8. What Survived the PropTech Bubble

What the CEO Admitted on Camera

In November 2021, Zillow CEO Rich Barton sat down for an earnings call and read investors a single sentence: “We’ve determined the price of failing quickly is far less than the price of continuing to lose money scaling an operation that isn’t going to work."" Moments later, the company announced it was shutting down Zillow Offers, its iBuying business, entirely. That year alone, it lost $881 million and had to let go a quarter of its staff.¹

Just months earlier, this business had looked like the boldest answer Silicon Valley had ever posed to real estate. Zillow had built its name on Zestimate, a free service that told you what your house was worth. Then someone asked: if you trust that estimate enough, why not buy and sell at that price yourself? Scan listings with an algorithm, calculate a fair price, buy with cash within days, renovate, and resell — that was iBuying. The homeowner got instant cash without the months-long wait of going through an agent; the company kept the spread and the fees.

The problem was that this business needed exactly one thing to work: “knowing the price” and “having the nerve and timing to actually buy and sell at that price” had to be the same skill. In practice, they were entirely different skills. Zestimate’s price estimates themselves weren’t bad — the error rate on listed homes ran around 2%, on par with a decent human appraiser. But once competition heated up, the algorithm stopped defending accurate estimates and started quoting prices meant to “win.” Homes went for 10–20% over market, in some cases more than $100,000 over. With thousands of units of that inflated inventory piled up, interest rates began climbing in the second half of 2021. A software company could simply pivot. Zillow was left holding thousands of actual houses with actual roofs and yards. There was no way to reverse course.

Around the same time, America’s two other iBuyers, Opendoor and Offerpad, hit the same wall. Their revenue fell by more than half within two years.² Opendoor fared worse still. The US Federal Trade Commission fined it $62 million for “buying above market and then advertising as if it had sold below market.”³ The algorithm hadn’t miscalculated. Between the calculation and the advertising, and between the calculation and the actual execution of the trade, sat human judgment and market liquidity — variables a spreadsheet doesn’t capture.

This chapter asks one question: across the five-year PropTech cycle that ignited in 2021, went out in 2023, and flared up again in 2025, what died, and what survived? And do the survivors share something in common?

Down to a Quarter in Three Years

Start with the numbers. In 2021, global PropTech venture funding hit an all-time high of $32 billion.⁴ In 2022, it fell to $19.75 billion, and in 2023, to $11.38 billion — a 42% year-over-year drop. Q1 2023 alone was down 77% from the same period a year earlier.⁵ In three years, the funding pipeline shrank to roughly a third of its peak.

This collapse wasn’t unique to PropTech. Every “hypergrowth story” built on low interest rates was liquidated in the same window. But in the real estate arena, this liquidation was unusually brutal, for a simple reason. Real estate is, by nature, capital-intensive and slow-turning. The zero-rate environment of 2021 created an illusion — layering “software-company growth multiples” onto a slow industry. A valuation formula built for startups whose revenue doubled every year got applied to companies that bought and sold bricks and mortar. When rates rose, the illusion lifted. A software company can absorb a modest rise in the cost of capital; a company that has stacked physical assets onto its balance sheet takes the shock of higher rates far more directly, and far faster.

2025 brought a reversal. Global PropTech funding rebounded 67.9% year-over-year to $16.7 billion, and funding flowing into AI-focused PropTech surged 176% by early 2026.⁶ But reading this rebound as a rerun of 2021 misses the point. As we’ll see, today’s capital isn’t chasing a story about “creating a new asset class” — it’s chasing a much narrower, much more verifiable promise: cutting the cost of existing workflows. Capital markets, having gone through one full boom-bust cycle, have come back as a far tougher grader.

The Second Collapse: When Real Estate Companies Pretended to Be Tech Companies

iBuying wasn’t the only casualty. A second collapse, walking in from the opposite direction, unfolded over the same five years: the office-sharing company WeWork and the modular-construction startup Katerra.

WeWork’s business was simple enough. Sign long-term leases (usually 10-plus years) with landlords, renovate, chop the space into small units, and sublease them short-term (month-to-month) — a subleasing business, in real-estate terms. But the company called itself a “technology platform that builds community,” and investors applied a software-company valuation multiple — dozens of times revenue — as though it were one.

What that hid was duration mismatch, one of the oldest and best-known risks in real estate. WeWork owed ten years of rent but collected revenue that tenants could cancel any time. In good times, this structure inflates profits like leverage. In a downturn, it works exactly in reverse: the outgoing payments stay fixed while incoming revenue dries up first. When the pandemic shook office demand, the structure collapsed in precisely that direction. Four years after its failed IPO attempt, WeWork filed for bankruptcy in 2023.⁷

Katerra, the modular-construction startup, arrived at the same conclusion by a different road. It raised massive investment from SoftBank but had already collapsed by 2021. WeWork and Katerra share a clear common thread: putting real-estate assets on the balance sheet, then demanding a software-startup valuation for them. Real estate carries heavy leverage and is acutely sensitive to the business cycle. Force that into the venture-capital growth formula — where losses can grow as you scale, as long as you capture market share — and real estate’s physical laws (cash flow, duration, the business cycle) don’t vanish; they just get disguised in startup language. It was a costly lesson that speaking in tech terms doesn’t make the financial statements speak a different language.

The Third Collapse: The Retreat of “We’re Not a Real Estate Company”

The third case unfolded more quietly, but just as surely. The PropTech brokerages that emerged in the mid-2010s all insisted, in one form or another, “we’re a software company, not a real estate brokerage.” Compass and Redfin led the pack.

In practice, it remained a business that still required people walking properties. Showing listings, mediating negotiations, closing deals — these still needed agents, and their salaries and marketing costs made up most of the income statement. Software was a layer laid over this labor-intensive industry; it never replaced the industry itself.

When rates rose and transaction volume dried up, the structure’s inability to cover fixed costs was laid bare. Compass carried out three or more rounds of layoffs between 2022 and 2023, with organizational restructuring continuing afterward.⁸ Redfin, after repeated rounds of restructuring, was ultimately acquired by mortgage giant Rocket Companies for $1.75 billion in 2025, giving up its status as an independent public company.⁹ What’s notable is that in the process, Redfin reverted to a traditional commission-based model. The original narrative of “using technology to break the brokerage-commission structure itself” landed on the far more modest ground of “using better tools within the existing industry.”

Lay the three collapses side by side and a pattern emerges. iBuying believed it could have speed and accuracy at once — and collapsed. WeWork and Katerra believed real-estate assets could be inflated like a startup — and collapsed. Compass and Redfin believed software could replace a labor-intensive industry — and instead of replacing it, ended up merely assisting it. All three stories share the same denominator: real estate is not software. However sophisticated the algorithm layered on top, the physical laws of a real asset — inventory, duration, labor, the business cycle — don’t disappear.

What the Survivors Share: The Non-Owners Won

Companies that came through these same five years relatively unshaken share a clear trait. Most of them didn’t hold assets directly — they sold data, platforms, and workflows.

CoStar Group is the prime example. CoStar sells real-estate market data and analytics; it doesn’t buy or sell buildings itself. It used the downturn as an opportunity, acquiring the 3D digital-twin company Matterport for $1.6 billion in February 2025, snapping up cheapened competitors and adjacent technologies to grow its footprint.¹⁰ The same trend shows up across PropTech as a whole. Through November 2025, PropTech M&A deal count hit 163, already surpassing all of 2024 (134) and closing in on the 2022 peak (170).¹¹ Private equity was involved in roughly a third of these deals — the so-called “vertical SaaS roll-up” strategy, buying up software companies serving a specific workflow niche one by one and stitching them together.

Reading this trend as an industry in decline would be a mistake. It’s closer to a signal of maturation. The industry is moving from an early stage where individual startups each grew their own business, to a consolidation phase where a small number of platforms holding data and workflow assets absorb the rest. A similar direction is visible in Europe, where leasing and asset-management software companies have grown by acquiring regional competitors across the continent. If iBuying maximized risk by “directly taking on inventory,” the survivors were the ones who didn’t take on inventory but instead sold information and tools to the people who traded it. They weren’t as directly exposed to asset-price swings as the market rose and fell, and demand for data itself held steady whether the market was hot or cold.

The 2025 Rebound, and Why It’s Different

Let’s return to the 2025–2026 rebound. The simple fact that PropTech funding is climbing again might raise a worried question: is the boom repeating itself? A slightly closer look at the data shows this rebound has a fundamentally different texture from 2021.

As we saw in Chapter 3, AI adoption in real-estate management jumped from 20% in 2024 to 58% in 2025 — but only 8% of companies had actually “fully automated” even one process.¹² That gap is the key. Adoption exploded, but the change that adoption produces remains far more incremental, far more verifiable. Today’s AI wave doesn’t start from a story about “creating a new asset class” or “disrupting the industry’s structure.” It starts from a far more conservative, far more measurable promise: cutting the labor cost of workflows that already exist — tenant screening, maintenance ticketing, lease-document analysis, market-research reports.

Investors’ posture has changed too. In 2021, the question was “how fast can this company capture the market?” Today it’s “how many percentage points of cost does this tool actually cut?” The scoring criterion itself has shifted from “technical novelty” to “measurable return on investment.” This is an explicit repeal of the 2021-era formula of “burn cash first, capture the market later.” And this shift echoes the lesson left by iBuying’s collapse: in real estate, speed alone can’t substitute for accuracy, and accuracy alone can’t substitute for market liquidity. Surviving requires both — plus verifiable numbers to back them up.

The boundaries of what counts as “PropTech” have widened over these five years too. Early PropTech was a narrow software niche — leasing platforms, brokerage apps. Recently, the definition has expanded to include construction, infrastructure, climate and energy, and industrial IoT. The data-center boom, the energy transition, and the demand for climate-risk data covered in earlier chapters are all being absorbed into “real estate technology” as a new growth axis. Fields once treated as entirely separate industries just five years ago are now being bundled together under the common denominator of “digitizing and optimizing the physical asset that is a building.”

Three Lessons That Will Still Hold in Five Years

Beyond the rise and fall of individual companies in this five-year cycle, there are three structural lessons that should still hold true three years from now, for readers in other countries too.

First, in real estate, “the accuracy of a prediction” and “the execution power to allocate capital based on that prediction” will remain separate. Zillow didn’t fail because its algorithm was wrong — it failed because, in the process of translating an accurate estimate into an actual trade, market liquidity and timing — variables statistical models struggle to handle — got in the way. This gap won’t be easily closed no matter how sophisticated AI becomes, because real estate isn’t an asset that sells whenever you want it to.

Second, “speaking in tech language” doesn’t make real estate’s basic physical laws disappear. WeWork, Katerra, Compass, and Redfin each confirmed this in their own way. Leverage, duration mismatch, and labor intensity operate the same regardless of what a company calls itself. This principle will apply just as equally to whatever new real-estate asset classes emerge going forward, including data centers.

Third, sellers of data, platforms, and workflows absorb the shock of the business cycle better than direct asset owners. This raises a practical question for anyone investing in or working within PropTech: does this business only make money when the market rises, or does demand for information and tools persist no matter which direction the market moves?

Return to that sentence Zillow admitted to.”We’ve determined the price of failing quickly is far less than the price of continuing to lose money.” The real message of that sentence isn’t that AI can’t predict real-estate prices. It’s that knowing a price and having the nerve and the liquidity to bet on that price are different skills. Tokenization, the subject of the next chapter, faces the same question. The technology to slice an asset into fragments already exists. The question is whether a market to actually trade those fragments exists to the degree the promise suggests.

---

Rule of the Game

Speed and accuracy don’t travel together. In real estate, “knowing the price” and “having the nerve and timing to actually buy and sell at that price” are different skills. Those who sell information about an asset, without taking it onto their own books, weather the shocks of the business cycle longer than those who hold the asset itself.

---

Sources

Footnotes

1. Zillow Offers shutdown and 2021 losses/layoffs — GeekWire, “After ditching home-buying business, Zillow Group partners with rival Opendoor”; Serhant, “The End Of Zillow Offers: A Sign Of Things To Come?” ↩

2. Opendoor/Offerpad revenue decline of more than half — compiled from Threads/earlystartupdays reporting; Mike DelPrete, “iBuyer” analysis. ↩

3. Opendoor FTC fine of $62 million — Inman, “Opendoor cutting prices in bid to get surplus inventory off books” and related reporting. ↩

4. 2021 global PropTech VC of $32 billion — CRETI, “2023 Proptech Venture Capital Report.” ↩

5. 2022–2023 PropTech funding decline ($19.75B → $11.38B, down 42%; Q1 2023 down 77% year-over-year) — Multifamily Dive, “Funding for proptech plummets 42%”; Commercial Observer, “Proptech VC Funding Down 77 Percent Annually.” ↩

6. 2025 PropTech funding of $16.7 billion (up 67.9%), AI-focused PropTech up 176% — Bisnow, “The Winter Of Proptech’s Discontent May Be Ending”; PropTechJobs, “PropTech Industry Landscape and Projections (2025-2030).” ↩

7. WeWork’s 2023 bankruptcy filing — general press coverage. ↩

8. Compass’s repeated 2022–2023 layoffs — InterviewPal, “Compass Layoffs 2026: 110 Jobs Cut After Anywhere Merger” and related reporting. ↩

9. Redfin’s repeated restructuring and its 2025 acquisition by Rocket Companies ($1.75 billion) — TechCrunch, “Redfin is laying off more workers as housing downturn persists”; compiled reporting on the Rocket Companies/Redfin acquisition. ↩

10. CoStar Group’s acquisition of Matterport ($1.6 billion, February 2025) — ProptechBuzz, “Costar’s $1.6B acquisition of Matterport.” ↩

11. 2025 PropTech M&A count of 163 (2024: 134, 2022: 170), private equity involved in roughly a third of deals — Levera Partners, “PropTech and Real Estate Software M&A: A Founder’s Guide to the 2025-26 Market”; CRETI, “The Great Proptech Shakeup.” ↩

12. Real-estate management AI adoption from 20% (2024) to 58% (2025), 8% of companies fully automated — MarketScale, “AI and automation fuel a new wave of real estate and property tech investment”; MRI Software, “PropTech trends for 2026.” ↩