In 1993, one company deliberately split itself in two. One half kept the buildings. The other half kept the business of running them.

Thirty-two years later, the half that kept the buildings owns interests in roughly seventy hotels. The half that kept the running of them reports a system of 9,805 properties and 1,779,936 rooms across 145 countries, while owning or leasing less than one percent of them.

That is not a story about hotels. It is the clearest natural experiment real estate has ever run on a question the industry is about to face everywhere: what actually makes a property company big.

The claim this piece defends is that ownership and scale are being quietly decoupled, and that the largest real estate companies of the next cycle will be operating platforms that own comparatively little. Not because owning is bad, but because owning has a growth ceiling built into its arithmetic, and operating does not.

Ownership Grows With Capital. Operating Grows With Systems.

Start with the mechanics, because everything else follows from them.

If your business is owning, growth is a function of capital. Every additional building requires the money to buy it, plus the debt capacity to carry it, plus an investor willing to fund it at a return that clears. Double the portfolio and you have roughly doubled the capital employed. There is no configuration in which this changes, because the asset is the capital. Growth is therefore bounded by the balance sheet and by whatever the capital markets will tolerate this year.

If your business is running buildings, the arithmetic is different in kind, not degree. The capability that lets you operate the hundredth building is largely the same capability that operates the five hundredth: the systems, the standards, the pricing logic, the reporting, the training, the brand. Once that capability exists, adding the next building consumes very little incremental capital. Growth is bounded not by money but by how transferable your operating capability is.

Call these capital-bound growth and system-bound growth. They produce completely different curves. Capital-bound growth is linear and expensive and pauses whenever markets tighten. System-bound growth compounds, because each new building added to the platform makes the platform slightly better and costs almost nothing to absorb.

Real estate has spent a century optimising the first one and treating the second as a service function. That is the assumption now coming apart.

The Experiment Already Ran, and It Finished

Hotels tested this to completion, and the results are on public record.

Marriott's 1993 restructuring separated the operating business from the property-owning business. What became Host Hotels and Resorts took the real estate. Marriott International took the running of it. Both still exist. Both are substantial. Their trajectories are not comparable.

Marriott's most recent annual filing describes it plainly as a franchisor, operator, and licensor that owns or leases very few of its lodging properties, and puts the system at 9,805 properties and nearly 1.78 million rooms across 145 countries, with a development pipeline of around 4,100 further properties and nearly 610,000 rooms. Its scale is measured in what it runs, not what it holds.

Host, meanwhile, is the other half of the same idea, and its filings are equally direct: all of its hotels are managed by third parties under management or operating agreements. It owns high-quality buildings and is one of the largest owners of Marriott and Hyatt hotels. It does not run them.

The instructive part is that neither company is failing. This is not a story of a winner and a loser. It is a story of two genuinely different businesses that were once bundled together and turned out to have different scaling properties once separated. One grows by raising capital. The other grows by signing contracts. Over three decades, that difference in mechanism produced a difference in size that no amount of operational excellence on the ownership side could have closed.

What Happens When You Try to Scale by Owning

The counter-example is even more useful, and it comes from housing.

For more than a decade, institutional investors have been buying single-family homes in the United States with access to capital on a scale individual landlords could never match. This was as close to an unconstrained test of the ownership model as the modern market has produced. Serious money, patient horizons, sophisticated operators, a fragmented target market, and years to work.

The result, according to a March 2026 report from the US Government Accountability Office, is that across six major metropolitan areas studied from 2018 to 2024, institutional investors ended up owning between one and three percent of all single-family homes. Their share of single-family rental homes specifically ranged from four percent in Seattle to twenty-two percent in Jacksonville, but as a proportion of the housing stock, the number stayed small.

Set the two facts side by side. Institutional capital, buying aggressively for years in the world's deepest housing market, reached low single digits of the stock. Marriott, buying almost nothing, reached 145 countries.

The difference is not ambition, competence, or access to money. It is that one model has to purchase every unit of growth and the other does not.

Why This Is Only Now Spreading Beyond Hotels

If the arithmetic is this clear, the obvious question is why the rest of real estate did not separate decades ago. The answer is a condition that has to be met first, and it is worth naming precisely.

Owning and operating can only split when operating capability becomes separable: codifiable enough to be written down, transferable enough to be taught, and measurable enough to be enforced at a distance by an owner who is not in the building. You cannot contract for judgement. You can contract for a system.

Hotels met that condition early because the industry had already reduced operation to standards. A brand specified how a room was cleaned, how a guest was checked in, what the property looked like, and how performance was measured, and the reservation and loyalty infrastructure gave the operator something the owner could not replicate alone. That combination is what made a management contract enforceable, and an enforceable management contract is what makes the separation possible.

Most other property sectors failed that test for a long time, not because the work was harder but because it was tacit. It lived in the heads of a regional manager and a maintenance lead. It varied by building. It could not be specified in a contract or verified from outside, so ownership and operation stayed bundled by necessity.

That is what has changed. As operating processes move onto shared systems, the tacit becomes explicit. Standards get encoded rather than remembered. Performance becomes observable at portfolio level rather than inferred from a monthly report. Once a way of running buildings can be specified, transferred, and verified, it can be sold to owners who will never set foot on site. The separability condition is being met, sector by sector, and the sequence follows the operational intensity of the asset: the more a sector's returns depend on how the place is run, the more valuable a proven operating platform is, and the sooner the split becomes attractive to both sides.

Two Ways to Get Big

The Prediction

Within the next decade, the largest real estate companies by any operational measure will not be the largest owners. They will be operating platforms running portfolios they do not hold, and the industry will get used to describing scale in units under management rather than assets under ownership.

Three consequences follow, and they are worth thinking about before the market forces them.

The first is that these become two different businesses with two different investor bases. An owner is selling exposure to an asset class. An operator is selling a contractual earnings stream that grows without proportional capital. Those are not variations on a theme. They attract different capital, are judged on different metrics, and should probably not be run by the same management team pretending it is one company.

The second is that mixed firms will face an uncomfortable question. If you own and operate, you are running a capital-bound business and a system-bound business inside one structure, and the capital-bound half will set the pace for both. The scaling advantage of the operating side stays invisible as long as it is only ever applied to buildings you happen to own.

The third is the risk most owners have not priced. If operating capability separates and you have not built any, you do not stay a bundled company. You become a PropCo by default: the party providing the capital while somebody else provides the business. That is a legitimate position, and Host has made a real business of it for thirty years. But it should be a decision, not an outcome you discover.

The buildings will still be there. Someone will still have to own them, and owning good buildings will remain a perfectly sound way to make money. What is changing is that owning them will stop being the thing that makes a real estate company large. Scale is moving to the part of the business that does not have to be purchased, one building at a time, forever. The next giants will be the firms that noticed which half of the business compounds.

Frequently Asked Questions

1. What does asset-light mean in real estate?
An asset-light model is one where a company runs property it does not own, earning management or franchise fees rather than returns on owned assets. Marriott is the clearest example: its 2025 annual filing describes a system of 9,805 properties and nearly 1.78 million rooms while stating that it owns or leases less than one percent of them.

2. Why does operating scale faster than owning?
Because owning requires capital for every additional building, while operating requires capital mainly to build the system once. Adding another property to an established operating platform consumes very little incremental capital, so growth is limited by how transferable the operating capability is rather than by the balance sheet.

3. What was the 1993 Marriott split?
Marriott restructured into a property-owning company, which became Host Hotels and Resorts, and an operating and franchising company, Marriott International. It is the clearest long-running example of separating ownership from operation, and both companies still exist in those roles today.

4. Does this mean owning real estate is a bad business?
No. Owning good assets remains a sound way to earn returns, and Host Hotels has built a substantial business doing exactly that. The argument is narrower: ownership is not the mechanism that produces very large companies, because it grows in proportion to capital deployed rather than compounding on a system.

5. Why is the separation of owning and operating spreading beyond hotels?
Because it depends on operating capability being separable, meaning codifiable, transferable, and verifiable from outside the building. Hotels met that condition early through brand standards and shared distribution systems. Other sectors are meeting it now as operating processes move onto shared systems that make standards explicit and performance observable at portfolio level.

6. How will real estate company size be measured in future?
Increasingly by units under management rather than assets under ownership. That is already the convention in hospitality, where scale is described in rooms and properties in the system regardless of who holds title, and it is the natural measure for any sector where operating and owning have separated.

7. What should an owner do if they do not want to become a pure asset holder?
Treat operating capability as something to be built and made transferable, rather than as an in-house cost of holding property. The practical test is whether the way your portfolio is run could be specified, taught, and verified well enough that a third-party owner would pay for it. If it could not, the capability is tacit, and tacit capability does not scale.