The third proptech wave isn't listings or fractional ownership. It's an intelligence layer that prices, matches, and runs real assets. Where it lines up.
Two proptech waves rearranged how we search and how we own. The third one runs the asset itself, and it needs three things in the same place.
Disclosure: a venture in our portfolio is building in this space, and our compute venture is named below. We flag both up front. Third-party points are attributed.
Proptech has had two big waves, and a lot of money chased both. The first digitized the search: portals, listings, the move from newspaper classifieds to a phone. The second rearranged ownership: fractional, co-ownership, new ways to slice a deed. Useful, both of them. Neither one actually ran the building.
The third wave is different, and it's the one a software studio with a compute thesis pays attention to. It's an intelligence layer that prices, matches, and operates real assets in real time. And it only works when three things finally sit in the same place.
The first two waves changed the interface and the ownership structure, but they left the asset itself running the old way. A listings portal makes a property easier to find. It does nothing to price it dynamically, match the right buyer or guest, or run it more intelligently once it's owned. Fractional ownership splits the deed. It doesn't make the underlying asset smarter, and a fractional building managed by a spreadsheet is just a spreadsheet with more owners.
Both waves stopped at the surface. The intelligence to actually operate the asset, to price it, match it, and run it, never showed up, because the pieces required weren't in place.
For proptech AI to do real work on a real asset, three ingredients have to meet.
The data. Structured, proprietary operating data: how this kind of inventory actually prices, fills, and absorbs over time. Not generic market comps. The real behavioral exhaust of running the asset. (Why that exhaust is a product in its own right is its own field note.)
The model. Something that turns the data into decisions: pricing, matching, occupancy forecasting, service prediction. The model is only as good as the data underneath it, which is why the two are inseparable.
The asset. A real property the model can actually act on, so the loop closes. Without a real asset, you have an interesting model and nowhere to apply it.
Listings digitized the search. Fractional split the ownership. The third wave runs the asset, and it needs the data, the model, and the asset in the same place at the same time.
Most proptech has one or two of these. Almost nobody has all three lined up, which is precisely why the third wave is still early and still open.
Run real-time intelligence across a portfolio of assets and you need inference, steadily and affordably. That's a compute problem, and it's not incidental. A pricing-and-matching model isn't a one-time training run; it's a service that has to keep answering, asset by asset, day after day.
This is where the studio's own compute thesis connects. Griddly, the decentralized GPU compute venture in the same portfolio, is built to make inference capacity more available and more affordable. A portfolio that owns both the real-asset venture and a compute venture has a path to feed the intelligence layer that the asset venture needs, instead of renting that capacity at someone else's margin. The connection is deliberate, and like the rest of the portfolio, parts of it are still being built. We'd rather say that than imply a finished loop.
A pure proptech startup can usually get the data or the model. Getting a real asset to run it on, plus the compute to run it at scale, is hard from a standing start. A studio that has originated a real-asset venture, an intelligence layer, and a compute venture under one roof is unusually positioned to put all three in the same place, which is the whole game in this third wave.
That's the bet. Not another listings site. Not another way to fractionalize a deed. The layer that prices, matches, and runs the asset itself, with the data, the model, and the asset finally in one place, and the compute to keep it running.
Read next: The Layer-1 model: financing the tech under a real asset
Nothing here is an offer to sell a security or investment advice. Several ventures referenced are in active development; descriptions reflect design intent, not a representation that every capability is live. Statements about the proptech market are general context.
Field notes on venture building, AI, and capital. No spam, unsubscribe anytime.
By subscribing you agree to receive AOS Insights e-mails. We use your address only for this newsletter - see our Privacy Policy.
