The absorption gap

I spent two days at PropTech Connect in London, and the most useful thing I heard was not about property technology at all. It was a ConTech truth, said out loud by the people who buy and build: the tools are no longer the bottleneck, and neither is the capital. What is scarce is the built environment's capacity to absorb either.

Day one had a name for it. Evan Petkov of Optiml+ put it in numbers, citing PwC and JLL research: 92% of firms are running AI pilots, around 5% hit the objectives they set, and 20% of firms capture roughly 74% of the value. He calls it the pilot penalty. Adoption is everywhere. Value is rare.

The rest of the day argued the same point from every seat in the room. SEGRO's first CIO, Richard Corbridge, made the case for going slow first: build the data foundation before the tools, and put business leaders, not IT, in the chair of the governance board. A panel on technical debt reframed the problem entirely. What looks like technology debt is usually organizational debt that technology exposes; build a better business, not better tech. The teams from King's Cross and Findable showed where the value actually hides: not in the model, but in the hundred thousand-plus building documents no agent could reach until someone made them legible.

The tools are becoming table stakes. The edge is decision quality, and the ability to prove it.

Day two moved to the supply side, and the numbers got larger. The National Housing Bank, live since April as a subsidiary of Homes England, arrived with up to £16bn of debt, equity and guarantees, aiming to crowd in more than £53bn of private capital toward 500.000 homes. Its pitch to the room was disarming: bring us your problem, not a product. Then came the honest part. A founder called Homes England "a fortress" to sell into. Its own chief executive conceded the institution is "quite conservative," with the innovation program still being designed. Capital at generational scale on one side; an institution still learning how to absorb it on the other.

The gap is not a London anecdote

Leave the conference and the same shape appears at every scale.

Look at the largest construction demand signal on earth. The four biggest hyperscalers have guided to roughly USD 725B of combined 2026 capex, up about 77% year on year, and US data center construction starts hit USD 77,7B in 2025. That is capital with no patience. Yet of the 16GW slated for 2026, only around 5GW is actually under construction, and 30 to 50% of the pipeline is expected to slip or cancel. The binding constraints are not funding or ambition. They are equipment lead times, interconnection queues, and permitting: roughly 48% of US data center development stalls at zoning, New York froze permits for anything drawing 50MW or more in July, and more than 20 projects were canceled in Q1 2026 alone, accounting for USD 41,7B and 3,5GW.

That is the absorption gap in concrete and copper. Unlimited money meeting a supply chain and a permitting system that cannot take it.

Venture capital tells the same story from the other end. ConTech investment in H1 2026 was USD 2,884B across 153 disclosed deals, down roughly 19% in capital and 20% in deal count year on year. Money did not leave the sector so much as get selective. Strategic investors took part in about 37% of transactions, a record share. Only two Series C rounds and eight late-stage transactions closed in the entire half. The mid-stage funnel is thin, and the cheque increasingly comes from an incumbent rather than a fund.

• Pre-seed to Series A70%
• Series B and later30%

Seven in ten deals landed before Series B. Only two Series C rounds closed in the entire half.

Source: Last Week in ConTech, The State of ConTech H1 2026

The sharpest contrast came from outside the UK. Dubai's Land Department demonstrated real-time, open transaction data and a property purchase closed in seven minutes. Its DIFC innovation hub runs licensing at heavy subsidy and an "inside man" model that, by its own account, lands pilots at something like 100 to 1 against cold outreach. One UK REIT's AI lead asked, from the floor, how to get access to Dubai as a testbed, because his own market could not show him tangible AI returns. Two capitals, both awash in money. One has engineered its absorption capacity as policy. The other has not.

Money is not the differentiator. What separates these markets is whether anyone built the capacity to absorb it.

Source: Stratly analysis

What it means

The defensible position is no longer the model or the feature. Everyone can buy the model. The moat is the layer underneath it: owning the point where data is created, and the decision that data enables. The distinction that matters is between a system that records what happened and one that captures why, because agent accuracy compounds on decision trajectory, not on archives. A system of record that cannot say why will not carry its data gravity into the agent era.

The market has already started paying for this, and the clearest evidence is where you would least expect it. In July, three robotics firms took €179,4M of the month's €234M in ConTech funding, about 77% of the total. Not one of them sells a machine. TerraFirma retrofits excavators and bids earthworks as a subcontractor. Gritt owns no hardware and fits AI control onto rented equipment. Monumental quotes a finished wall at fixed price and timeline, and absorbs the cost if its robots fall behind. All three price the outcome and carry the liability, which is exactly how a general contractor already buys a trade package. The unlock was not the robot. It was a commercial model the buyer's organization could absorb without changing.

The same logic is repricing exits. There are two tracks now. Point solutions leave at single-digit revenue multiples, often as acqui-hires. Companies with irreplaceable proprietary data leave at a different number entirely: Autodesk paid roughly 27x forward ARR for MaintainX. Buyers have also started treating vendor survivability as a procurement criterion in its own right; McCarthy Building Companies chose Palantir over startups partly on those grounds.

So, two consequences.

If you are building: stop competing on pilot count and model access, because both are commodities now. Build where the absorption gap is, own the authoring layer rather than the archive, and price the outcome you can carry liability for, not the feature you can ship. Then make your buyer's adoption cost near zero, because their capacity to absorb you is your real constraint, not your roadmap.

If you are backing: the pilot penalty is the diligence lens. When 20% of firms take 74% of the value, the question is not whether a company has AI. It is whether it changes a decision, whether the buyer can defend the result, and whether the commercial model fits how the customer already purchases. With two Series C rounds in a half, a mid-stage plan that assumes a conventional next round is a plan with a gap in it.

The capital and the models arrived first. The returns will go to whoever closes the distance between them and the built outcome.

The Scan

Capital got selective, and strategic. ConTech took USD 2,884B across 153 disclosed deals in H1 2026, down about 19% in capital year on year, with strategic investors in a record 37% of transactions and only two Series C rounds in the half. Last Week in ConTech, The State of ConTech H1 2026

India bets on delivery. EY-Parthenon and CREDAI project that generative AI will compress both sales and project delivery across Indian real estate and construction. The same absorption question as London, from a different data-maturity starting point, in a market moving faster than its record-keeping. Construction Week India, June 2026

Gulf builds a testbed, on purpose. Dubai PropTech Hub, a DIFC initiative, signed a partnership in August giving startups access to Innovo Group's live construction projects and engineering teams. Not capital, not a program: real jobsites to prove on. Read it next to the Land Department's open transaction data and the pattern is clear. Absorption capacity as policy. PropTech Connect, August 2026

Latin America is the same gap, inverted. London has capital waiting on capacity. Latin America has demand waiting on infrastructure: a USD 709B construction industry, around 70% of contractors and developers delivering late, labor informality running from 40% in Brazil to 60% in Colombia and Mexico, and a housing deficit of some 70 million people. The forward read lands on this edition's spine: field capture going voice-first and WhatsApp-native, and AI agents moving from systems of record to systems of action. Zacua Ventures and Leonard (VINCI), June 2026