Who wins and who loses in the next real estate cycle

Of course, I’m tempted to tell you AI implementation will determine the winners of the next cycle. It would be a useful thing for me to believe, given what I do for a living. It’s not what the data says.

The 2026–2030 residential cycle is a capital cycle. It will be decided by who has the balance sheet to survive a $162 billion multifamily maturity wall^[1], the operating-cost structure that comes with scale that smaller operators cannot replicate, and the relationships to buy when nobody else can. AI compounds the firms already winning. It does not rescue the ones already losing. Their outcome has already been set, even if it hasn’t shown up on the balance sheet yet.

I’ll call those firms zombies. Operators that rode a thirteen-year bull cycle and convinced themselves the result was earned. Today, they appear alive from the outside. They do not yet understand that the thing keeping them upright was the rate environment, and those days are gone. Some of them will figure it out in time to sell. Most won’t.

The honest version: AI is going from differentiation to table stakes. Operators who don’t build it into their workflow over the next twenty-four months will fail faster than those who do. AI is real and you should be paying attention. It just won’t determine who wins. Capital and scale will.

The cycle we’re in

Three pressures are compounding at once, and the combination is what makes this cycle structural rather than cyclical.

First, rent growth has gone flat. 2021-vintage underwriting did not price for flat. Multifamily rents were up 0.2% year-over-year through November 2025^[2]. Yardi Matrix forecasts 1.2% for 2026 and no meaningful improvement before 2028^[2]. Sun Belt markets that absorbed the 2020–2022 acquisition frenzy still face 4–5% supply additions in 2026 and 2027^[3][4]. Multifamily starts are down more than 40% from 2023^[4], which tightens the market eventually but does not solve the cash flow problem now.

Second, a historic debt maturity wall is right in front of us. More than $930 billion in commercial real estate loans matured in 2025, well above historical norms^[5], and most multifamily borrowers received extensions rather than refinancing^[1]. The 2026 multifamily maturity calendar steps up roughly 56% from $104 billion in 2025 to $162 billion, with 2027 at $168 billion^[1]. This is not a one-time event. It is a structural repricing that will run through 2027 and into 2028 as extensions expire and lenders run out of patience.

Third, rates have reset. What many believed to be a temporary correction is turning out to be a more permanent reset. The 10-year U.S. Treasury has sat above 4% for more than three years, with only brief dips below. That’s roughly double the post-GFC average and more than eight times the August 2020 low. The trickle-down effects are two-fold: cap rates have expanded, and debt service has compressed leverage.

On the cap rate side, the same NOI now supports a lower valuation. A 2021-vintage property purchased at a 4.5% cap on $4.5M of NOI was worth $100 million. At today’s mid-5s cap rate it appraises at roughly $84 million. Same property, same income, sixteen million dollars of value erased.

On the debt service side, a 3.5% coupon at 1.25x debt service coverage carried 75% LTV. A 6% coupon at the same coverage caps closer to 60–65% LTV. Higher rates mechanically reduce the principal a new loan can support, before any tightening from the lender side.

The two compress together, and the math is brutal. A $100 million property bought with 75% leverage now refinances at 65% of $84 million, about $54.6 million, against a $75 million maturing balance. The $20.4 million gap^[5] is the structural fact of this cycle. It is new equity, mezzanine capital, or a discounted payoff. It is rarely a refinance at the same basis (implication would be ~89% LTV).

The industry has elected to kick the can. Most 2025 maturities received extend-and-pretend modifications instead of forced workouts^[1]. The gap doesn’t disappear with an extension. It compounds into 2026 and 2027 as those modifications expire, with the same cap rate and DSCR math working against the borrower from a deeper hole. Kicking the can only works while lenders remain patient.

On top of all of this, operating costs have reset permanently higher. Multifamily insurance went from roughly $30 per unit per month pre-pandemic to $65 by late 2023, a 119% increase in four years^[6]. For coastal Florida and Texas properties, insurance alone now eats 150 to 300 basis points of NOI that was not in the 2021 underwrite.

The downstream effects show up everywhere capital touches dirt. LP allocations have moved to a small number of mega-managers (more on this in the next section), institutional equity for new development is sidelined waiting for visibility, land deals stop penciling because new development basis can’t compete with the distressed asset bases coming off the maturity wall, and the buyer pool for stabilized assets froze around the Senate housing bill on the BTR side^[21]. Three pressures, all pulling the same direction: flat rents, a massive maturity wall in front of us, and a permanent reset in rates. Firms with low-coupon debt and flexible capital have room to maneuver and capitalize on blood in the water. Firms with 2021-vintage loans maturing into 2026 rates do not. The latter group is the first wave of zombies.

Capital is the variable that decides this cycle

The defining shift in residential real estate over the past three years is that institutional capital has consolidated into a small number of mega-platforms, and that consolidation is what determines the 2026–2030 winners.

The top ten private equity funds captured 45.7% of capital raised in 2025, up from 34.5% in 2024 and the highest concentration since records began^[7]. Real estate tracks the same pattern: the top ten real estate funds raised $68 billion in 2025, 40% of total commitments^[8]. The top fifteen managers control 45% of real estate assets^[8]. Blackstone alone raised more than $52 billion for closed-end real estate funds in the five years through 2024, nearly twice its closest peer^[9]. Of the $4.2 billion that new real estate managers raised through Q3 2025, two startups accounted for more than half^[8]. The capital is not arriving for the rest.

The cause is not mysterious. Limited partners aren’t getting their money back fast enough from earlier vintages, so they’re consolidating their next allocations into fewer, larger relationships, mostly with managers who already have multiple strategies and the operational capacity to deploy size quickly. Management fees hit record lows in 2025, and the math compounds: $100 billion at 1.5% generates $1.5 billion annually, while $1 billion at 2% generates $20 million^[7]. Bigger platforms can survive on thinner fees. Smaller managers cannot.

The downstream consequences are direct. Greystar manages more than 946,000 apartment units and owns more than 122,000^[10], simultaneously the largest manager, owner, and developer by units started. The NMHC top-fifty managers oversee 24% of the nation’s apartment stock as of 2026, up from 21% in 2025^[11]. A three-point jump in a single year is not noise. It is smaller owners folding operations into the largest platforms because their cost structures no longer work standalone.

SFR shows the same pattern, but with sharper edges. Invitation Homes owns over 86,000 homes^[12] at a 36.1% operating margin^[13]. American Homes 4 Rent owns roughly 61,000 homes at 24.3%^[14]. Both are structurally higher than what scattered-site operators produce, and the gap between them suggests the returns to scale within SFR are non-linear. The largest operator captures a disproportionate share of the margin. Roughly 1,200 basis points across an extra 25,000 homes, with a similar gap between AMH and the scattered-site cohort.

Invitation Homes acquired ResiBuilt in January 2026^[15] to bring purpose-built BTR development in-house. AMH operates an in-house development platform that has been doing the same for years. Both are migrating their growth from scattered-site SFR acquisition into purpose-built BTR development because SFR stopped penciling. By 2025, the cost advantage of buying SFR versus building purpose-built BTR had compressed to under $20K per home. At that spread, the operational drag of acquiring scattered single-family homes one at a time outweighs the basis advantage; BTR becomes the clear preference as in-house development restores margin through control over land, site, and product spec. Vertical integration here is not strategic vision. It is what you do when the off-the-shelf supply chain stops giving you a margin.

The pattern repeats in homebuilding. D.R. Horton and Lennar dominate new-home delivery in the top fifty US metros, and the gap is widening. Captive mortgage and title arms plus the balance sheet to absorb 2024–2025 margin compression let them outspend smaller builders on incentives by a factor of five^[19]. Lennar’s per-home incentive spending went from $12,000 in Q3 2022 to $62,800 in Q4 2025, a 420% increase. Gross margin fell from 22.1% to 17.7% over the same period^[19]. That is real pain, but it is pain Lennar can absorb and small builders cannot. That is what consolidation looks like in real time: the largest player tolerates the punishment because it knows the smaller ones can’t.

Management consolidation, SFR/BTR consolidation, and homebuilder consolidation are the same story playing out in three different segments. The next four years of winners are being selected mostly on three criteria: low-cost capital access, operational scale, and balance sheet to buy distressed assets off the maturity wall. That list is short, and it is getting shorter.

Where AI actually matters

Now AI, on its merits. The industry narrative says AI is the main event. The evidence does not, at least in its current form. The reason is that most firms are buying it the wrong way.

The dominant pattern inside real estate today is unguided. Companies buy a stack of generalist AI subscriptions, distribute logins across the org, and assume value will emerge. It mostly doesn’t. There are two reasons. First, employees aren’t given a structured way to extract leverage from the tools, so usage stays anecdotal and the productivity gains never compound. Second, generic vendor tools, by construction, are built to work for any company, and therefore optimize for none. They sit on top of the workflow rather than inside it. Real leverage comes from the opposite direction: integrating a firm’s data into a single knowledge base and automating the firm’s specific workflows on top of that base. Everything else is marginal.

The empirical record is consistent with this. MIT’s Project NANDA published “The GenAI Divide” in mid-2025, analyzing 300 public AI deployments alongside 150 leader interviews and 350 employee surveys. The headline finding: 95% of enterprise AI pilots produce no measurable P&L impact^[16]. Only 5% generate meaningful value. The failure is organizational, not technical. Most budgets go to sales and marketing, where ROI is lowest. The highest returns come from back-office automation, where budgets are thinnest^[16]. Vendor-sourced AI tools succeed roughly 67% of the time. Internal builds succeed at about a third of that rate^[16]. Even if the true failure rate is 70% rather than 95%, it is still the dominant outcome. AI is not producing reliable enterprise returns yet, at scale. My read is that the failure rate inside real estate is higher than the cross-industry average, not lower, because the workflows in real estate are even more company-specific.

The academic productivity literature reinforces the same boundary. The most rigorous study to date, by economists at Stanford and MIT, tracked 5,179 customer support agents using a generative AI assistant: 14% productivity gain on average, 34% for novice workers, almost no effect on experienced ones^[17]. Real gains, real signal, narrowly bounded to specific tasks performed by less experienced people. That is not firm-level transformation, and it is not what closes the operating-cost gap on the maturity wall.

Then there is Zillow, the cautionary tale most people invoke wrong. The conventional reading is that AI failed at scale. The actual lesson is sharper. Zillow Offers spent roughly four years building an algorithm trained on Zestimate valuations to bid on homes at scale, acquired close to 10,000 homes in Q3 2021, shut the business down two months later, and took $569 million in writedowns^[18]. The Stanford GSB post-mortem identified the structural problem: algorithmic intermediation is only profitable in the most liquid and easy-to-value markets, which is precisely where the returns to adding technology are lowest^[18]. The lesson is not that AI doesn’t work. The lesson is that betting the firm on a single algorithmic decision in an illiquid market doesn’t.

Each of these threads points to the same distinction. There are two AI stacks emerging inside firms, and they do different work.

Vendor AI is overhead. The list grows by the quarter: leasing chatbots, generic copilots, off-the-shelf revenue management, deal-sourcing platforms, comp aggregators. It will commoditize, and probably already has. Every operator will buy it. Nobody will gain advantage from it. Treat it as the next ERP or CRM line item: necessary, undifferentiated, eventually a pure operating cost.

Workflow AI is leverage. Workflow-specific systems integrated into a firm’s proprietary data and processes. The firm’s underwriting cadence, asset-management reporting, deal screening logic, IC memo production, loan-doc tracking, lender reporting, partner update assembly. This kind of AI doesn’t commoditize because the workflow is the firm. It cannot be bought as a subscription, because the value is in the integration, not the model. Operators who build it over the next twenty-four months own a position generic vendor tools cannot replicate.

The signal that this distinction is real is already showing up at the top. On May 4, 2026, Anthropic, Blackstone, Hellman & Friedman, and Goldman Sachs announced a $1.5 billion joint venture to embed engineers inside enterprises and build workflow-specific AI deployments at scale^[24]. OpenAI is reportedly pursuing a near-identical structure with TPG and Bain. Jon Gray, Blackstone’s President and COO, said the venture is built to “break down one of the most significant bottlenecks to enterprise AI adoption by expanding the number of highly skilled implementation partners.” The largest residential capital allocator on earth made the same call this section is making, in the same news cycle as this post: vendor stack alone is not the deployment that compounds. The implementations that compound are embedded and workflow-specific.

The defensible claim is narrower than the headline. AI is a survival tool for the mid-market, not a moat. Operators who deploy nothing will enter 2029 with a measurable operating-cost disadvantage they cannot close. Operators who deploy generalist subscriptions will enter 2029 with the same cost disadvantage and a higher software bill. Operators who build workflow AI over the next twenty-four months will widen their cost advantage. AI compounds existing advantages. It does not create new ones.

Generalist AI implementation versus workflow-specific AI does not decide which firms win the cycle. Capital does that. But it decides whether mid-market operators close the cost gap with the firms that already have it, or fall further behind. That distinction deserves its own treatment, and it’s the subject of the next post in this series.

Winners and losers

Three structural winners. They share three traits in some combination: low-cost capital access, operational scale, and a balance sheet built for the maturity wall.

Vertically integrated residential operators with balance sheet. The institutional SFR REITs, the largest apartment operators with in-house management, and the public homebuilders with captive financial services. Invitation Homes at 36% operating margin^[13], AMH at 24%^[14], both well above what scattered-site operators can produce. Lennar’s incentive spending up 420% from Q3 2022 to Q4 2025, gross margin down 440 basis points over the same window, and Lennar still consolidated share^[19]. The compression is real pain, but it is pain the giants can absorb and small builders cannot. That asymmetry is the entire homebuilder consolidation story in this cycle, and it isn’t slowing down.

Mega-platform managers with credit strategies. Blackstone, Brookfield, Carlyle, Ares, and a handful of peers benefit from three converging dynamics: the retreat of regional banks from CRE lending, the maturity wall, and the rise of private credit. Real estate debt funds had their strongest fundraising year since 2021, closing $51 billion in 2025^[8]. Debt fund returns outpaced equity at 5.5% YTD versus 4.0% for ODCE through Q3^[8]. Borrowers facing refinance gaps need rescue capital, preferred equity, and mezzanine financing. The firms with the platform to provide it at scale capture risk-adjusted returns that equity strategies cannot match, and end up owning residential equity at attractive basis three to five years out when the rescue capital converts. This is how Blackstone built its SFR exposure in 2012–2014. The same playbook is running now in multifamily.

Supply-constrained coastal multifamily. New York finished 2025 with 12-month absorption of 27,704 units^[20], leading all US markets, while deliveries in NYC and Chicago have been held to 1–2% of existing stock for years^[4]. New York, Chicago, Kansas City and similar markets are delivering rent gains of 9–13% in 2025^[2], while Sun Belt markets with 4–5% supply additions still face modest deflationary pressure^[4]. The flight from coastal to Sun Belt was a real trade from 2015 to 2022. The flight back is happening quietly now.

Four structural losers. They share the inverse: undersized balance sheets, narrow capital access, and exposure to 2021-vintage assumptions that will not survive contact with 2026 reality.

Small and mid-cap homebuilders without land optionality. They cannot match the incentive spending or absorb the gross margin compression that the giants have absorbed. Expect consolidation through 2028, mostly via land pipeline sales to the majors rather than corporate acquisitions.

Small and mid-sized BTR and multifamily developers without pipeline. Land optionality is what insulates a developer from the gap between a 2024 underwrite and a 2027 reality. Without it, you’re building on contracts at a land basis that no longer pencils, with construction debt at rates that didn’t exist when the dirt was tied up. This category took the worst of both ends of the cycle: the equity allocator pool collapsed at the same time the institutional buyer pool for stabilized BTR froze around the Senate housing bill^[21]. Expect a wave of land dispositions into the public builders and the mega-platforms over the next eighteen months.

Pure allocators at mid-market scale. The firms raising $500 million to $2 billion per fund, deploying through third-party operators, and charging traditional 2-and-20 fees face a squeeze on both ends. Mega-managers have the LP capital^[8]. Vertically integrated operators have the operational alpha^[22]. Most mid-market allocators will consolidate, wind down without raising successors, or pivot to operator models. That pivot is already happening.

Sun Belt operators with 2021-vintage leverage. The late entrants who bought at 3.5–4% debt, underwrote rent growth at 5%, and assumed 4.5% exit caps are approaching maturity dates with rates double, rents flat, and cap rates in the mid 5s^[5]. Q3 2025 distressed CRE volume reached $126.6 billion, with $22.8 billion in multifamily alone^[5]. The late-cycle Sun Belt levered buyer is the archetype zombie of this cycle: still operating, still underwriting new deals on the same playbook, with the maturity dates that will eventually decide the outcome already in the calendar.

The case against this view

The strongest opposing case says the picture is overcomplicated. The real driver is who has dry powder and who has maturity dates. Everything else, including operational sophistication, AI deployment, and vertical integration, is noise around the capital structure. Blackstone wins because Blackstone has $1.3 trillion in AUM^[9]. Mid-market operators lose because they don’t.

The case has merit. Fundraising concentration tracks deployment power closely^[7][8], and the firms making the largest distressed multifamily acquisitions in 2025 are the mega-managers, not the specialists. But it overreaches. Operational margins inside institutional real estate are widening, not narrowing; the spread between Invitation Homes, AMH, and the scattered-site cohort is hundreds of basis points and compounds over holding periods. And the mega-manager advantage is most decisive in acquisition, not operation. Blackstone can buy cheaply off the maturity wall, but the returns depend on operating execution Blackstone does not perform itself, which is why its residential acquisitions are mediated through platforms like Tricon and joint ventures with specialized operators. Capital dominates acquisition timing. Operational integration dominates holding-period returns. The firms that combine both dominate everywhere.

What would change the call

Three signals would force a different conclusion.

If the Senate housing bill passes in its current form, the BTR side of the picture collapses. The current draft forces institutional developers to sell BTR communities to homeowners after seven years^[21], destroying the hold-and-operate economics BTR has been priced on since 2020. Firms that priced forward commitments into 2024 take losses. Legislative risk is live as of this writing.

If public REITs start reporting measurable AI-attributable productivity gains in earnings releases, numbers that hold up to analyst scrutiny rather than vendor-sponsored case studies, AI moves from table stakes to moat. The threshold to watch is roughly 15–20% measured ROI against the current MIT empirical baseline of near zero^[16]. As of this writing, no material number of public REITs are reporting these numbers. That gap, between vendor demo and reportable enterprise productivity, is the gap Stirrup Lane was built to close.

If the maturity wall resolves through extensions rather than distress through 2028, the capital-consolidation thesis weakens. The softer-landing view, articulated by some CoStar analysts, is that lenders will keep working collaboratively rather than forcing dispositions^[23]. The counter-evidence is the run rate. Q3 2025 alone produced $126.6 billion of distressed CRE volume, an 18% year-over-year increase^[5]. That is a single quarter at a roughly $500 billion annualized pace, a level not seen since the GFC. The data is running toward distress, not away from it. The scenario is not settled, but the slope is.

---

What it adds up to

The 2026–2030 cycle is less interesting than the industry wants to believe. It is a capital cycle, a rates cycle, and a margin cycle. The firms that win it will win it for reasons recognizable to a real estate investor in 1995. Beta is available through the mega-managers, the institutional REITs, and the top five homebuilders. Alpha sits with operators who can buy distressed Sun Belt assets off the maturity wall and hold through the 2027–2028 supply tightening, or with credit strategies that provide rescue capital into refinance gaps.

AI doesn’t change any of that. Deploy it anyway. Not because it wins the cycle, but because the cost of not deploying it is the operating disadvantage that nobody enters the next cycle from. Build AI into the workflow. Then go win the cycle on the things that have always won cycles.

Capital. Basis. Execution.