One Region Up, Four Down: Why a 24.8% National Drop Doesn't Buy a Houston Medical Buildout Any Slack

What the headline averages away

Dodge's September 21 release put total construction starts at a seasonally adjusted annual rate of $1.34 trillion for August, a 24.8% decline from July. Nonresidential building starts fell 32.0% to $624 billion. Commercial starts dropped 37.6%; manufacturing fell 80.8% after July's megaproject surge; institutional starts pulled back 18.1%.

Sarah Martin, Director of Economic Research at Dodge, framed the month as a return to trend rather than a break in it: “After a pop in activity last month, construction starts largely normalized throughout August. Abstracting from the month-to-month volatility, the story remains consistent. Data center, semiconductor and energy construction are driving growth, while several other sectors are facing subdued activity alongside deeper labor shortages, and accelerating material prices.”

That is the correct reading of the national number, and it is not the number that governs a buildout on Fannin or in Katy. Two figures underneath it are.

Figure one: the region

Dodge reports starts by region, and August split cleanly:

• South Central: +28.8% month over month — the only region to improve
• Northeast: −52.6%
• West: −51.4%
• South Atlantic: −24.2%
• Midwest: −1.1%

Construction capacity does not travel well. A superintendent, a sheet-metal crew, an electrical foreman, and a fire-sprinkler sub are local assets with local calendars. When starts collapse in the Northeast and the West, that does not free up a single Houston pipefitter. It just makes the national average look like relief that a Texas owner will never receive.

The practical translation: the bid environment an owner in Boston or Phoenix is about to enjoy is the opposite of the one in Houston. An owner who delays a Houston medical buildout because a national headline said starts fell 24.8% is timing a market they are not standing in.

Figure two: healthcare's bounce is a rebound, not a boom

Healthcare starts rose 96.1% month over month in August. That number invites a victory lap it does not deserve, and the honest read matters more than the flattering one. Dodge's own language is precise: “Following a weak July, healthcare starts normalized in August.” It is a snap-back off a low base, not a surge.

The annual figures are the counterweight, and they point the other way. Institutional starts are down 3.9% year to date through August and down 4.3% for the twelve months ending August 2026. Over that same twelve months, commercial starts are up 48.6% and total nonresidential is up 17.9%.

Put those two facts side by side and the strategic position becomes clear. Healthcare owners are not the ones bidding up Houston's trades. Someone else is, and healthcare is paying the price anyway. A medical, dental, or veterinary buildout is a price-taker in a regional labor market whose capacity is being consumed by commercial and energy work with a different balance sheet and a different tolerance for cost.

The capacity is already spoken for

Associated Builders and Contractors released its August backlog and confidence survey on Tuesday, September 15, and it quantifies exactly who holds the calendar.

• Construction backlog rebounded to 8.5 months after a cyclical low in July, though still below the level of the preceding several months.
• Roughly one in six ABC members is currently under contract on a data center — the highest proportion on record.
• Firms holding data center contracts report 9.9 months of backlog. Firms without report 8.3 months.
• The share of contractors intending to cut staffing over the next six months rose to 12.3%, the highest in any month since December 2025.

ABC Chief Economist Anirban Basu noted what the survey turned up without being asked: “There was a sharp increase in the number of contractors who, unprompted, mentioned labor shortages in the survey. With construction job openings at a near two-year high, it appears that labor scarcity is reemerging as a major headwind for the industry.”

ABC economist Zack Fritz named the shape of it directly: “This is most likely a reflection of the ‘haves and have nots’ dynamics of data centers. Those with data center work are struggling to find specialty trade contractors, while other contractors are facing continually softening demand.”

Read that last sentence as a healthcare owner, because it describes your bid list with uncomfortable precision. The specialty trades a clinic depends on — medical gas, controls, low-voltage, imaging-room shielding, the electrical sub who has actually done an OR or an operatory — are the scarce resource, and the firms competing hardest for them are not building clinics. Meanwhile the contractors with slack capacity are the ones facing softening demand, which is a polite way of saying the sub with immediate availability may be available for a reason worth understanding.

And the money got more expensive the same week

On Wednesday, September 16, the Federal Reserve raised its benchmark rate by 25 basis points — the first increase since 2023 — with inflation running stubbornly above the 2% target.

Michael Guckes, chief economist at ConstructConnect, described the first-order effect on the owner's pencil: “In the bigger picture, higher financing costs will weaken the profitability calculations of owners and developers considering new commercial real estate projects. Projects that were on the edge of meeting their profitability goals may now be falling short, such projects will not go to bid.”

He also flagged the second-order effect on projects already moving, which is the one contractors feel: short-term credit and revolving loans tied to the prime rate move up immediately. “If the change in rates severely impacts the project owner, this could result in an unexpected delay in contractor payments,” Guckes said. “In a worst-case scenario, higher rates could force the owner to put the project on hold, or even abandon it, which would leave the contractor saddled with immediate expenses and accounts receivable that are difficult to recoup.”

There is a credible counter-case, and it deserves equal space. Brian Strawberry, chief economist at FMI, argued that the overnight rate is not where commercial real estate actually borrows: “A hike would immediately raise the cost of floating-rate construction loans, which is real money on projects carrying debt, but the bigger effect is likely lower long-term rates. A hike that convinces markets the Fed is serious about inflation tends to pull the 10-year down, and that is where construction actually borrows.”

Both can be true, and for a medical owner the distinction is operational rather than academic. Floating-rate construction debt costs more starting now. Permanent take-out financing may cost less later. That combination rewards a short, certain construction duration and punishes a long, drifting one — which is a schedule problem, not a rate problem.

Two more cautions from the same reporting. Kenneth Roberts of Venable saw no relief coming on the cost side: “I don't see any data that suggests that in the next six months, the inflation that the construction industry has experienced in the last year is going to decrease.” And Guckes warned against reading one hike as the end of it: “In recent decades the Fed has not raised the Federal Funds Rate one time by 25 basis points or 50 basis points and then held at that rate for a sustained period of time. Rather, the Fed has a track record of making multiple small rate increases relatively quickly.”

Strawberry also noted which sectors are insulated, and it closes the loop on the capacity problem: data centers and semiconductor plants run largely on corporate cash flow and government incentives, so “borrowing costs barely touch it.” The rate hike slows the owner who competes for Houston's trades on borrowed money. It does not slow the buyer who is currently outbidding them.

Three data points, one instruction

Stack the week and the conclusion is unusually unambiguous:

1. The national slowdown is not local. South Central was the only region up, at 28.8%.
2. The scarce specialty trades are under contract to someone else, at a record rate, and the firms holding that work are the ones struggling to staff it.
3. Borrowing costs rose on September 16, with a named expectation of more, against an input-cost index already running 8.9% year over year as of the September 11 reporting covered in Brief #37.

Every one of those pushes the same direction: waiting is the expensive option, and the national headline is the thing most likely to talk an owner into it.

Where the recoverable time actually is

If the answer is “don't wait,” the immediate objection is fair: the project isn't ready. Drawings aren't done, the program is still moving, the lender wants a number. That objection is about preconstruction duration, and there is now a credible measurement of what preconstruction discipline is worth.

On Wednesday, September 16, Suffolk Construction and MIT published a joint white paper, “Construction in the Age of AI.” Applied together to a modeled 180,000-square-foot project, six interconnected levers produced an estimated 17% to 20% reduction in project cost and a 22% to 25% reduction in schedule. The six levers: design automation, offsite manufacturing, permitting, scheduling, skilled labor and subcontracting, and supply chain and procurement.

The finding that matters for a clinic is not the headline percentage — the study was a retrospective analysis of a single completed multifamily project, the authors state the results are not causal, and a 6,000-square-foot dental office is not a $180 million tower. The finding that matters is the structure. The paper identifies design automation as the “clearest upstream enabler” of the other five, because a resolved, data-complete design is what permit compliance checking, procurement, schedule generation, and subcontractor coordination all feed from.

Suffolk CTO Jit Kee Chin put the compounding plainly: “This is not like a math exercise for 1+1+1 equals 3. It's 1+1+1 equals, potentially, 6.”

Translated out of the research and into a Houston buildout: the single highest-leverage hour on a medical project is spent resolving the design before it goes to permit and buyout, because every downstream decision inherits that resolution or inherits its absence. An unresolved ceiling plan becomes an RFI, then a mechanical revision, then a re-quote on a package whose price moved 8.9% in the meantime, then a two-week slip on a floating-rate construction loan that just got more expensive. That is the compounding running in reverse, and it is the one most clinic projects actually experience.

By the Numbers

• Dodge Construction Network, August starts (released September 21, 2026): total starts −24.8% m/m to a $1.34 trillion SAAR · nonresidential building −32.0% to $624 billion · residential −5.2% · nonbuilding −26.7%
• Regional, month over month: South Central +28.8% (only region up) · Midwest −1.1% · South Atlantic −24.2% · West −51.4% · Northeast −52.6%
• Sector, month over month: healthcare +96.1% (normalizing off a weak July) · institutional −18.1% · education −14.3% · commercial −37.6% · offices and data centers −31.3% · manufacturing −80.8%
• The annual counterweight: institutional starts −3.9% year to date and −4.3% for the twelve months ending August · commercial +48.6% and total nonresidential +17.9% over the same twelve months
• ABC backlog and confidence (released September 15, 2026): backlog 8.5 months, rebounding off a July low · 1 in 6 ABC members under contract on a data center, highest on record · data center firms 9.9 months backlog vs. 8.3 months without · share intending to cut staffing 12.3%, highest since December 2025
• Federal Reserve (September 16, 2026): +25 basis points, first increase since 2023, with inflation above the 2% target
• Suffolk and MIT, “Construction in the Age of AI” (published September 16, 2026): six combined levers modeled at 17–20% cost reduction and 22–25% schedule reduction on a 180,000 SF case study · design automation identified as the clearest upstream enabler · authors state results are not causal
• Cost backdrop (ABC analysis of BLS Producer Price Index, reported September 11, 2026, covered in Brief #37): construction input prices +8.9% year over year

Action Items

1. Stop pricing your schedule off national headlines. Before the next owner meeting, replace “the market is slowing” with a regional number. South Central was up 28.8% in a month the country fell 24.8%. If a decision to delay is being justified by a softening market, the justification does not survive contact with the regional table.
2. Lock the specialty trades before you lock the general contractor. Medical gas, controls, low-voltage, shielding, and the electrical sub with actual clinical experience are the constrained resources — not the GC. Pre-qualify them, get dated letters of intent, and confirm crew availability against your target window. A record one in six contractors is on data center work; the calendar you need is being filled right now by someone else's project.
3. Interrogate immediate availability. Backlog for firms outside the data center boom is 8.3 months and their demand is described as “continually softening.” A sub who can start next week may simply be well-run and local — or may be the one that market is passing by. Ask for three current references on healthcare-occupancy work and confirm the crew that did them still exists.
4. Separate your floating-rate exposure from your permanent financing assumption. The construction loan repriced on September 16; the take-out may not move the same direction. Ask your lender for the all-in carry cost of a four-week and an eight-week schedule extension in writing, and put that number in front of every scope decision. Schedule slip is now a financing line item, not a scheduling inconvenience.
5. Buy design resolution before you buy anything else. The Suffolk/MIT structure — design automation as the upstream enabler of permitting, procurement, scheduling, and subcontractor coordination — holds at clinic scale even though the percentages do not. Resolve ceilings, med gas routing, structural penetrations, and equipment cut sheets before the permit set goes out. Every unresolved item becomes an RFI, then a re-quote at a higher index, then a slip on a more expensive loan.
6. Assume this is not the last rate move. Guckes is explicit that the Fed moves in waves, and Roberts sees no cost relief in the next six months. Build the go/no-go around a decision date rather than a market forecast: name the date by which scope, buyout, and financing must be settled, and treat every week past it as a priced event rather than a delay.

The Takeaway

There is a specific way that a good owner loses money this quarter, and it does not look like a mistake while it is happening. It looks like prudence. A national headline says starts fell almost 25%. The Fed raises rates. The instinct — a reasonable, experienced, defensible instinct — is to pause, let the market come to you, and re-bid in ninety days into a softer field.

The data from this single week says that field will not be softer in Houston. The only region in the country that grew is the one you are standing in. The specialty trades you need are under contract to a buyer at a record rate, and that buyer does not borrow money. The input-cost index is running 8.9% a year, so the ninety days is not free. And the rate move that prompted the pause is, by the chief economist who covered it, more likely the first of several than the last.

Healthcare's 96.1% monthly bounce is not a boom, and this brief will not pretend otherwise — institutional starts are still down on the year. But that is exactly the point. A medical, dental, or veterinary owner is not the force setting Houston's price for labor or for gear. That does not make them exempt from it. It makes them the participant with the least leverage and the most to gain from moving early and deliberately.

The projects that come in on schedule and on budget out of this quarter will not be the ones that read the market best. They will be the ones that resolved the design, locked the constrained trades, priced the cost of waiting honestly, and named a decision date and kept it.

Forward Always.