addresses, and the two look interchangeable only in a spreadsheet.

A rationed pipeline turns a growth plan into an allocation problem, and allocation rewards whoever is already large. A chain that cannot open the stores its technology budget assumed has two honest options: pay above market for the addresses that exist, or make the trading estate earn what the new one was supposed to. The second is what the software was always better suited to, and almost nobody bought it on that basis. What the record-low pipeline removes is the third option, in which the fleet quietly grows into the forecast and no one has to choose between the other two. That choice was always coming. Construction has set the date.