WSJ's Take On the Week Sep 20, 2026 35m 13m saved
With Jon Petersen, equity analyst at Jefferies
Real estate investment trusts have underperformed four years running, and the Jefferies analyst who covers them says that is the reason they are not expensive now.
The reflex when the Federal Reserve raises rates is to sell REITs, which is roughly what happened in 2022. Jon Petersen's argument is that this hike lands on a sector where higher-for-longer rates have been the base case for years and where almost nobody has started a new building.
"Once the dot-com bubble burst, REITs outperformed the broader market on a total return basis for seven straight years in a row because everybody kind of stepped back from tech for a long period of time and real estate felt like the safe place to be."
Jon Petersen, equity analyst at Jefferies, on WSJ's Take On the Week, covers the whole REIT sector: senior housing, warehouses, malls and data centers. He is also the firm's data center analyst, which is why the interview runs from assisted living demographics to where an inference token physically lands.
The full episode is covered here so you can skip it.
Here are the 11 calls that matter.
Key Takeaways
Real estate is where listeners said they already invest, and the hosts read it as the non-AI ballast in a portfolio
REITs have underperformed four years in a row, which is Petersen's reason for thinking the sector is not overvalued going into this hiking cycle
The number that matters for property is the 10-year Treasury, now around 5% against the mid-4s the market had settled into
Nobody is building, and that is the sector's defense: REIT earnings growth of about 7% a year against a 4–5% long-term average
Senior housing is growing fastest, with Welltower and Ventas seeing same-property net operating income growth of 15–20%
This is the year the first baby boomers turn 80, and that cohort grows above 5% a year for a decade against a 2% long-term average
Warehouse rents signed in the 2021–22 boom are about to reset downward on five-year leases, while data-center supply chains have become a new source of warehouse demand
Inference, not training, is what fills the listed data-center REITs — the enterprise data sits in Northern Virginia, not North Dakota
The hyperscaler leases rolling off today were seven-year deals; the ones being signed now run 10, 15 and even 20 years
If the AI trade breaks, Petersen thinks real estate is a healthy place to be, and REIT leverage is the lowest it has been in a long time
A 3.5% dividend yield plus 7% earnings growth is a 10.5% total return, which he says beats the S&P 500's long-run number
1. Why Listeners Asked
The episode exists because the audience asked for it. The hosts had surveyed listeners on how they invest, offering individual stocks and ETFs as examples, and real estate came back more often than they expected. Many people's largest single investment is the house they live in, and beyond that the hosts listed multifamily housing syndicates and direct commercial property.
The appeal they described is defensive. One host summed it up in three words: "Heavy asset, low obsolescence." He then put it another way — "The non-AI stuff in your portfolio." The other host framed it as an inflation hedge and a version of the halo trade Josh Brown had described on an earlier episode.
They then explained the vehicle for anyone listening who had not met it. A REIT owns property such as offices, malls or data centers, and gets a tax break for passing most of its income through to shareholders as dividends, which is why REIT dividends are among the highest in the market. The value moves on both the property and the payout. For listeners who do not want to pick, there is a REIT ETF that holds a broad basket.
2. The AI Doomer Week
Before the guest, the hosts covered the week's other story: an Anthropic researcher who left, saying the safety work on the newest models was not good enough and that the pace was too fast. That set off a wider argument about whether model development can run away from the companies doing it.
The hosts gave the cynical reading airtime as well. Some people think the slow-down talk is convenient — a way to explain away a delayed IPO or thin near-term profitability, and a signal to competitors to stop spending on the next frontier.
Through a markets lens, the hosts said the news was that Anthropic and OpenAI are now openly discussing a deliberate slowdown, and the question is what that does to their own listings and to AI exposure across the market. The reaction was small: some correction in chip stocks in the days after, a rally in software names that had been priced as AI roadkill, and neither move outsized. By the end of the week, the Federal Reserve had overshadowed all of it.
3. The 12-0 Hike
The Fed raised rates by a quarter point, which the hosts said was already priced in and is probably not the last hike of the cycle. Officials have penciled another increase into their projections, and the vote was unanimous at 12-0.
The market's first move was to buy longer-dated Treasurys, pushing the 10-year yield down. As Kevin Warsh spoke to reporters and sounded hawkish about his own inflation credentials, yields went back up, and by the next morning they had drifted down again.
One host read the decision as a relief: had the Fed not moved, the fear that it does not know how to handle this, or that it is too political given the president's pressure for lower rates, would have kept building. The other offered a historical parallel — Alan Greenspan was appointed by a president who wanted lower rates, and his first decision was a hike. Warsh has said he wants to say less and map the path out less, which is a Greenspan-like posture, so a first move that was also a hike fits.
The variable they both kept returning to was oil. Gasoline prices matter to voters heading into midterms, and one host said the tightest correlation right now is between bond yields and crude, which makes whether oil stays above 100 the most important near-term input. He also used the day's move as a caution: the Fed hiked and long-term yields fell, because what the market is pricing is whether the hikes eventually slow the economy enough to bring rates back down without causing a recession.
4. Not 2022 Again
The first question to Petersen was the one in the episode title: when the Fed last raised rates in 2022, commercial real estate cratered and arguably never recovered. Is that about to happen again?
His answer starts with where the sector was standing in each case. In 2022 it was standing on a very good year.
2021 was a recovery year, and the hike killed the momentum
Back in 2022, we'd actually come off a year that was incredibly strong for the REIT sector. So during the pandemic, maybe go back a couple years before that, in 2020, REITs had gotten hit particularly hard with retail stores being closed, offices being closed. 2021 was the year of huge recovery. There was inflation in the economy, rates were pretty low. People were feeling really good about the REIT sector, and then the Fed came in and raised rates and it kind of killed the momentum.
Jon Petersen
This time the sector is standing on four years of losing to the market, which is what makes his case.
Four years of underperformance is the reason the sector is not expensive
What I'd say though is the REIT sector has underperformed the last four years in a row. I don't think we're, as a sector, very overvalued at these levels. And so to a degree, I would say higher interest rates or higher for longer interest rates have been the base case investor scenario in the REIT sector for the past few years. So I don't think we'll see a repeat of what we saw in 2022.
Jon Petersen
5. Cap Rates Are Bond Math
Asked for the plain explanation of how rates reach property values, Petersen made two points. The first is which rate matters: property is financed long, so the 10-year Treasury is the reference, not the Fed's own short-term rate.
The sector watches the 10-year, not the policy rate
In the REIT sector, people are looking at the long-term interest rate. Generally with real estate, the 10-year Treasury is what we're looking at is kind of the most important metric. And the reason why is real estate tends to be financed on a long-term basis with longer term debt, so there isn't a lot of exposure to short-term interest rates for the REIT sector.
Jon Petersen
The second is the mechanism. A cap rate is a building's net operating income divided by its price — the yield a buyer gets. When borrowing costs rise, buyers demand a higher yield, and a higher yield on the same income is a lower price.
A building prices like a bond, and the arithmetic runs the same way
And if interest rates go up, then your cost of debt on financing that building would go up. And therefore, if you're going to buy the building, you would want to buy it at a higher cap rate, which is going to bring values down. Think of it very bond-like, right? Cap rates go up, values go down. Same with bonds and yields.
Jon Petersen
Asked what long rates have actually done, he described a straight line up, and flagged the 5% level as a psychological one rather than a mathematical one.
5% may be a mental block for buyers who had accepted the mid-4s
It's been kind of a straight upward trajectory, particularly in the past few weeks. Now that we're around 5%, for the last couple years, we've kind of settled out in the mid-fours, and I think people have got to a point where they're kind of comfortable that maybe mid-fours is where we're going to be now that we're hitting 5%. I don't know if that creates a new sort of mental block for the community that drives a little bit of underperformance here in the near term.
Jon Petersen
6. Supply Is The Killer
Petersen's core argument is not about rates at all. It is that high rates already did their damage on the supply side, and that damage is now a support.
Rents fall when new buildings outrun demand, and that is what breaks the sector
One thing that always kills real estate is supply growth. If there's a lot of supply growth and that supply growth outweighs demand, then rents tend to decline and things can be kind of weak.
Jon Petersen
The building stopped, which is why the sector enters this cycle healthy
We sort of went through a softening of the cycle 2022, 2023, 2024 on higher supply, but now we've gotten to the other side where supply growth is very limited because it just hasn't made economic sense to start new real estate projects over the past few years.
Jon Petersen
That shows up as earnings growth well above the sector's own history, which is his answer to the higher interest cost.
7% earnings growth against a 4–5% long-term average
So across the entire REIT sector, our expectation for earnings growth over the next few years is about 7% a year. If you go back, the long-term average in the REIT sector is about 4% to 5%. So we are seeing elevated earnings growth right now that can offset the higher interest costs.
Jon Petersen
7. The 80-Plus Trade
Asked which sub-sectors have the wind behind them, Petersen named one that surprises people.
Senior housing is the fastest-growing REIT sub-sector
So the one that's growing the fastest right now is senior housing, which may surprise people.
Jon Petersen
The demographics are not a secret, and he said so — anyone can look them up on the census website. What matters is the threshold.
The first baby boomers turn 80 this year, which is the age assisted living starts
This is the year when the first baby boomers are turning 80, that 80 plus demographic is what we follow for senior housing. That's when people tend to move into assisted living.
Jon Petersen
That cohort grows above 5% a year against a 2% long-term average
And so, the expectation for population growth of that demographic is expected to be above 5% for the next 10 years. The long-term average was about 2%.
Jon Petersen
The supply story is the same as everywhere else, with a sharper turn. Developers overbuilt senior housing through the 2010s precisely because the demographics were obvious, private equity financed a lot of it, and then the pandemic hit the category about as hard as a category can be hit. Construction stopped. One host called it shocking that nobody built into the silver tsunami; Petersen's point was that they had, too early.
Construction has not restarted, because rents are still below the level that justifies it
And now we're in an environment where everybody knows, people that are in the industry know that the fundamentals are getting better, but there still is a gap between where rents are and where they need to be to justify new construction.
Jon Petersen
What that gap produces in the meantime is pricing power for the buildings that already exist.
Welltower and Ventas are growing same-property NOI at 15–20%
But the companies like the publicly-traded companies like Welltower and Ventas that invest in these senior housing facilities, they're seeing same property net operating income growth in the 15% to 20% range
Jon Petersen
For a sector this dull, that is an extraordinary number
We get excited about high single digits earnings growth and they're doing high teens.
Jon Petersen
8. Warehouses And Malls
Asked whether the pandemic trades are still running, meaning empty offices, dead malls and booming warehouses, Petersen said the effects linger but the direction has turned in places.
Warehouses had the best of it. Everyone shopped from home, goods had to be stored, supply-chain problems pushed more investment into logistics, and rents rose hard in 2021 and 2022 before new supply arrived.
Those boom-year rents are about to be marked back to market
In the logistics sector, you typically sign five-year leases. So if rents were really strong in 2021 and 2022, maybe you're getting ahead of me, but you add five to those numbers over the next few years, we're actually going to be resetting some of those rents.
Jon Petersen
Prologis and STAG still have tailwinds, but the growth moderates from here
There still are some tailwinds that is keeping companies like Prologis and STAG Industrial going right now, but it's going to start to moderate over the next few years as they start to roll those rents over.
Jon Petersen
The new source of warehouse demand is not e-commerce at all. It is the physical supply chain behind data centers.
Data-center construction became a warehouse story this year
It sort of came out of nowhere. I suppose as the data center analyst, I should have seen it coming, but in the first half of this year, there was a huge wave of demand for new logistics real estate related to data center supply chain.
Jon Petersen
He named the tenants: Schneider Electric leasing warehouse space to manufacture switchgear, including a large lease in Nashville this year; Corning needing somewhere to keep fiber optic cable.
The hyperscalers need staging space beside their own campuses
And then you have the hyperscalers themselves that have been looking for logistics capacity near their big campuses where they're building data centers because they need places to stage equipment, make sure the GPUs are all ready to go before they actually take them to the data center and plug in.
Jon Petersen
He also noted the geography has shifted with US manufacturing investment — away from the coastal population centers that e-commerce favored and toward the middle of the country, a trend he traced to the Biden administration's stimulus bills and said continues now.
On malls, which the Journal had written about that week, his answer was the same one word.
Nobody has built a mall in years, and the stock shrank
But in the mall sector, seven, eight years ago, we sort of woke up and realized that we had more department stores that we needed. We had more retail square footage than we needed, and we stopped building new malls.
Jon Petersen
He watched it happen to his own first employer
The mall that I worked at, I worked at Sears when I was a teenager in Colorado. The mall that I worked at is no longer there. I believe there's a Costco and maybe some condos on top of it.
Jon Petersen
Private buyers cleared the weak stock, which left the good stock scarce
There were some smart private investors out there that raised funds and went and bought out a lot of these old malls, milked them for cash flow for a number of years, and then tore them down and redeveloped them into something else. And so what we've been left with is the class A malls that have the tenants that people want to go see.
Jon Petersen
Growing retailers are running out of places to go, which is pushing rents and values up. Asked who owns the survivors, he named Simon Property Group as the dominant class A owner, with Macerich the other large one.
9. The Tier 1 Data Center
The hosts put a debate to him: when you buy a data center REIT, are you buying the gigawatt-scale AI campuses in the news? He said no, and then explained why that is not a disappointment.
The frontier training campuses are not what the listed REITs own
If you're looking for a gigawatt scale data center that's training the next frontier model, that's probably happening in Louisiana or North Dakota or West Texas or some of those kind of places.
Jon Petersen
Your prompt still travels through the markets the REITs do own
But think about it this way. When you go onto ChatGPT and you prompt it and you ask it a question and you send it through, it has to go through the internet to get to that data center in North Dakota. That's going to take you through the major data center markets in the United States. That's where Digital Realty and Equinix primarily own their data centers.
Jon Petersen
He listed those markets: Northern Virginia, Atlanta, Chicago, Dallas, Phoenix, Portland and Silicon Valley — which is also where the AWS and Microsoft Azure cloud availability zones sit, and where corporate data lives.
As spending shifts from training to inference, the tier 1 markets get the demand
So as we move, training will always exist to some degree. The frontier models will continue to push the frontier, but inference is going to become a lot more important. And as more tokens are used for inference instead of training, that's going to drive, I think, more of this demand into the major tier 1 data center markets because that's where the enterprise data is, that's where the people are.
Jon Petersen
His analogy for why proximity matters was the early web, when a slow page meant you went to a different site. AI today is slow in the same way, and he expects users to behave the same way.
Latency becomes a switching decision between models
And we're going to get to a point that if Gemini is running faster than ChatGPT, people are going to get annoyed with ChatGPT and they're going to start using Gemini. And if that starts to slow down, maybe they'll go try to use Anthropic.
Jon Petersen
Asked whether those saturated markets can even take more capacity, he gave both sides. Digital Realty holds large land banks bought years ago in exactly those markets, and as transmission capacity comes online that land becomes developable.
He was wrong about the land price at the time
They bought land that years ago when they bought it, I criticized them for overpaying for that land.
Jon Petersen
If nothing new can be built, the existing buildings are worth more
The other point is kind of the opposite where, well, if nobody can build new data centers, then that just makes the value of existing data centers that much higher.
Jon Petersen
The payoff he expects is in renewals. Digital Realty signed leases with social media and cloud tenants at the bottom of the market in the late 2010s, and those come due at the end of the decade.
The renewals mark up, which is the mirror image of the warehouse problem
So companies like Digital Realty, we think are actually very well positioned over the next few years to push rents quite a bit higher on the leases that are coming due.
Jon Petersen
A host pressed the obvious counter: today's AI leases become tomorrow's reset. Petersen agreed and then gave the term structure, which is the part that decides when the bill arrives.
Seven-year leases are rolling off; today's are 10, 15 and 20 years
And one thing I would point out is a lot of the leases that are rolling off today that were signed with hyperscalers were seven-year leases. So you have to go back seven years ago. Today, hyperscalers are signing leases that are more like 10, 15, even 20 years.
Jon Petersen
The reckoning is real and it is 15 years out
So you're absolutely right. There is going to be a day of reckoning where the leases that are signed today are going to be marked back down to market, which will probably be lower in the future, but that's 15 years from now.
Jon Petersen
Asked how data-center stocks have traded through the AI rally, he said 2023 and 2024 were strong because the REITs were one of the few ways to express the AI infrastructure trade, and that it has cooled since — because the competition for that attention arrived.
The former Bitcoin miners took the excitement
So companies like Hut 8 or Core Scientific or TeraWulf, these companies were all Bitcoin miners in the past and are kind of pivoting entirely away from that business.
Jon Petersen
Those are not REITs, and a host drew the distinction out for listeners. They trade differently, and Petersen described a different shareholder base: heavy retail participation, TMT investors, utilities investors.
Dedicated REIT buyers and generalists want different returns from the same theme
So if I were to talk to a REIT dedicated investor, they're very excited about Digital Realty and Equinix. They underwrite the business. They say this is going from a high single digit earnings growth business to a kind of low to mid-teens earning growth business.
Jon Petersen
Which is why the boring version struggles for attention
And so it's kind of hard to get people to focus on the data center REITs because there's kind of these sexier data center companies that are out there.
Jon Petersen
10. The Dot-Com Precedent
The episode's title question came back at the end: are REITs actually defensive, or have they been pulled into the AI trade too?
Petersen's answer was a precedent. Four years of underperformance, he said, is partly rates and partly that the market has been focused on technology's growth. That has happened before.
REITs lost through the dot-com run-up, then beat the market for seven years
In the dot-com run up, REITs underperformed. Once the dot-com bubble burst, REITs outperformed the broader market on a total return basis for seven straight years in a row because everybody kind of stepped back from tech for a long period of time and real estate felt like the safe place to be.
Jon Petersen
He said the rotation is already visible in miniature, week to week, whenever the AI trade pulls back — and that REITs were outperforming through the first half of the year, until roughly a month ago when the rate news turned against them.
If this is a bubble and it bursts, property is a healthy place to be
So I do think when we get to the other side of this, and if we are in an AI bubble and it starts to burst, real estate probably is a healthy place to be.
Jon Petersen
He raised the obvious objection himself: the seven good years ended in the financial crisis, and real estate was the crisis. His answer is that the balance sheets are not the same.
Leverage is the lowest it has been in a long time
Balance sheets were over levered at that time period. Debt maturities were not stacked correctly. REITs are at their lowest leverage levels that they've been in a very long time.
Jon Petersen
Which is why he would rather own them into a recession than not
If we had another recession, I think that REITs are very well positioned here.
Jon Petersen
11. Dividend Plus Growth
Asked about the investors who buy REITs for income rather than price, Petersen gave the current yield and said it is lower than rates would justify — and then said the yield is not why most people own them.
The average REIT yields about 3.5%, covered by cash flow
So dividends today are fairly well covered by cash flow. The average dividend yield of a REIT today is about 3.5%. That is, I think relative to interest rates, a bit lower than maybe it should be.
Jon Petersen
The income case is really a growth case played out over a decade
And over time, what you hope is you buy a REIT today at a 3.5% dividend yield. Hopefully it has pretty strong growth over the next 5, 10 years. And that 3.5% yield can turn into a 7% yield over time as they increase dividends.
Jon Petersen
Put together with the earnings growth, that is his total return math.
7% growth plus a 3.5% yield is about 10.5%, ahead of the S&P 500's long-run return
So if you can get 7% growth and a 3.5% dividend yield, that should be about a 10.5% total return, assuming multiples stay the same, which I think for most investors should be a pretty healthy level. Long term, that's ahead of where the S&P 500 has returned.
Jon Petersen
The last question was for index investors: is it worth going down to the sub-sector level rather than buying a REIT index? He started by describing his own job, which is really several jobs.
One analyst, four unrelated industries
My title is I am a real estate analyst, but I wake up in the morning and I talk to people about retail trends and then I talk to people about senior demographics and then I talk to people about the data center industry and I'm just bouncing back and forth between all sorts of different sectors.
Jon Petersen
Follow the sub-sector with the best earnings growth and you do better
So to the premise of your question, if you spend time in the REIT sector and pay attention to which sub-sectors are seeing the best earnings growth, you're going to get better returns over time likely.
Jon Petersen
But picking individual names inside a sub-sector is probably not worth the work
If you're keyed into those trends, maybe one thing I would say is being a stock picker within those sectors can maybe that's a little more effort than it's worth.
Jon Petersen
His practical suggestion was to buy the basket — notice that logistics is working, then own the logistics names rather than choosing between them.
Bonus Insights
The hosts translated the jargon as it came up
Twice they stopped to define terms for listeners: NOI as net operating income, supply as new buildings being built, and GFC as the 2008 financial crisis. One host also pointed out that a REIT sits between a bond and a stock — dividend income plus the chance the property value rises, which together are the total return.
Nobody's models go out that far
When Petersen said today's data-center leases reset 15 years from now, one host asked whether their discounted cash flow models even reach that far, the other said he did not know what he was having for lunch, and Petersen said he might be living in a Welltower senior living facility by then.
The office-to-apartment conversion trade is a local one
After the sign-off, the conversation kept going on whether office REITs are a buy because the buildings become apartments. Petersen's view was that it works in some markets, and he named New York and San Francisco, but not as a general thesis. The studio they were sitting in, one host noted, has no window to the outside.
Dead malls as a genre
One host raised the online subculture of people sneaking into abandoned malls to film them, which Petersen took as the cue for how the private funds bought those malls, ran them for cash flow and then redeveloped the sites.
Petersen's bottom line is that the hike arrives at a sector that has already taken its punishment: four years of underperformance, no new construction, the lowest leverage in a long time, and earnings growth running above its own history — which is why he thinks a burst AI trade sends money back into property the way a burst dot-com trade did.
Products, Companies & Tools Mentioned
Welltower and Ventas (The listed senior housing owners; Petersen says they are growing same-property net operating income at 15% to 20% a year)
Digital Realty and Equinix (The two data-center REITs that own the tier 1 markets where enterprise data sits; he expects Digital Realty to mark rents up sharply on late-2010s leases coming due)
Prologis and STAG Industrial (The logistics owners still carrying tailwinds, with rent growth he expects to moderate as 2021–22 leases roll)
Simon Property Group and Macerich (The two big mall REITs; Simon dominates class A malls, which are the stock that survived)
Hut 8, Core Scientific and TeraWulf (Former Bitcoin miners pivoting into AI data centers; not REITs, far more volatile, and taking the attention the data-center REITs used to get)
Schneider Electric and Corning (Named as the new warehouse tenants: switchgear manufacturing, including a large Nashville lease this year, and fiber optic cable storage)
ChatGPT, Gemini and Anthropic (His latency example: users will switch models the way they once switched websites that loaded slowly)
Amazon Web Services and Microsoft Azure (Their cloud availability zones are in the same tier 1 markets the data-center REITs own)
Jefferies (His firm; its REIT coverage spans retail, senior housing and data centers, and its sector earnings growth forecast is about 7% a year)
Listen to the full episode
🟣 Apple Podcasts | 🔗 Episode page
Listen to the full episode:
More Jon Petersen interviews and podcast appearances
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: