Two numbers came out of July, and they tell very different stories. The market grew 78 percent compared with a year ago. But take out one category, and that growth shrinks to 1 percent. Only one of those numbers describes the deals your team is actually working on.
Here is what happened. MSCI's monthly Capital Trends report, published this week and written up by Bisnow on August 25, showed $74.4 billion of U.S. commercial real estate sales in July. That is the strongest July since 2005, better even than the 2022 peak, and up 78 percent from last July. It is a big number. The catch is in the very next line: data centers alone made up $33.8 billion of it. Once you set that one sector aside, the market was up just 1 percent, and prices were flat compared with a year ago.
So if your team spends its time in apartments and industrial, the headline you are about to see everywhere does not really describe your week.
Where the growth really came from
The data center number is not a small quirk in the data. Sales in that sector were up more than 1,900 percent from a year ago. And large portfolio and entity-level deals, where an investor buys a whole company or a bundle of properties at once rather than a single building, were up 376 percent. Those two categories did almost all of the work.
One deal explains most of it, and you have probably already seen it in a slide. On July 21, a group made up of AIP, MGX, and BlackRock's Global Infrastructure Partners bought Aligned Data Centers from Macquarie for about $40 billion. That covers 51 campuses and more than 6.4 gigawatts of power capacity, with another $5 billion committed for expansion. It is a huge, important transaction. It is also the sale of an entire data center platform, which has almost nothing in common with, say, a 220-unit apartment community in Tampa.
A couple of other data points are worth keeping in perspective. JP Morgan told clients that these monthly figures usually get revised upward by about 30 percent, so they feel comfortable heading into the fall. That may be true, but a revision does not change what actually sold. Green Street counted $164 billion of large deals in the first half of the year, up 30 percent, and the same question applies there: which property types actually moved? JLL's latest data, meanwhile, shows that the properties that do attract competitive bidding are still getting it. All of that is useful, and none of it is the same as saying every corner of the market had a $74.4 billion month.
One number does deserve your attention: the 10-year Treasury is still above 4.5 percent. That is the real signal in this report. Do not let an exciting data center headline talk you into assuming a lower cap rate on an ordinary apartment deal when you eventually sell it, which would quietly make the deal look better than it is.
What actually changed hands
Set the data centers and the big platform deals aside, and here is how the month looked across the property types most teams actually buy.
| Sector | July 2026 volume | vs. July 2025 |
|---|---|---|
| Data centers | $33.8B | +1,911% |
| Industrial | $9B | Flat |
| Urban office | $2.2B | +48% |
| Suburban office | $5.5B | +28% |
| Apartments | — | −16% |
| Retail | — | −13% |
| Hotels | — | +61% |
| Senior housing | — | +55% |
| Total | $74.4B | +78% |
| Total, without data centers | — | +1% |
Industrial sales came in at $9 billion, essentially flat versus last July. Apartment sales fell 16 percent, and retail fell 13 percent.
A few property types did have strong months. Urban office sales rose 48 percent to $2.2 billion, suburban office rose 28 percent to $5.5 billion, hotels were up 61 percent, and senior housing was up 55 percent. Those are real gains. They are also not the kinds of deals most multifamily and industrial teams are trying to close this quarter.
If you want a steadier number to point to, the rolling twelve-month total is $654 billion, up 34 percent. That is a better figure for a year-end investment committee memo, though even it includes a full year of large data center and infrastructure deals. Here is the trap to watch for: if your IC slide says "volume is back," and the next line assumes an aggressive going-in cap rate (the yield you are paying at the moment you buy) on a mid-tier Sun Belt apartment, you have let a data center month quietly set the tone for an apartment model.
The real story of July is composition, not momentum. A handful of very large, very specific assets changed hands, while the property types your analysts model every day either stayed flat or moved the other way.
The average cap rate isn't your cap rate
A quick reminder before the numbers: a cap rate is roughly the annual return a property produces relative to its price, and a basis point is one hundredth of a percent. Lower cap rates generally mean higher prices.
The average cap rate across all July sales was 6.89 percent, up 6 basis points from June. But that average hides a lot. Hotels sold at 8.32 percent. Office averaged 7.54 percent. Industrial moved the most, widening 28 basis points from June to 7.44 percent. None of those averages is a number you can safely drop straight into your model.
An average blended across hotels, office, industrial, and a $40 billion data center platform is really just a temperature reading for the market as a whole. It is not the rate that comparable buildings in your submarket and asset class are actually trading at. Apartments, for example, have been selling at tighter cap rates than the rest of the market all year. RealPage, working from the same MSCI data, put the average apartment cap rate at 5.79 percent in the second quarter. That is 110 basis points below the all-property July figure. Blend the two together and you are inventing a market that does not exist.
The same caution applies when you are on the buying side of an offering memorandum. The OM will often quote a "market cap rate" pulled from a national chart. Your job is to find the three or four sales that actually look like this property, in this class, over the last two or three quarters, value it off the trailing income rather than the seller's projections, and then decide whether the asking price holds up.
Underwrite the deal, not the headline
This $74.4 billion headline is going to land in your inbox this week. Some of those emails will attach an apartment OM that assumes rent growth the property's own financials do not support, with "volume is back" as the justification. But the volume came back in a sector you are not buying.
The work itself does not change. Rebuild the property's net operating income from the T-12 (the trailing twelve months of actual income and expenses). Check the rent roll against the income statement to make sure they agree. Put the big line items, like insurance, taxes, and payroll, on a forward-looking run-rate basis instead of last year's numbers. Then ask a simple question: does the seller's growth story still hold up in a market where apartment sales are down 16 percent and industrial cap rates just moved 28 basis points in a single month?
The tedious part is still gathering and rebuilding the numbers. The judgment part is still the same too: is 2 percent rent growth realistic on this specific property, and does the exit need another 25 to 50 basis points of cushion because interest rates are staying high? A confident but wrong number is still the worst outcome you can have, and a splashy headline just makes it easier to feel sure of yourself.
This is exactly where putting the OM into the Excel template your IC already trusts pays off. Every figure is tied back to a page in the source documents, so your analyst can spend time on the handful of assumptions that actually make or break the deal. From there, AcquiScore can flag the deal as Proceed, Caution, or Pass against your buy box, which helps when your inbox is full of "see, July was huge" cover notes. It does not make the decision for you. It just keeps a data center month from setting the tone on a deal it never really belonged in.
What to do this week
If a new deal package shows up with a national volume chart in its first few pages, take that chart out of your IC appendix. Replace it with the actual breakdown: $33.8 billion of it was data centers, apartment sales were down 16 percent, industrial was flat at $9 billion, and cap rates were mixed rather than falling across the board.
Then go back to a deal you are already working on. Load the T-12 and rent roll into your Excel model, with every figure on net operating income, occupancy, and in-place rent traceable to the source. See whether the seller's growth story still stands once you refuse to borrow a 78 percent headline that was never really about your market.
Frequently Asked Questions
Why did July CRE sales look so strong if apartments were down?
Because $33.8 billion of the $74.4 billion total was data center deals, led by the very large Aligned platform sale. Take that one sector out and MSCI's July sales were up only 1 percent from a year earlier. Apartment sales actually fell 16 percent, and industrial was flat at $9 billion. The strong headline reflects a change in what sold, not a broad rise across the whole market.
Should I use the 6.89 percent July average cap rate in my model?
No. That figure is an average across every property type. Hotels sold at 8.32 percent, office at 7.54 percent, and industrial at 7.44 percent, while apartments have been trading much tighter, around 5.79 percent in the second quarter on the same data. For your model, use recent closed sales of similar properties in your own submarket and class, ideally from the last two or three quarters.
How should a buy-side team read a month dominated by one huge deal?
Keep the big platform deals separate from ordinary building sales. The Aligned transaction was a roughly $40 billion sale of an entire data center company, with 6.4 gigawatts of power attached, so a typical 250-unit apartment deal is not comparable to it. For year-end context, the rolling twelve-month total of $654 billion, up 34 percent, is a steadier guide than any single month.
How does AcquiOS keep a headline number from creeping into the underwriting?
AcquiOS builds your model from the OM, the T-12, and the rent roll, inside your own Excel, with every number tied back to a page in the source documents. Market data is used to test the assumptions in the file, not to overwrite them with a headline volume figure. AcquiScore then ranks the deal against your written buy box, so the month stays in the market commentary instead of turning into your cap rate.