Cold Storage: Reading Food-System Data for Demand

Every other property type in this series suffers the same foundational absence: no federal series counts its capacity. Self-storage lacks a square-footage census, hotels a room count, campgrounds a site inventory. Cold storage is the exception, and the exception is spectacular. Every two years the Department of Agriculture surveys the nation's refrigerated warehouses and publishes their capacity in cubic feet, by state, by warehouse type, by cooler and freezer split; every month it publishes what is actually sitting in them, commodity by commodity, in pounds. The asset class the market treats as a niche is, in data terms, the best-instrumented industrial real estate in America, and this article reads those instruments the way its manifest title promises: as food-system data, because cold storage demand is the food system's working inventory made spatial. What follows uses no brokerage absorption figure, no market-size projection and no vendor's coverage; the federal instruments carry the whole argument, which for once they can.

The reading matters now for a reason the headline series understates and its composition reveals. Capacity is growing, 3.99 billion gross cubic feet on 1 October 2025, up 7.9% in two years (USDA NASS, Capacity of Refrigerated Warehouses, February 2026). But the growth's location, in the ownership sense, has flipped, and the flip is this article's finding.

The capacity census: what exists, measured in cubic feet

Start with the instrument itself. The 54th biennial survey counts 931 refrigerated warehouses: 476 public warehouses holding 2.46 billion gross cubic feet, 62% of national capacity, and 455 private and semiprivate facilities holding 1.53 billion (USDA NASS, February 2026, reference date 1 October 2025). Usable capacity, net of racking and workspace, runs 3.25 billion cubic feet, 82% of gross, and the temperature split is decisive for the real estate: 79% of usable space is freezer, 21% cooler, which is why the buildings are the most expensive boxes in industrial construction and why their conversion to or from ordinary warehouse use is effectively a rebuild. NASS states its frame honestly, usable reports from about 374 firms representing about 38% of capacity with estimation for the balance, and the survey's reliability note belongs in any memo that leans on it, per the provenance discipline this library applies to its own sources.

Freezer and cooler: one asset class, two thermodynamic products

The 79-to-21 freezer-to-cooler split inside usable capacity is not a detail; it is the product segmentation. Freezer space holds the long-cycle national inventory, proteins and frozen produce bought against seasons and shocks, and its demand follows production and trade. Cooler space holds the short-cycle chilled chain, dairy, produce in transit, prepared foods, whose demand follows consumption geography and turns in days. The two command different construction, different racking, different energy profiles and different customers, and a market can be long freezer and short cooler at once. The survey publishes the split by state, which makes the first question about any local imbalance answerable from the same table that posed it; the second question, which temperature band the proposed box serves, belongs in the memo's first paragraph, because everything downstream, the shipper roster, the utilization pulse, even the power arithmetic, divides along it.

The composition shift: the boom is private

Now run the series backward, type by type, and watch the industry reorganize. In October 2017 public warehouses held 2,834,824 thousand cubic feet and private and semiprivate facilities 767,331, a 79-to-21 split. By 2019 the figures were 2,683,640 against 961,851; by 2021, 2,670,456 against 1,063,864; by 2023, 2,514,133 against 1,181,615; and by 2025, 2,456,683 against 1,531,046 (USDA NASS, Capacity of Refrigerated Warehouses, February 2026, historical table). Total capacity grew 10.7% across those eight years. Every cubic foot of the net growth, and more, is private: own-account capacity doubled, up 99.5%, while the public third-party tier shrank 13.3%. The cold storage boom the trade discusses is real, and it is a private-fleet story, processors, retailers and food distributors building refrigerated capacity inside their own networks, while the investable, leasable, third-party tier that most cold-storage real estate underwriting actually concerns contracted in absolute terms through the strongest food-logistics decade on record.

The finding rewrites the demand question a lender should ask. National capacity growth is not evidence for a public-warehouse project, because the growth is happening on the other side of the ownership line; the relevant series for a third-party facility is the public tier's own trajectory, down, plus the local question of which shippers in the catchment lack own-account capacity and must buy space. Both halves are readable: the type split by state sits in the same survey tables, and the shipper base is the food-system employment the payroll files count, which this article reaches below. And the flip's timing is itself informative: the public tier's decline predates the pandemic, visible already between 2017 and 2019, so the reorganization is structural strategy, not a shock response, and nothing in the series suggests it has finished. A demand memo that quotes the 7.9% headline without the composition has, with perfect sourcing, reversed the signal.

One reading caution inside the series: the biennial line is lumpy. Total capacity eased from 3,734,321 thousand cubic feet in 2021 to 3,695,752 in 2023 before the jump to 3,987,732 in 2025, and part of every biennial move is frame, warehouses entering and leaving the survey's universe, rather than construction alone. NASS's own reliability note says as much. The honest use of the series is direction and composition across surveys, not a two-year growth rate quoted to a decimal, and the composition, public against private, is robust across every vintage precisely because it does not depend on the frame's edges.

The monthly pulse: what is actually in the freezers

Between the biennial capacity censuses runs the monthly Cold Storage report, the utilization pulse no other asset class possesses. On 30 June 2026 the nation's refrigerated warehouses held, among the majors, 1,433,068 thousand pounds of natural cheese, 99% of a year earlier; 969,847 of frozen fruit, up 6%; 1,343,610 of frozen vegetables, 92% of a year earlier; and 332,130 of butter, down 7% (USDA NASS, Cold Storage, July 2026 edition, June 30 holdings). Frozen poultry stood at 1,157,505 thousand pounds, 2% above a year earlier after a 4% monthly build, and total red meat ran 3% above its year-ago line, the protein freezers carrying the report's largest volumes. For demand analysis the stocks series does what occupancy data do elsewhere, imperfectly but publicly: a commodity group running persistently above its year-ago line is filling freezer space somewhere specific, and the report's commodity detail points at which supply chains, and therefore which regions and which warehouse types, are carrying the load. The June snapshot reads as a working example: fruit stocks 6% above a year earlier and vegetables 8% below describe two produce chains in different inventory positions, poultry building into the grilling season while cheese holds flat describes the protein and dairy cycles doing ordinary work, and none of it required an interview. The cadence is the gift: monthly, free, decades deep, and revised transparently. The report's own historical-records table underlines how deep: series for the major commodities reach back a century, which means the current month's holdings can be placed against every cycle the industry has ever run, a context no private dataset in any asset class of this library can offer.

The industry, counted at its payrolls

The operator layer confirms the boom and locates its labor. County Business Patterns counts 1,376 refrigerated warehousing establishments with paid employees in 2023, employing 66,236 people, only 21.9% of establishments under five employees, the census signature of an industrial-scale industry (U.S. Census Bureau, County Business Patterns 2023, NAICS 493120, 2025). The payroll series adds the arc: 1,440 reporting units in 2019 grew to 1,802 in 2024, up 25.1%, with employment up 11.3% to 73,374 and average pay up 24.4% to $65,258 (BLS, QCEW annual averages, NAICS 493120, 2020 and 2025). Note what the pair of growth rates says beside the capacity series: establishments grew far faster than employment, and capacity grew faster than either on the private side, which is automation arriving, the high-bay, rack-supported, sparsely staffed build that has dominated recent construction. The demand consequence is a supply-quality split the way this series found in flex and hotel stock: the new automated tier competes for high-throughput contract business, the legacy tier holds the flexible, short-commitment demand, and the two are one asset class only in the NAICS table. The general warehousing comparison sharpens it: the sector this article's flex industrial piece measured runs 90.4 employees per establishment against refrigerated warehousing's 48.1, cold's automation lead in one ratio.

Own-account or third-party: the decision that moved the market

The composition shift has a microeconomic engine worth stating, because a lender can read its local state. A shipper builds its own freezer when its volume is large and steady enough to fill one, when its network is stable enough to fix the location, and when control of service levels is worth the capital; it buys third-party space for overflow, seasonality, market entry and everything below the scale threshold. The national series says the threshold has been falling, automation and cheap design repetition making own-account viable for mid-sized shippers who once leased, which is how the private line doubled while the public line shrank. Locally, the make-or-buy balance is legible: the shipper roster's size distribution from the employer censuses says how many catchment shippers sit above plausible own-account scale, the permit record says which have already acted, and the remainder, the sub-threshold and the seasonal, is the third-party facility's true addressable base. A public-warehouse pro forma whose anchor tenant is one large shipper is, on this arithmetic, lending into the segment most likely to leave; the durable base is the many-shipper middle, and the roster counts it. The survey even offers a national check on the local reading: where a state's public capacity holds while its private line grows, the make-or-buy migration is early there, and where public capacity is already falling, late; the state tables date each market's position on the curve.

The power line: the opex that prices the box

Refrigeration is electricity made cold, and the operating line that dominates every cold-storage pro forma has a federal price series. Industrial electricity averaged 6.81 cents per kilowatt-hour in 2019 and 8.62 in 2025, up 26.6%, with the climb concentrated since 2021 (U.S. EIA, Monthly Energy Review, Table 9.8, 2026). For demand analysis the power line does two jobs. It is the wedge between the tiers, because the automated new build's kilowatt-hours per pallet run far below the legacy stock's, so rising power prices shift contract demand toward exactly the tier the establishment boom built, a mechanism visible in three public series without one operator interview. And it is a siting variable: state industrial rates differ by multiples, the same EIA tables publish them, and the cold-chain map's drift toward cheap-power states is the rational response the capacity survey's state detail lets an analyst check rather than assert.

The geography: where the cold chain sits

The state capacity table reads like a map of the food system rather than of population, which is the point. California leads with 400 million cubic feet of gross capacity, then Georgia at 304, Washington at 301, Wisconsin at 297 and Texas at 254 (USDA NASS, February 2026): produce and ports in California and Washington, poultry in Georgia, dairy in Wisconsin, scale and distribution in Texas. Warehouse counts follow the same logic, California with 106 facilities, Georgia 62, Texas 56, Florida 50. Set this against the per-capita maps of this series' consumer asset classes and the difference is structural: cold storage concentrates at production and gateway nodes, not at consumption, because the freezer inventory sits where the food enters the system, and the last-mile chilled layer that serves consumption is thinner, newer and largely inside the private fleets the composition shift measured. For a catchment analysis the instruction follows: a third-party facility's demand base is the food production, processing and import volume within trucking range, all of it counted by the USDA and Census programs the agricultural facilities article reads, plus the grocery and food-service distribution employment the payroll files locate.

The size-group discipline

The survey's size-group tables carry the concentration fact every underwriter should hold: national capacity lives overwhelmingly in large facilities, and the small-warehouse tail, numerous in the count, is thin in cubic feet. The analytical habit this series applies to employment size classes transfers directly: never reason from facility counts when the quantity that matters is capacity, and never compare a market's warehouse count with another's without the size distribution beside it. A metro with six facilities can out-store one with twenty, and the survey's own tables, not an estimate, say which is which. The same discipline guards the competitive analysis: the legacy box across the highway is a competitor in the count and, depending on its cubic feet, its temperature split and its dock throughput, possibly no competitor at all for the contract the subject facility wants.

The pharma boundary, drawn where the data draw it

Temperature-controlled real estate has a second life outside the food system, pharmaceutical and biomedical cold chains, and the honest statement is that the instruments this article celebrates do not cover it. The USDA survey's universe is food-system warehousing; regulated pharmaceutical storage answers to a different federal regime and publishes no capacity census. The boundary matters for scope discipline in both directions: a food-side analysis should not borrow pharma growth narratives to pad its demand case, and a pharma-side project cannot lean on the USDA tables for its supply picture and must build its inventory from the industry codes, licensure and parcel records the long way, the way every other asset class in this library does. The one shared layer is the power line, which prices both.

The gateway layer: imports in the freezer

A growing share of what sits in coastal cold storage arrived by ship, and the trade side has its own named public sources: the Census Bureau's foreign-trade statistics count refrigerated commodity imports by customs district, and USDA's trade data programs track the agricultural flows by product and port. For a gateway-market facility the import series are the demand base's second half beside domestic production, and the state capacity table already shows their weight, in Washington's 301 million cubic feet serving Pacific trade and produce and in Georgia's poultry-and-port complex. The catchment method extends naturally: shipper roster, plus the customs district's refrigerated import trend, plus the drayage geography that ties the port to the boxes, each layer public, each dated, and the gateway premium or its absence becomes a measured feature of the market rather than a broker's adjective. The inland mirror deserves its sentence too: the rail-served distribution freezers of the middle of the country live on domestic production flows the agricultural census maps, and their gateway question is the intermodal ramp, not the pier, with the same method applied to different public tables.

Reading a catchment for a refrigerated box

The catchment method adapts the series' standard playbook to a business-to-business asset. The demand base is the shipper roster: food manufacturers, processors, distributors and importers within the service radius, counted by code and size from the County Business Patterns method and the QCEW demand-driver files, because every refrigerated pallet has a payroll behind it. The roster's composition matters as much as its count: processors generate steady base-load storage, importers generate seasonal surges, distributors generate turns, and a catchment's mix of the three, readable from the codes, predicts the contract structure a facility can actually sell. The competitive set is the capacity survey's state table brought local: the public warehouses in range, from the survey frame and the address-level reconciliation this library always runs, with the private fleets noted as demand subtractions rather than competitors, since an own-account freezer is a customer the market already lost. The labor layer is the same QCEW files at county grain, cold-chain work is specialized and the wage line prices it. And the workforce geography, where warehouse labor actually lives relative to the box, is the LODES commuting method applied to an industry whose shifts make commute range a real constraint. The output is the standard five-line demand statement, every line sourced, for an asset class where the usual excuse, that no data exists, is the one excuse the USDA removed.

The labor layer, cold-chain specific

The staffing gate this series applies everywhere binds differently here, and the payroll numbers show how. Refrigerated warehousing's average pay of $65,258 in 2024 sits well above general warehousing's, the premium for work done at temperature with certifications attached, and the industry's employment grew half as fast as its establishments, automation absorbing the growth. For a specific facility the labor question is therefore less whether workers exist than whether the specialized tier does, refrigeration technicians above all, whose scarcity idles freezers as effectively as power failures. The county QCEW rows price the local cold-chain wage; the commuting method places the workforce; and the pro forma whose maintenance line assumes technicians at general-warehouse wages has made the sector's characteristic quiet error, visible in one comparison of public series.

A note on reading two clocks

The asset class's instruments run on two clocks, and the method keeps them straight. The capacity census is biennial and structural: composition, geography, tiers. The stocks report is monthly and cyclical: utilization, commodity pressure, season. Mixing the clocks produces the genre's standard errors, a biennial capacity jump read as a demand surge, a monthly stock build read as structural shortage, and the memo avoids both by labeling every figure with its clock. The discipline costs one column in a table and buys the file its longevity: a committee rereading it in three years should find the structural claims still standing on the biennial series and the cyclical claims dated to their months, each aging the way its clock intended.

One further habit closes the method: reconcile the two clocks once per file. Stocks divided by usable capacity gives a crude national utilization band, and the same division at state grain, holdings pressure against surveyed capacity, flags the markets where the freezers are structurally full against those with headroom. The ratio is rough, the commodity mix confounds it, and stated with those caveats it remains the only public utilization signal in industrial real estate, one more instrument the asset class did not have to earn but gets to use.

The lending classification, and why it is deserved

Cold storage sits explicitly on the special purpose side of SBA lending guidance: the reported examples from SOP 50 10 8 include cold storage facilities where more than 50% of total square footage is equipped for refrigeration, carrying the 15% borrower contribution on 504 projects against the standard 10%, with 20% for a new business (504 Capital Corporation, guidance on SOP 50 10 8, 2025; NAGGL, 2025). The classification is the composition shift's financial echo. A freezer building's improvements, envelope, refrigeration plant, floors engineered for thermal cycling, are worthless to the next ordinary-warehouse tenant and a rebuild for the next cold one, so the collateral's value concentrates in the demand analysis this article has been assembling: the shipper base, the public-tier trajectory, the power economics and the tier quality of the specific box. The public SBA 7(a) and 504 datasets, which MMCG Analytics' SBA layer is built on, carry the code's lending history, subject to the standing rule that no performance rate is shown for any cohort of fewer than ten loans. MMCG Analytics supplies the data and the analysis; the credit decision rests with the lender.

The classification also sets the workout arithmetic the demand file should anticipate: a defaulted freezer re-lets only into the third-party tier this article showed contracting nationally, so the exit analysis leans on the same local series as the entry, the shipper roster's depth and the public tier's regional trajectory, and a market that supports the loan on entry arithmetic but not on exit arithmetic has told the committee which covenant matters.

Method: the five numbers a cold storage memo should carry

First, the national frame with its composition: total, public and private capacity from the biennial survey, stated by type so the 7.9% headline cannot reverse the signal. Second, the utilization pulse: the relevant commodity groups' stock lines from the monthly report, current against year-ago, matched to the facility's intended business. Third, the shipper roster: food-system establishments by code and size in the service radius from the employer censuses, with the own-account subtraction noted where a major shipper runs its own fleet. Fourth, the competitive tier map: public warehouses in range with their vintage and automation generation from permits and imagery, because the tier split prices contracts. Fifth, the power line: the state industrial rate and its five-year trend from the EIA tables, the opex that decides which tier wins the marginal pallet.

Each carries a source and a date; the contract assumptions on top are the operator's, stated as such; and the provenance standard and the pillar's discipline govern, with one local grace note: this is the pillar's only chapter whose capacity denominator comes from a federal survey rather than from the analyst's own reconciliation, and the memo should enjoy it.

A worked sequence for one freezer

Run for a proposed 5-million-cubic-foot public freezer at a regional food hub, the sequence goes as follows. The survey's state table gives the public-tier context: the state holds, say, 180 million cubic feet, 60% public, flat across two surveys while the private line grew. The shipper roster within 150 miles counts 240 food manufacturers and processors and 60 distributors, three of national scale, one of which, the permit record shows, broke ground on its own freezer last year, an own-account subtraction the memo names. The monthly stocks relevant to the hub's business, poultry and frozen vegetables for this geography, run at or above year-ago lines. The competitive tier map finds four public facilities in range, two automated builds under five years old and two legacy boxes, and the state industrial power rate sits below the national 8.62 cents with a flat trend, favoring the operating model. The memo that results states: a public tier locally stable against the national contraction, a counted shipper base with one named subtraction, utilization signals at trend, a competitive set split by tier, and a power position that advantages the new build, five sourced lines under whatever contract assumptions the operator defends. The tier map's vintage column does one more quiet job: it dates the market's last construction cycle, and a catchment whose newest public freezer is fifteen years old has been ceding its throughput contracts to whichever neighbor built since, a competitive fact the permit record states without a tour. The protective case reads itself: a catchment whose anchor shippers have gone own-account, whose public tier is shrinking faster than the nation's and whose power rate is climbing is a market where the box will chase contracts downhill, and every clause came from a public file.

The 2026 read, and the cadence

Read in August 2026, cold storage is the series' cleanest case of a headline and a structure disagreeing. The headline: capacity at a record 3.99 billion cubic feet, up 7.9%, establishments up a quarter since 2019, the boom real by every count. The structure: all net growth private, the public tier down 13% over eight years, automation splitting the stock into tiers, and power costs up 26.6% pricing the split. Demand for the leasable tier is therefore a local, shipper-by-shipper question this article's method answers with rosters and subtractions rather than with the national curve. The cadence is the asset class's luxury: stocks monthly, payrolls quarterly, power monthly, the capacity census every two years, and a memo built on those instruments is refreshed by the government on a schedule no other property type in this library enjoys. The next article in this queue leaves the food system for the housing one, and the closing member of the pillar's core, multifamily, will find its own federal instruments waiting, permits, starts, completions and vacancies, the series this whole library began with.

Frequently asked questions

What public data measures cold storage capacity and demand?

USDA NASS publishes the only federal capacity census in commercial real estate, the biennial Capacity of Refrigerated Warehouses (3.99 billion gross cubic feet at 1 October 2025, by state and warehouse type), plus the monthly Cold Storage report of holdings by commodity. County Business Patterns and QCEW count the operators; EIA prices the power.

Is cold storage capacity growing?

Yes, and the growth is private. Total gross capacity rose 10.7% from 2017 to 2025, but private and semiprivate capacity doubled (767 million to 1.53 billion cubic feet) while public third-party capacity fell 13.3% (2.83 billion to 2.46 billion). The boom is own-account cold chains, not the leasable tier.

What do the monthly Cold Storage stocks show?

Utilization by commodity: on 30 June 2026, natural cheese stood at 99% of a year earlier, frozen fruit at 106%, frozen vegetables at 92%, butter at 93%, poultry at 102% and red meat 3% above its year-ago line. Persistent above-trend commodities identify which supply chains are filling freezer space.

Where is U.S. cold storage concentrated?

At production and gateway nodes, not population: California leads with 400 million gross cubic feet, then Georgia (304), Washington (301), Wisconsin (297) and Texas (254), the geography of produce, poultry, dairy and ports. The last-mile chilled layer near consumers sits largely inside private fleets.

Why is cold storage a special purpose property for SBA lending?

Reported SOP 50 10 8 examples classify facilities with more than 50% of square footage equipped for refrigeration as special purpose, with a 15% borrower contribution on 504 projects (20% for new businesses). The refrigeration plant and envelope have no value to ordinary-warehouse users, which concentrates the credit on demand analysis.

How do power prices affect cold storage demand?

Electricity is the dominant operating cost, and industrial rates rose 26.6% from 2019 (6.81 cents per kWh) to 2025 (8.62 cents). Rising power prices shift contract demand toward automated, efficient new builds and against legacy stock, and state rate differences make power a siting variable the same EIA tables quantify.

Sources

1. USDA National Agricultural Statistics Service, Capacity of Refrigerated Warehouses (54th biennial survey, reference 1 October 2025), February 2026; capacity series by type 2017 to 2025 and state tables. https://www.nass.usda.gov/Publications/Todays_Reports/reports/rfwh0126.txt
2. USDA National Agricultural Statistics Service, Cold Storage, July 2026 edition (holdings on 30 June 2026 by commodity), 2026. https://www.nass.usda.gov/Publications/Todays_Reports/reports/cost0726.txt
3. U.S. Census Bureau, County Business Patterns 2023, NAICS 493120 (refrigerated warehousing and storage), released 26 June 2025. https://www2.census.gov/programs-surveys/cbp/datasets/2023/cbp23us.zip
4. U.S. Bureau of Labor Statistics, Quarterly Census of Employment and Wages, annual averages 2019 and 2024, NAICS 493120, private, United States, 2020 and 2025. https://data.bls.gov/cew/data/api/2024/a/industry/493120.csv
5. U.S. Energy Information Administration, Monthly Energy Review, Table 9.8, average industrial electricity price, annual averages 2019 to 2025, 2026. https://www.eia.gov/totalenergy/data/browser/csv.php?tbl=T09.08
6. 504 Capital Corporation, SBA 504 Loans for Special Purpose Properties and Real Estate (guidance on SOP 50 10 8 classifications, including the more-than-50% refrigeration standard), 2025. https://504capital.com/blog/financing-special-purpose-properties-sba-504-loans/
7. National Association of Government Guaranteed Lenders, SBA Notice Revising SOP 50 10 8 (Procedural Notice 5000-872764), 2025. https://www.naggl.org/sba-notice-revising-sop-50-10-8/
8. U.S. Census Bureau, County Business Patterns 2023, NAICS 493110 (general warehousing, the employees-per-establishment comparator), 2025. https://www2.census.gov/programs-surveys/cbp/datasets/2023/
9. MMCG Research, SBA 7(a) Performance Series: MMCG analysis of the public SBA 7(a) loan register, 2026. https://mmcganalytics.com/sba-default-rates/
10. USDA National Agricultural Statistics Service, 2022 Census of Agriculture (the production base the agricultural-facilities method reads), released 13 February 2024. https://www.nass.usda.gov/Publications/AgCensus/2022/
11. U.S. Census Bureau, Nonemployer Statistics 2023, released 15 May 2025. https://www.census.gov/newsroom/press-releases/2025/2023-nonemployer-statistics.html