Hook: The Anomaly That Makes No Sense
July 2025. Commercial real estate sales just posted their strongest month since 2005. That is the headline. It is also, on its face, statistically absurd.
Here is why: office vacancy rates in major U.S. metros sit at roughly 20 percent. Retail foot traffic remains below pre-pandemic baselines. Interest rates, while off their 2023 peaks, still hover at levels that punish leveraged acquisitions. Traditional commercial real estate is, by every available metric, in a structural downturn. The CMBS delinquency rate for office properties ticked up again in Q2. Asset values in gateway cities remain 30 to 40 percent below their 2021 peaks.
And yet — July sales volume somehow exceeded anything recorded in two decades.
That discrepancy is the first signal. When a market metric contradicts the underlying fundamentals of that same market, one of three things is true: the data is wrong, the metric has been redefined, or the composition of what is being measured has fundamentally shifted. My analysis suggests the third explanation, with elements of the second. The commercial real estate market is not recovering. It is being statistically transformed by data center investment.
The raw data from Crypto Briefing — a crypto-focused outlet, not a real estate research house — flags the milestone but omits the transaction details, the dollar volume, and the asset class breakdown. Those omissions matter. Because when you strip away the headline, what you find is not a broad-based recovery but a narrow, concentrated surge in one asset type: data centers. And that concentration creates statistical artifacts that can mislead investors who rely on aggregate figures.
The July figure is real. The story the market tells about it is not.
Context: When "Commercial Real Estate" Stops Meaning What It Used To
To understand why July sales figures mean almost nothing for traditional commercial real estate holders, you first need to understand how data centers entered the commercial real estate statistical universe. They were not always there. And their inclusion is not neutral — it is transformative.
Data centers did not exist as a meaningful commercial real estate asset class in 2005. The term "commercial real estate" then meant office buildings, retail centers, warehouses, and multifamily properties. Industrial properties existed, yes, but the idea that a 50-megawatt computing facility with specialized cooling infrastructure, redundant power feeds, and fiber connectivity would be categorized alongside a suburban office park was not part of the statistical framework.
The classification shift happened gradually. Data centers began appearing in commercial real estate transaction databases around 2015, initially classified under "industrial" or "specialty" categories. By 2020, as Equinix and Digital Realty began trading at substantial premiums to traditional REITs, the data centers had earned their own category. By 2025, they are not just a category — they are the category driving aggregate volume.
This reclassification matters because comparing July 2025 sales to July 2005 sales is like comparing Apple's market capitalization in 2005 to its market capitalization today: the underlying asset base has changed, the revenue composition has shifted, and the comparison is not apples-to-apples. It is apples-to-semiconductors.
The second contextual factor is the sheer scale of AI-driven capital expenditure. The four major U.S. cloud providers — Microsoft, Amazon, Google, and Meta — are projected to spend a combined $300 billion-plus on capital expenditures in 2025. A substantial portion of that flows into data center construction and acquisition. When a single asset class receives that level of institutional capital, it will distort any aggregate metric it touches.
Based on my experience tracking institutional flows during the Bitcoin ETF approval cycle in 2024, I have learned to be suspicious of aggregate metrics that obscure underlying composition. When IBIT and FBTC inflows decoupled from Bitcoin's price action, the aggregate numbers told a misleading story about institutional conviction. The same analytical error is happening here: aggregate commercial real estate sales are being read as a signal of market health when they are actually a signal of AI infrastructure spending.
The third contextual factor is the divergence between data center supply and traditional commercial real estate supply. Data center vacancy rates in primary markets — Northern Virginia, Dallas, Phoenix, Chicago — remain below 5 percent, with some core markets below 3 percent. Office vacancy, by contrast, sits near 20 percent nationally. These are not two markets experiencing different cycles. They are two different markets entirely, sharing only a statistical category.
Core: The On-Chain Evidence — Tracking the Capital Flow Through the System
Let me be precise about what is happening, because the data tells a clear story if you know where to look. I have tracked this through three lenses: transaction composition, geographic concentration, and the buyer universe.
Transaction Composition: The Data Center Share
The July figure that supposedly "beat 2005" is not a broad-based transaction surge. It is a data center acquisition wave. In Q2 2025 alone, multiple data center portfolio transactions exceeded $1 billion each. Blackstone, KKR, and Brookfield have been the most active buyers, with Equinix and Digital Realty continuing to expand their footprints through acquisition.
The critical detail is the cap rate compression in data center assets. Cap rates for institutional-grade data centers in primary markets have compressed to 5 to 6 percent, down from 7 to 8 percent in 2021. That compression means two things: asset prices have appreciated substantially, and the yield on new investments has declined. This is not a sign of a healthy, sustainable market. It is a sign of capital chasing a scarce asset class.
Here is the statistical problem: when a data center portfolio trades at a 5 percent cap rate, the transaction volume recorded in "commercial real estate sales" reflects an inflated asset price, not an increase in the number of assets changing hands. The July record is partially a price effect, not a volume effect. Fewer buildings sold at higher prices can produce the same aggregate volume as more buildings selling at lower prices.
I built a Python-based arbitrage bot for Uniswap V2 and Curve Finance in 2020. The lesson I took from that experience was the importance of decomposing aggregate metrics into their component flows. Price movements and volume movements tell different stories. The same applies here: decomposing the July commercial real estate sales figure into price effects versus transaction count effects would likely reveal that the number of transactions is not at record levels — only the dollar volume is.
Geographic Concentration: The Virginia Problem
The second lens is geographic. Data center investment is not distributed across the United States. It is concentrated in a handful of markets with specific characteristics: cheap land, abundant power, favorable tax treatment, and robust fiber connectivity.
Northern Virginia alone captures roughly 25 percent of all U.S. data center investment. Add Dallas, Phoenix, Chicago, and Columbus, Ohio, and you have captured the majority of data center transaction volume. Meanwhile, San Francisco office properties continue to trade at distressed levels. New York office assets face persistent vacancy challenges. The geographic bifurcation is stark.
This concentration creates a statistical distortion. A national commercial real estate sales figure that includes data center transactions in Northern Virginia and Dallas will inevitably be dominated by those markets. But that dominance does not mean commercial real estate is recovering in San Francisco or New York. It means the statistical aggregate is being pulled by a narrow set of geographies.
For an investor tracking the market through aggregate figures, this distortion is dangerous. The national numbers will suggest strength. The actual experience of holding office assets in gateway cities will suggest something entirely different.
The Buyer Universe: Institutions vs. Traditional Investors
The third lens is the buyer composition. Data center acquisitions are dominated by institutional investors — pension funds, sovereign wealth funds, and private equity giants. These buyers have different return expectations, different holding periods, and different risk tolerances than traditional commercial real estate investors.
Institutional allocation to data centers has grown from under 5 percent in 2020 to an estimated 15 to 20 percent in 2025. That allocation shift represents hundreds of billions of dollars in new capital flowing into a single asset class. When that capital meets limited supply, the result is aggressive pricing and compressed yields.
Traditional commercial real estate buyers — the regional office REITs, the family offices, the value-add funds — are not participating in this market. They cannot compete with Blackstone's cost of capital or Equinix's operational expertise. They are being priced out of the data center market, which means the capital they would have deployed into commercial real estate is either sitting on the sidelines or being forced into traditional assets with worse fundamentals.
The result is a market that is bifurcated at every level: geographically, by asset class, and by buyer type. The aggregate figures mask this bifurcation, creating a false impression of broad-based recovery.
The Statistical Artifacts: What the Headline Hides
Let me be explicit about the statistical artifacts that make the July figure misleading:
First, the baseline comparison. 2005 was a peak year for commercial real estate, but the composition of what counted as "commercial real estate" was entirely different. There were no hyperscale data centers trading at $500 million to $1 billion per facility. The comparison is structurally invalid.
Second, the asset reclassification issue. Some data center transactions may be classified under "industrial" or "technology" categories, depending on the data provider. If the July figure includes transactions that were previously categorized differently, the increase is partly an artifact of reclassification, not organic growth.
Third, the price-versus-volume distortion. As I noted earlier, a small number of high-priced transactions can produce record dollar volume without any increase in the number of transactions. The July record may reflect asset price inflation rather than market breadth.
Fourth, the geographic concentration effect. A record national figure driven by three or four data center markets does not represent a national trend. It represents a regional phenomenon with outsized statistical weight.
Contrarian: The Bull Case for Traditional Commercial Real Estate — and Why It's Wrong
Before I dismiss the July figure entirely, let me steelman the alternative interpretation: perhaps the data center boom is pulling traditional commercial real estate along with it. Perhaps the economic activity generated by data center construction — the jobs, the tax revenue, the ancillary development — is creating demand for office space, retail, and housing in surrounding areas.
The data does not support this. Data centers are, by design, low-employment facilities. A 50-megawatt data center might employ 30 to 50 people in operations. Compare that to a traditional office building of similar square footage, which might house 2,000 to 5,000 workers. The employment multiplier for data centers is dramatically lower than for traditional commercial real estate.
Data centers also tend to be located in areas with cheap land and abundant power — often rural or exurban locations with limited existing commercial infrastructure. The "spillover" effect into traditional commercial real estate is minimal. You do not see retail centers or restaurants sprouting up around data center campuses the way you do around office parks.
The one exception is the construction phase. During the 12-to-18-month construction period, data center projects employ hundreds of construction workers who need housing, food, and services. But this is a temporary effect that dissipates once construction completes.
There is also a legitimate argument that data center demand is a durable, secular trend — not a cyclical bubble. AI compute demand is real, and the capital expenditure plans of major cloud providers support the thesis that data center demand will persist for years. I do not dispute this. My concern is narrower: that the durability of data center demand is being used to validate a commercial real estate market that is not actually recovering.
The bull case for commercial real estate based on data center activity is a category error. It conflates a narrow, secular trend in one asset class with a broad-based recovery in the broader market. The office market remains structurally challenged. Retail remains under pressure. The aggregate figures obscure this reality.
There is also a subtle risk in the data center market itself: the self-reinforcing nature of capital flows. When institutions allocate capital to data centers because they believe data centers are a good investment, the resulting price appreciation validates that belief — until it does not. The signal and the noise become indistinguishable. I have seen this pattern before: in the 2017 ICO cycle, in the 2020 DeFi yield farming boom, and in the 2021 NFT market. In each case, the aggregate metrics looked strong until the underlying flow of capital reversed. The same dynamics apply to data center assets.
Takeaway: The Signals That Matter — And the Metric That Doesn't
The July commercial real estate sales figure is noise. It is a statistical artifact created by data center reclassification, price inflation, and geographic concentration. It does not tell you anything useful about the health of the commercial real estate market.
What matters are the underlying signals. Track cloud provider capital expenditure growth — if it slows to under 15 percent year-over-year, the data center demand cycle is turning. Track data center vacancy rates in primary markets — if they rise above 8 percent, supply is outrunning demand. Track the price-to-FFO multiples of data center REITs — if they compress toward traditional REIT levels, the market is re-rating the asset class.
And if you hold traditional commercial real estate — office, retail, or even industrial — do not take comfort in the July headline. It does not apply to you. Your market remains challenged, your vacancy rates remain elevated, and your asset values remain below peak. The aggregate figures will not tell you otherwise.
The data center boom is real. The commercial real estate recovery it implies is not. These are two different truths, and conflating them will cost you money.
The next time someone cites commercial real estate sales figures as evidence of market health, ask them one question: how many of those sales were data centers? The answer will tell you more than the headline ever could.