U.S. Real Estate Investing Enters Garbage Time: Rates, AI, and Demographics—and the Two Plays That Still Work Over the Next Decade

Let me open with a line that will likely offend many of my peers in the industry: U.S. real estate investing is collectively entering garbage time. This is not an emotional call. It is the conclusion of three independent sets of data pointing in the same direction at once: borrowing costs will not fall back quickly, AI is draining the pool of young buyers who actually set prices, and the engine of U.S. population growth is stalling. For the past 20 years, real estate investors enjoyed three enormous tailwinds—low interest rates, steadily rising prices, and the concentration of population and high-income jobs in major metros. Looking forward from October 2026, all three are weakening, and weakening structurally.

The three tides that lifted housing for two decades are receding together, exposing what they hid.

The roadmap is simple. The first three sections argue the three pressures: first, the 30-year fixed mortgage rate is unlikely to return below 6% within three years, while gross rental yields in most counties already sit below the 10-year Treasury yield; second, AI does not need to lay anyone off—merely hiring less is enough to erode the marginal homebuyer; third, fertility and immigration are falling together, shrinking the pool of buyers a decade out. The final two substantive sections cover the two plays I believe are genuinely worth making over the next ten years: senior housing and modular construction. What they share is independence from price appreciation—they rely instead on a locked-in demographic shift or a cost advantage others cannot replicate.

One boundary up front: I am describing a five-to-ten-year trend, not a forecast that prices fall nationwide next year. Your city may well rise another 5% next year, and mortgage rates may suddenly drop 50 basis points in some month; none of that contradicts this article. The essential difference between real estate and stocks is transaction cost—commissions, excise tax, title, escrow—which means direction can only be judged on a ten-year horizon. The article closes with a framework you can apply yourself.

Rates: Mortgage Rates Are Unlikely to Fall Below 6% Within Three Years

For more than a decade, the reason U.S. housing was such a good investment came down to one word: cheap. You put $200,000 down to control a $1 million asset, the tenant pays your mortgage, the property rises 5% a year—$50,000—but you only put in $200,000, so your return on equity is 25%. That is the power of leverage. Leverage only works when the borrowing rate is below the asset's total return (rental yield plus appreciation). Once borrowing costs more than the asset returns, leverage flips from amplifying gains to amplifying losses.

Leverage only works when borrowing costs less than the asset returns; reversed, it magnifies losses.

The problem today is that both ends are deteriorating at once. According to the Freddie Mac PMMS, the 30-year fixed mortgage rate was 7.40% for the week of October 8, 2026, and remained at that level on October 15. Meanwhile, the Case-Shiller national home price index tracked by the St. Louis Fed's FRED database shows prices in July 2026 up nearly 60% from January 2020 (about 58.8%; the FHFA measure shows 59.3%). The National Association of Realtors reports a September 2026 existing-home median price of $417,500, while the U.S. Census Bureau puts 2025 median household income at $87,500—a ratio of 4.77x, close to five times income. Homes got expensive and borrowing got expensive; the math simply no longer works.

Mortgage rates follow long-term Treasuries, and what long-term Treasuries fear most is nominal GDP growing too fast—as long as the AI capital-expenditure cycle continues, there is no reason for mortgage rates to easily return below 6%.

Nominal Growth and AI Capex Are Pinning Long-Term Yields High

The common plan is "buy now at high rates, refinance when the Fed cuts in two years." The hidden assumption is that mortgage rates will follow the policy rate down. But the anchor for the 30-year mortgage is the 10-year Treasury, which prices the market's expectation of future nominal growth and inflation. According to the Bureau of Economic Analysis's third estimate released September 30, 2026, U.S. nominal GDP grew at an 8.5% annualized rate in the second quarter of 2026; stripping out government and trade, "private domestic final sales"—consumption plus private fixed investment—grew 4.6% in real terms; within that, nonresidential (business) fixed investment grew 9.0% in real terms, total private fixed investment grew 7.7%, and residential investment grew only 2.8%. These figures tell us two things: the economy is expanding quickly in nominal terms, and the engine is AI, data centers, chips, and the power grid—not housing.

Mortgage rates are tethered to long bonds, and the AI capex boom keeps pulling both upward.

This means capital is being pulled toward the other end of the real economy. The gap between 9% growth in business investment and 2.8% in residential investment is direct evidence that capital is flowing out of real estate and into AI infrastructure. As long as this investment cycle continues, expectations for nominal growth will not cool, long-term yields will not come down easily, and mortgage rates will stay pinned high. My call on rates is not based on any single Fed meeting; it is based on the length of this capex cycle, which runs far longer than a rate-cutting cycle.

AI Companies Are Now Competing With the Treasury for the Same Bond Buyers

There is a second driver of high rates: supply. The largest borrower in the bond market used to be the U.S. government. Now Meta, Oracle, Google, and the rest of the AI complex are all borrowing there too. The Dallas Fed's analysis of February 10, 2026, estimates that AI-related investment-grade corporate bond issuance this year is expected to be around $300 billion (the median Wall Street forecast), adding up to the equivalent of $360 billion in 10-year Treasury duration supply—roughly one-eighth of the duration supply from Treasury issuance. Investors have only so much money; to absorb both government and AI-company bonds at once, the only equilibrium is higher yields. The more aggressive AI investment becomes, the harder it is for long-term yields and mortgage rates to fall. This is a transmission channel many real estate professionals have not noticed: AI affects home prices not only through jobs—it first raises your monthly payment through the bond market.

Gross Rental Yields Have Already Fallen Behind Risk-Free Treasuries

High rates are only half the problem; rents have not kept pace. ATTOM's single-family rental report released in March 2026 shows that among 341 major U.S. counties, 54.8% saw three-bedroom single-family gross rental yields decline year over year. Several tech-heavy, high-income counties stand out:

After leaks, rental yield trails Treasuries; the only reason left to buy is faith in appreciation.

Note these are gross yields, before property tax, insurance, maintenance, and vacancy. On the other side, FRED data shows the 10-year Treasury closed at 5.09% on October 16, 2026, and traded around 5.13% intraday on October 19—roughly 5.1%. So you face a surreal choice: buy a house, deal with tenants, leaks, and roof replacements, and earn under 4% gross; or buy Treasuries, do nothing, and earn about 5%. The only reason left to buy is a belief that prices will keep rising. The next two sections ask whether the two pillars of that belief—high-income employment and population growth—still hold.

AI: No Layoffs Needed, Just Fewer Hires

I have been talking about AI's effect on real estate for the past year, and many people do not believe it. Under my earlier video arguing that East Coast home prices would see a major decline, a flood of comments pushed back: look, Boston is still rising, and finance has not seen mass layoffs. This reflects a huge misconception: AI does not need to lay off everyone currently employed. A department that used to need 20 people can now get the work done with 15 plus AI. The boss does not need to fire five people tomorrow; he simply does not backfill when someone retires or leaves, and three years later the department has naturally shrunk from 20 to 15. This mechanism is attrition. It is nearly invisible in the unemployment rate but immediately visible in new hiring.

AI need not fire anyone; by not replacing those who leave, teams shrink on their own.

AI's first impact is not on the employee who has sat in the office for 15 years, but on the young person standing at the door waiting to come in—it is removing the bottom rung of the career ladder.

The evidence comes from research published by the Stanford Digital Economy Lab on August 12, 2026. Using ADP payroll data covering millions of U.S. workers (through June 2026), the study found that 22-to-25-year-olds in the occupations most exposed to AI already show an employment gap of about 19% relative to peers in low-exposure occupations; under the updated methodology, that gap was 15% in July 2025—meaning it widened from 15% to 19% within a year. Critically, experienced workers show no comparable gap. The implication: AI has not yet caused society-wide mass unemployment, but it is precisely squeezing those just entering the workforce.

The connection to real estate runs through a concept I return to constantly: the marginal buyer. According to Census Bureau Vintage 2025 data, the Seattle-Tacoma-Bellevue metro had 4.16 million people as of July 2025, but Seattle home prices are not set by those 4.16 million. They are set by the last small slice of people who both want to buy and can afford to. They are typically around 30, just entering their phase of rapid income growth—software engineers, lawyers, finance professionals, consultants, or young families with a new baby upgrading from an apartment to a single-family home. Prices clear at the margin, and the marginal buyer's income curve caps the clearing price. I worked through this transmission in detail in Quantifying AI's Impact on Seattle Home Prices; here I will stress one conclusion. If AI means a 55-year-old engineer does a little less work, I am not worried at all. But if AI delays a 25-year-old's entry into a high-income career by five years, he does not simply skip buying this year—he skips buying at 30 and buys at 35 instead. Pushing the first-time-buyer age back five years across a cohort is in a completely different league from a one-point rise in the unemployment rate: the latter is cyclical, the former structural.

Prices are set by the marginal buyer, and AI is thinning the cohort who would fill that seat.

This also explains why "Boston is still rising" does not refute the argument. The effects of attrition and deferred hiring take three to five years to show up in housing demand, while today's prices reflect the cohort that entered the workforce three to five years ago. In other words, today's prices are pricing past employment, not future employment.

Demographics: Fertility, Immigration, and "Who Buys Your House in Ten Years?"

For decades, U.S. real estate rested on another assumption treated as common sense: the population would always grow, so someone would always be there to buy your house. That logic stood on two pillars—domestic births and international immigration. Both are now giving way simultaneously.

The fertility decline is the longest-running variable. According to the CDC's report of April 9, 2026, the U.S. general fertility rate in 2025 was 53.1 births per 1,000 women aged 15 to 44, down 23% from 2007. This means every age cohort born after 2007 is nearly a quarter smaller than the generation before—and they will reach first-time-buying age through the 2030s. Fertility is not a variable policy can reverse in a few years, so this curve is essentially already written into the next twenty years of housing demand.

Every cohort born after 2007 is nearly a quarter smaller, a curve already written into future demand.

The immigration decline is faster and closer. Census Bureau data released January 27, 2026, shows that over the July 2024–June 2025 period, net international migration fell from 2.7 million the prior year to 1.3 million, a 53.8% drop; annual population growth fell from 3.2 million to 1.8 million, and the growth rate dropped from 1% to 0.5%. And it may not be over: the Brookings Institution estimated in January 2026 that net migration in 2025 may have fallen to between negative 295,000 and negative 10,000, which, if ultimately confirmed, would be the first negative reading in at least half a century (note that the Congressional Budget Office's estimate for the same period remained around positive 400,000, so the qualifier "if confirmed" must stay). I analyzed this specifically in What Negative U.S. Net Migration Means for the Seattle Housing Market: cities like Seattle, which depend heavily on international migration and H-1B talent inflows, are far more sensitive to this variable than the national average.

Young people are finding it harder to enter high-income careers, first-time buying keeps getting later, fertility keeps falling, immigration keeps shrinking—ten years from now, who buys your house? And why would they pay more than you did today?

Putting the three sections together, the three tailwinds of the past 20 years are receding in sync:

Once more on time horizon: real estate is not a stock you can sell with a click when you stop liking it. Its transaction costs mean what investors truly need to judge is not whether next year brings +3% or -3%, but the direction of the next decade. And none of the three directional calls is optimistic.

Housing's high transaction costs mean the call is the decade-long deep current, not next year's ripples.

Play One: Senior Housing—The Most Certain Demographic Dividend

So far I have argued that young-buyer demand is shrinking. But there is another group in America whose numbers are growing fast and who hold enormous wealth: the baby boomers. According to Pew Research in January 2026, the first boomers, born in 1946, turn 80 in 2026. According to the Federal Reserve's Distributional Financial Accounts, boomer household net worth stood at about $97.4 trillion in the second quarter of 2026, or 52% of total U.S. household net worth of $185.7 trillion. The largest and wealthiest generation in American history is collectively entering the age at which demand for senior housing rises sharply.

The core change in real estate ahead is not that people suddenly stop needing housing—it is that housing demand is being reallocated from 30-year-olds to 80-year-olds.

The demand side has already materialized. According to NIC data for the third quarter of 2026 (reported by Senior Housing News on October 1, 2026), senior housing occupancy across the 31 primary markets it tracks reached 90.4%, rising for 21 consecutive quarters; independent living stood at 91.7% and assisted living at 89.1%. The supply side, however, has barely moved: senior housing inventory grew only 0.4% year over year, annual new development is only about 10,445 units, and NIC estimates that meeting total senior housing demand will require more than 100,000 new units per year into the 2030s.

Senior housing occupancy has climbed for twenty-one quarters while supply barely moves; demand waits outside.

A 30-year-old and an 80-year-old want completely different things from a home. The young buyer cares about commute and school district; the senior cares about elevators, whether someone will help after a fall, how easy meals are, access to medical care and social life, and whether the children can visit easily. This means senior housing cash flow does not depend on price appreciation—it depends on a population that has already been born, already aged, and already accumulated wealth. That makes it a rare real estate category where the demand side is fully certain. I first made this call in The Two Biggest Opportunities in U.S. Real Estate for 2026; today's NIC data only makes it firmer. For investors, the transmission is straightforward: occupancy above 90% with inventory growth under 0.5% gives operators sustained pricing power—exactly what other residential categories are losing.

Play Two: Modular Construction and the Shrinking Supply Window

Many viewers find a contradiction in my videos: Li Laosi, you keep saying real estate is in garbage time, so why are you still building? The answer is that my cost structure differs from traditional developers'. In a market where neither prices nor rents are rising much, profit can only come from two places: lower costs than everyone else, and faster speed than everyone else.

In a flat market, profit comes from only two places: lower cost and faster speed.

Garbage time does not mean no real estate can make money; it means the era of profiting from market appreciation is over, and from here on you have to create the profit yourself.

Take our own projects. Hard costs for traditional stick-frame apartments in the Seattle area run about $300 per square foot—Terrapin Construction Group's 2026 contractor guide puts the published range for Seattle wood-frame multifamily hard costs at $295 to $395 per square foot, so $300 sits at the low end—while with modular construction I need only $200 per square foot, one-third cheaper. On speed, the factory produces a roughly 100-unit apartment building in about four weeks, and factory production runs in parallel with site work: while the site is being excavated and foundations poured, the factory is already producing modules, which are then trucked in, craned into place, connected to utilities, and finished. The total construction cycle is about one year versus roughly two for conventional methods—consistent with McKinsey's 2019 estimate that modular can shorten schedules by 20% to 50%, with our projects at the upper end of that range.

What does saving a year mean? A year of construction loan interest, a year of developer overhead, a year of land carry cost—and most importantly, rent starts a year earlier. In a 7.4% rate environment, time itself is the largest cost line, and modular cuts that line in half. Where others must charge $3,000 in rent to make money, I may survive at $2,500. At that point I need to bet neither on 10% annual appreciation nor on rates falling next year.

At high rates, time itself is the largest cost; halving the schedule halves the interest bill.

The West Coast is also presenting a rare window: the apartment pipeline is visibly contracting. According to Kidder Mathews's third-quarter 2026 report (CoStar data), Seattle delivered only 6,906 apartment units in the first three quarters of 2026, down 38.4% from the same period last year, while net absorption reached 9,320 units—the market is absorbing apartments faster than new supply arrives. An honest caveat: net absorption of 9,320 units is itself down 20.9% from 11,782 a year earlier, so this is not demand rising; it is supply shrinking faster than demand. I laid out how this supply gap formed in Single-Family Is Stuck, Multifamily Has Bottomed; it is pushing multifamily into a phase with genuine rent support.

At the same time, total housing need has not disappeared. Freddie Mac, based on third-quarter 2024 data, estimated a U.S. housing supply shortfall of 3.7 million units (published November 2024 and not updated since); the Washington State Department of Commerce estimates the state will need about 1.1 million additional homes by 2044. So the most surreal feature of U.S. real estate today is this: on one side, a shortage of homes; on the other, developers afraid to build, because at today's land, labor, material, and interest costs, many projects simply do not pencil. The real question for real estate going forward is not whether prices rise, but whether, at today's costs, you can build homes ordinary people can afford. Whoever can do that captures the profit of an era that lacks housing but lacks anyone willing to build it.

Counterarguments and Responses

Objection one: "Mortgage rates have already peaked—you wrote about the peak signal yourself—and once cuts arrive, housing will be a good investment again." This is the challenge I hear most and the one most worth answering seriously. I did analyze signals that mortgage rates may have peaked in Bessent's Long-Bond Buyback Confirms the Call. But "peaked" and "back below 6%" are two entirely different claims. A peak says rates are unlikely to rise much further. This article's claim is that with nominal GDP growing 8.5% annualized, business fixed investment growing 9%, and roughly $300 billion of AI-related bond issuance expected, the 10-year Treasury is unlikely to fall from 5.1% to a level that pushes mortgage rates below 6%. A move from 7.4% to 6.5% still leaves a property yielding 3% to 4% gross behind Treasuries; leverage only works again when rates truly fall below the asset's total return. So the peak changes the direction of rates, not the arithmetic of the investment.

A peak in rates is not a return to low rates; beyond the summit may lie a long plateau, not a descent.

Objection two: "Boston is still rising, East Coast finance hasn't seen mass layoffs, so the AI-hurts-housing thesis is disproven." This conflates today's prices with tomorrow's demand. The Stanford research shows clearly that AI's employment impact is concentrated among 22-to-25-year-olds just entering the workforce, with no gap for experienced workers. The people buying in Boston today largely entered the workforce three to five years ago and have accumulated down payments and income; what AI is compressing is the cohort that would otherwise appear on the buyer's side three to five years from now. Attrition and deferred hiring take time to transmit, and prices respond to employment with a natural lag. As an aside, I have placed my optimistic, base, and pessimistic scenarios for Boston on the site's Housing Map, so readers can test this article's logic against them.

Objection three: "The immigration drop is one administration's policy; it will reverse with the next election and can't be treated as a long-term trend." Immigration policy is indeed cyclical, and Brookings's "net migration turned negative" estimate still awaits final confirmation—I preserved those uncertainties in the text. But note that this article's demographic argument rests on two pillars, and the fertility pillar is immune to the policy cycle: a 23% decline from 2007 to 2025 is a curve nearly twenty years long, and it has already shrunk the base of people reaching first-time-buying age in the 2030s by close to a quarter. Even if immigration fully recovered to 2.7 million, it would only restore "1% population growth"—it would not change the fact that the domestic homebuying population is contracting.

Objection four: "The 5% on Treasuries is nominal; housing offers leverage and an inflation hedge, so they can't be compared directly." That argument holds when rates are below asset returns, but today the opposite is true. When you borrow at 7.4% to buy an asset yielding 3.1% to 4.5% gross, leverage amplifies a negative spread, not a positive one; only a high enough appreciation rate fills the hole. The inflation hedge presumes rents and prices keep pace with inflation, yet ATTOM's data shows gross rental yields falling in 54.8% of counties, and the price-to-income ratio is already near 5x. In other words, neither leverage nor the inflation hedge is a free amulet; both work only when the spread between returns and rates is positive.

Leverage only magnifies the spread; borrowing above the yield amplifies the loss, not the gain.

The Framework: How to Test Whether a Real Estate Investment Still Makes Sense

Distilled, this article's method is a five-question checklist. For any real estate project, ask yourself these in order:

1. Is the spread positive? Take the project's gross rental yield, subtract all operating costs, and compare the result to the current 30-year mortgage rate (Freddie Mac PMMS) and the 10-year Treasury yield (FRED). If net yield is below your borrowing rate, the project depends on appreciation, and you must know exactly what you are betting on.

2. Who is the marginal buyer behind the appreciation? Do not look at total metro population; look at who will be on the buyer's side three to five years from now. Is their industry high- or low-exposure to AI? Use employment data broken out by age and occupational exposure, like Stanford's, rather than the headline unemployment rate, to judge that cohort's income curve.

3. Is the pool of buyers ten years out growing or shrinking? Check the region's dependence on net migration and its domestic birth trends (Census Bureau, CDC). Cities heavily dependent on international migration are far more sensitive to variables like a 53.8% drop in net migration than the national average.

4. Does the project's demand come from a demographic shift that has already happened? Senior housing demand comes from a generation already born, already aging, and already holding $97.4 trillion; school-district demand comes from a shrinking cohort of young families. The more the demand has "already occurred," the more reliable the call.

5. Do you have a cost or speed advantage others lack? In a market where neither prices nor rents are rising much, profit can only come from hard costs below the market average ($200 per square foot versus $300) or a shorter schedule (one year versus two). If your costs are the same as everyone else's, you can only earn what the market hands out—and the market is ceasing to hand it out.

In the past, money was made on the big market trend: rates kept falling, population kept growing, prices kept rising, and all you had to do was buy—the market made the money for you. In the future, those who actually profit will either capture a very clearly defined demographic shift, like senior housing, or hold a cost and efficiency edge others cannot match, like modular. Your answers to these five questions determine whether you are investing on the momentum of the former era or have already stepped into the latter.

When the market stops carrying you, only a certain demographic shift or a self-made cost edge moves you up.