If AI Takes the Jobs, Where Do We Put the Savings? Why We're Betting on Real Estate

Anthropic CEO Dario Amodei has warned that artificial intelligence could wipe out as much as half of all entry-level white-collar jobs in the United States. If even part of that scenario comes true, a rather personal question follows: what do we want decades of savings to depend on? Axios

Today, living close to work adds value. Tomorrow it may count for less.

People also choose a property for the shops and services nearby, a park, the rambla (Montevideo's waterfront promenade), the neighbors, and the rent it can bring in. What we want to know is which of those reasons hold up if AI reshapes employment.

The short answer

Our answer is to diversify. At INGAR, we believe a well-bought property can hold a share of your wealth, as long as the net yield, the price and the demand all add up.

According to the INGAR Pocitos Index for July 2026, a typical 67 m² apartment priced at USD 240,000 would generate roughly USD 1,020 gross and USD 660 net per month. Expenses and vacancy would eat up close to 35% of gross income. This is an estimate built from median listing figures, not a closed deal or a promised return.

What AI could change about work

Amodei's warning is a projection. In 2026 it became harder to dismiss: the most advanced models now handle complete assignments end to end. In software benchmarks, they take a task, open tools, edit files, run tests and deliver a result that would cost a human expert hours.

METR measures that shift using software tasks. It calculates how long an expert would take, and how complex a task can get before a model stops completing it at least half the time. Between 2019 and 2025, the frontier models' mark roughly doubled every seven months. For models released since 2024, the doubling time drops to 89 days — barely three months. METR: Time Horizon 1.1

Take it with a grain of salt: these are narrow software tasks and the bar is a 50% success rate. Failing one time out of two doesn't replace a professional. But how fast that mark is climbing is hard to ignore.

What does the picture look like outside the labs? The IMF estimates that about 40% of jobs worldwide are exposed to AI-driven change; exposure doesn't mean elimination. IMF Stanford has yet to find broad job losses in the United States, but it does see a signal among workers aged 22 to 25: in highly exposed occupations, employment ran roughly 19% below less exposed roles. The gap came mainly from reduced hiring, and the study can't prove how much of it is due to AI. Stanford Digital Economy Lab

The ILO sums it up well: no mass displacement so far, though there are real risks for young workers, for inequality and for job quality. ILO empirical review

Nor does the pressure fall evenly across tasks. The ILO finds greater exposure in roles built around cognitive, clerical and digital work. Manual trades and craft occupations sit farther from that direct impact. The OECD reaches a similar conclusion: today's capabilities fall short when a job requires reading a shifting context, making complex judgment calls and owning the consequences. That describes the present, not a permanent exemption for any profession. For an income-producing investment, the link is direct: if jobs and incomes change, so does the ability to pay rent. ILO OECD

The future may bring more productivity, new companies and better jobs. It may also leave fewer entry-level positions and funnel a large share of the gains to a few. You don't have to pick a prophecy to notice that the range of outcomes has widened.

Real estate versus stocks, bonds, gold and crypto

Stocks let you own a piece of the companies that win with AI — and also the ones whose business it destroys.

A broad index fund spreads that risk out.

Bonds, deposits and cash offer more predictable payments or liquidity, in exchange for exposure to inflation, currency swings and the issuer's ability to pay.

Gold and crypto can store or move value, but neither produces income on its own. Real estate is less liquid and ties up more capital, but you can live in it, rent it out and improve it.

We don't think a property should replace everything else. It interests us because its outcome doesn't hinge on guessing which company, currency or technology wins next.

What holds up a property's value

A property can house people or activity, produce rent, absorb improvements and serve as collateral. That usefulness exists whether or not its market price moves.

Location adds something that can't be mass-produced. In Montevideo today, proximity to work, transit and services helps explain prices. If AI and remote work cut down some commutes, being near an office may count for less. Denying that would weaken the argument; the real question is which other reasons for choosing a place survive.

We already ran the experiment

The pandemic gave us an imperfect test. A report by Susan Wachter for Penn IUR estimated that the shift toward remote work was consistent with an 8% to 10% drop in office rents and values, and cited a projected 28% decline for New York offices. Meanwhile, between December 2019 and July 2022, US housing rose 45% in nominal terms. Low rates and stimulus played their part, but demand for places didn't vanish: it moved, and it was especially strong where there were amenities, access to open space or downtowns with street life. Penn Institute for Urban Research

Even if the daily commute matters less, being near other people still answers a human need. In 2025, the WHO reported that one in six people worldwide is affected by loneliness, and linked it to some 871,000 deaths a year. WHO Commission on Social Connection A meta-analysis led by Julianne Holt-Lunstad found a 26% higher likelihood of mortality associated with loneliness, 29% with social isolation and 32% with living alone. Perspectives on Psychological Science

None of that proves a given neighborhood will hold its value. Communities move, habits change, and some areas empty out. Social connection explains why we keep looking for places; it doesn't tell you which one to pick. That's why a location only becomes a defensible investment when it meets concrete conditions:

• the intended use is legally permitted;
• the infrastructure and access are there;
• the location solves a real need;
• there are people or businesses willing to pay for that use;
• the purchase price doesn't already bake in unlikely gains.

More floors can also get built, zoning rules can change, and new transit links can make other areas attractive. And a property can sit empty, need repairs or take months to sell. Just because its price doesn't flash on a screen every day doesn't mean it never falls.

Housing versus stocks: what the record since 1870 shows

Land was shelter, production and a source of income long before there were listed shares. The modern stock market is usually traced back to 1602 and the Dutch East India Company. Oxford Academic

Being older doesn't make it the better investment.

The rigorous comparison starts in 1870. Across 14 advanced economies, real house prices stayed roughly flat until the mid-20th century, then climbed sharply. About 80% of the post-World War II boom is explained by land, not by the cost of rebuilding structures. American Economic Review A sought-after location is the hard part to replicate — though plenty of areas stall or lose activity all the same.

Another study, covering 16 economies from 1870 to 2015, found average real returns of close to 7% a year for both housing and equities. The Rate of Return on Everything But research on actual British properties came up with just 2.3% net, and costs equal to 32% of gross income. The Review of Financial Studies An index pools thousands of properties; a buyer ends up with exactly one. The history can be good and the specific purchase bad.

Returns on a Pocitos apartment: a worked example with INGAR data

Data as of July 2026. The figures in this example are frozen so the math stays reproducible. The linked Index may show a later edition.

Let's run a simple exercise using the INGAR Pocitos Index, updated to July 2026. The index describes the market with a typical 67 m² apartment at a median price of USD 240,000; the estimated yield worked out to 5.1% gross and 3.3% net per year.

The INGAR Index lets you compare apartments across 49 neighborhoods. Pocitos ranks 9th by price per m². That doesn't tell you where to buy, but it keeps you from reading USD 3,548 or a 5.1% yield as numbers floating in a vacuum.

Applied to that price, estimated gross income comes to about USD 1,020 a month. After property management, vacancy, maintenance and taxes, roughly USD 660 a month is left. Nothing dramatic "went wrong": between the headline number and the money in your pocket sit the ordinary costs of owning and renting out a property.

The gap, USD 360 a month, works out to about 35% of gross income. A century earlier and on another continent, the study of British properties found average costs of 32%. The proportion is nearly identical: costs aren't a footnote, they're a structural part of real-estate investing.

A note on method: the published median rent (USD 883) comes from a sample of rental listings that is separate from the sales sample. The yield is estimated using the median price per m² from each of the two universes, which is why you can't get it by dividing USD 883 by USD 240,000.

This isn't a closed transaction or a promised return; it's an example built from median asking figures. A real unit may land better or worse depending on price, condition, common charges and achievable rent. Our guide to real rental yields lets you swap those averages for the numbers on a specific property.

Why we at INGAR are betting on this industry

We don't know which asset will perform best over the next twenty years. We do know that people will keep looking for places to live, meet, be cared for, rest and belong to a community. Technology can change which places they choose; it doesn't make one place interchangeable with another.

That's why we're betting on this industry.

Not because "bricks never go down," but because a good property brings together three things that rarely come as a set: present-day usefulness, the ability to produce income, and a location no one can copy.

There's also the fact that we can act on it. We can study a single block, negotiate the price, upgrade a unit, choose who to rent to and manage expenses better. A small stake in a distant company offers no such control.

Our conviction comes with conditions. We look for locations with demonstrable demand, prices the rent can actually support, and a financial structure that can ride out bad years without a forced sale. When those three line up, we believe real estate is among the best tools for building and preserving wealth over the long run.

We're also betting on the sector because of how it absorbs technology. Search, comparison and a good chunk of the paperwork will keep getting faster and cheaper. Our work will have to earn its keep some other way: using those tools to analyze better, verify on the ground and stand behind a recommendation. That combination only strengthens our conviction about the industry.

None of this forces a choice between property and stocks. A financial portfolio provides liquidity and a way to share in the growth AI itself creates. Real estate does something different: it's a useful, local, finite asset we can act on. This isn't fear of technology; it's refusing to depend on a single version of the future.

How to decide where to invest without predicting the future

You don't need to call the arrival date of an intelligence capable of doing most human tasks, or picture a world without money. Evaluating a property calls for closer-to-home questions:

1. Does the net yield work even if the price never rises?
2. Would this location still be useful if jobs and incomes change?
3. How much capital ends up tied up, and how much stays liquid?
4. How much vacancy, expense and debt can the investment absorb?
5. What currency is the result measured in?
6. What would have to happen for you to admit the purchase was a mistake?

If you're still deciding where to look, our neighborhood comparison by price and yield helps narrow the search down to specific areas.

AI widens the range of possible futures. That's why we want part of our wealth to rest on something that is useful today, can produce income, and lets us influence the outcome. After that comes the hard part: buying well.

Frequently asked questions

Is real estate a good investment if AI reshapes employment?

It can be, when the net yield, the price and the demand all work without counting on future appreciation. The property still carries risks: concentration, vacancy, costs and limited liquidity.

Is real estate safer than stocks?

There's no guaranteed winner. Stocks bring liquidity and diversification; a property concentrates more capital, but you can use it, rent it out and improve it.

What return does a Pocitos apartment deliver?

As of the July 2026 data cut, the typical apartment in the INGAR Index yielded 5.1% gross and 3.3% net per year: about USD 1,020 and USD 660 a month on USD 240,000.

Can AI push property values down?

It can change jobs, incomes and commuting patterns, and with them demand in certain areas. The outcome depends on use, location, infrastructure, price and actual demand.

---

Informational content only. This is not financial, tax or legal advice, nor a recommendation to buy any particular asset. Past performance does not guarantee future results.