Everyone stopped building. That is the part of the 2026 multifamily story that is not in dispute: construction starts have collapsed to a 15-year low, and most owners read that as a warning. I read it as a signal — but a narrow, specific one, not the lazy "national supply cliff is coming" version you'll hear at every conference this year. The freshest data cuts against the simple cliff story, and a sophisticated sponsor will catch anyone who tells it.

So here's the honest version of the thesis, with the caveat built in: starts cratered, which thins the future pipeline. But completions through 2028 were just revised up, not down, because the under-construction pipeline is deep and lead times run about 24 months — there is no clean near-term national vacuum. The real scarcity is specific. It's in market-rate product, and it's geographic — the Sun Belt overbuilt and is still digesting it, while the Northeast and measured secondary Midwest and Southeast submarkets never had the supply problem. The opportunity that follows from that is precise, not broad: break ground now, in a submarket that never overbuilt, to deliver into a thinner 2028 market-rate window — and you have to start now, because the 24-month clock is the whole point.

This is a piece about ground-up construction financing, not bridge. If you want the bridge market read — rates, lenders, when bridge is the right tool — that's the Newark $5.55M bridge closing post; a construction takeout is one of the things bridge solves, but I'm not going to re-teach bridge mechanics here. This is about whether to finance a new building, and whether the math works.

The data, stated honestly

The headline is real. Multifamily starts in Q1 2026 ran roughly 55,000 units — the lowest quarterly level since 2011, about 73% below the early-2022 peak (CoStar, via Commercial Observer, May 2026). For scale: starts peaked around 708,000 units in 2022 (CRE Daily, December 2025). Capital got expensive, construction costs stayed high, and the math stopped working for a lot of would-be developers. That's why starts fell — hold that thought, because it's also the counter-case.

Here's where the lazy version breaks. If you straight-line "starts collapsed" into "supply vacuum in 2027," the data embarrasses you. Completions through 2028 were revised up, not down. Yardi Matrix raised its completion forecasts for 2026, 2027, and 2028 by 6.4%, 8.1%, and 8.9% respectively — to roughly 459,000 units in 2026, 440,000 in 2027, and 448,000 in 2028 (Yardi Matrix, February and May 2026). The reason is mechanical: the deals already under construction are a deep backlog, and at a ~24-month lead time, what delivers in 2026–2028 was largely committed before starts fell off a cliff. The collapse in starts doesn't show up as a national shortage until after that backlog clears. So there is no clean, broad national vacuum in the near term — and anyone selling you one is selling.

The scarcity that is real and defensible is narrower:

  • It's market-rate-specific. New market-rate supply in 2028 is projected to run about 31% below 2025 levels (Yardi Matrix, May 2026). That's the strongest "vacuum" stat available, and it's the one that matters for a market-rate ground-up deal — affordable and other product categories distort the all-in completion totals.
  • The timing lines up with the recovery. A project you start in 2026 delivers into 2028, and Yardi's conservative asking-rent forecast has growth at roughly 0.5% in 2026, 1% in 2027, and 2.3% in 2028 (Yardi Matrix, May 2026). I'm deliberately anchoring to Yardi rather than the more bullish 1.5%/2.2% numbers floating around; if you want to flag the range to an investor, fine, but underwrite to the conservative figure. The point stands either way: rents are recovering on the timeline your building would deliver into, not the timeline you break ground on.

That's the case, with the caveat showing: not a national cliff, but a market-rate-specific, geographically concentrated thinning that a 24-month build is positioned to catch.

Where the thesis actually works — and where it doesn't

Geography is doing most of the work here, so be specific about it. The Sun Belt overshoot metros — Austin, Phoenix, Nashville, Dallas, Denver — absorbed record deliveries and face a long absorption runway, while the Northeast and measured Midwest are positioned to stabilize sooner (Matthews, March 2026). Translate that into a build decision:

  • Build into submarkets that never overbuilt. Northeast secondary markets — the New Jersey / New York secondary submarkets we work in — and selective, measured Southeast and Midwest secondary markets where the pipeline was always thin. These are the places where a 2028 delivery meets a market that's short on new market-rate product instead of swimming in it.
  • Do not break ground into the overshoot metros. A new building delivering into Austin or Phoenix in 2028 is competing against a glut that's still being absorbed — that means concessions, slow lease-up, and rent growth that arrives late, exactly when you need stabilization. The same site plan that pencils in a thin Northeast submarket can be a value trap in an overbuilt Sun Belt one. Same product, different market, opposite outcome.

The discipline is the differentiator. This thesis is not "build multifamily." It's "build market-rate multifamily in a submarket that didn't overbuild, to deliver into 2028." Miss any of those three and you're not executing the thesis — you're just developing into a soft market.

The construction loan as the vehicle

Ground-up is financed differently from an acquisition or a stabilized perm, so a quick orientation on the structure (the specific terms for a representative deal are in the worked example below, pending live numbers):

  • Loan-to-cost (LTC), not loan-to-value. A construction loan sizes off total development cost — land, hard costs, soft costs, contingency, carry — because there's no stabilized value to lend against yet. LTC is the lever; the balance is your equity.
  • Draw structure. You don't get the loan in a lump sum. The lender funds in draws against construction progress, inspected and released as work completes, so interest accrues only on what's actually advanced. Your cost basis builds over the construction period rather than on day one.
  • Interest reserve. Because a building under construction produces no income, the loan typically carries an interest reserve — funded into the loan — to service the debt through completion and the start of lease-up. Whether the reserve is funded into the loan or carried separately changes your equity requirement, so it's a term to pin down, not assume.
  • Term and extensions. A construction loan is term-limited to the build-plus-lease-up horizon, usually with extension options to bridge to a stabilized takeout. The exit — a permanent loan or a sale — is what you're underwriting toward from the first draw.

The point for this thesis: the construction loan is the instrument that lets you commit to a 24-month delivery now. It's also where the discipline lives, because every one of those levers — LTC, draw pace, reserve, term — has to be sized against a takeout you've actually verified. Which brings us to the math.

Worked example

⚠️ PLACEHOLDER — DO NOT PUBLISH AS-IS. This example is intentionally blank. Per the brief (MUST-FIX #1), the construction-loan terms and the representative project inputs are [DOMINICK TO SUPPLY] — they are not to be invented or pulled from memory. Once the inputs below land, every dependent figure derives from a single driving assumption (total development cost + stabilized rents), so the recompute point is one place. Verify the market exit cap rate for the chosen submarket at draft.

Inputs — [DOMINICK TO SUPPLY]:

  • Representative project: unit count · TDC per unit (→ total development cost) · stabilized monthly rent per unit · vacancy assumption · operating-expense assumption · target market/submarket
  • Construction loan: loan-to-cost (LTC) % · rate + index · whether the interest reserve is funded into the loan · term + extension options · draw structure
  • Market exit cap rate for the chosen submarket (verify at draft)

Math chain (to be computed once inputs land):

Stabilized NOI    = (units × monthly market rent × 12 × (1 − vacancy)) − operating expenses
Yield on Cost     = Stabilized NOI ÷ Total Development Cost
Development spread = Yield on Cost − Market exit cap rate          ← the key result
Construction loan = Total Development Cost × LTC
Equity required   = Total Development Cost − Construction loan

The number that decides the deal is the development spread — yield-on-cost minus the market exit cap rate. That spread is your compensation for taking development risk instead of buying a stabilized asset; if it's thin, the build doesn't justify the risk no matter how good the supply story sounds. We'll trace every figure back to TDC and stabilized rent so a reviewer can check the whole chain from the two driving assumptions.

The honest counter-case

I won't pretend this is a clean win, because three things genuinely cut against it.

Building is hard — and that difficulty is the reason starts fell in the first place. The 15-year-low in starts isn't a market oversight you get to arbitrage costlessly; it's the rational response of developers to high construction costs, expensive capital, and execution risk. Cost overruns, entitlement delays, and lease-up slippage all land on the equity, and a 24-month build is 24 months of things that can go wrong. The supply window is real, but it's compensation for taking on exactly the risk that scared everyone else out. Go in with controlled costs and a real contingency, or don't go in.

The takeout gap is real, and it's worse for small deals. A construction loan is not a permanent home — it's a term loan you have to refinance out of at stabilization. And the agency perm that everyone assumes is waiting may not be. As of April 15, 2026, Freddie Mac retired the standalone Small Balance Loan program and folded it into Conventional Small, which carries a $2 million loan floor. A ground-up deal that stabilizes to a sub-$2M permanent loan has no clean agency takeout — Conventional Small won't write it at all. That sponsor needs a community-bank or credit-union perm, a debt-fund mini-perm, Fannie Mae Small, a sale, or to have built large enough to clear $2M in the first place. This is the same DSCR-and-floor reality we traced for value-add exits in The Takeout Is the Plan, and it applies with full force to ground-up: size the takeout before you size the construction loan. (And if your maturing balance lands below $2M for any reason, the refi-gap math is the same problem from the other side.)

The recovery is conservative, by design. I anchored rent growth to Yardi's low numbers on purpose. If you need 4% rent growth at delivery to make the spread work, you don't have a deal — you have a bet on a forecast more bullish than the data supports. The deals that work here work at 2.3% 2028 growth, or they don't work.

The window belongs to the disciplined sponsor

The build-year thesis is not a green light to develop. It's a narrow opening for a specific sponsor: one with a real site in a submarket that never overbuilt, controlled and contingency-backed costs, a market-rate product that fits the thinning 2028 window, and a permanent takeout already sized — including an honest answer to the sub-$2M floor. Hit all of that and the 24-month clock works for you instead of against you. Miss the discipline and the same clock buries you in an overbuilt market with no clean exit.

That disciplined sponsor is exactly who we finance. If you have a ground-up small multifamily site under control and you're trying to figure out whether the development spread is real after construction costs and a verified takeout, send me the project budget, the rent assumptions, and the submarket. We'll size the construction loan against a takeout we've actually confirmed — and if the spread isn't there, or the stabilized loan lands under $2M with no clean agency perm, we'll tell you that on the first call.

Dominick Prevete, Founder — Blue Sky Capital Advisors (908) 220-6404 · dominick@nationalloanprovider.com 31 years in real estate finance, ~$2B in sales volume led, 100+ bank and private-lender relationships. Lending in all 50 states. NMLS information available upon request.

FAQ

Is now actually a good time to build multifamily? It depends entirely on where. Nationally, the picture is mixed: multifamily starts cratered to a 15-year low in Q1 2026 (about 55,000 units, the lowest quarterly level since 2011), but completions through 2028 were revised up, not down, because the existing under-construction pipeline is deep and takes roughly 24 months to deliver. So there is no clean national supply cliff. The real scarcity is two things at once — it's specific to market-rate product (new market-rate supply in 2028 is projected about 31% below 2025 levels per Yardi Matrix, May 2026), and it's geographic. Building can make sense in a Northeast secondary submarket or a measured Southeast secondary market that never overbuilt, where you'd deliver into a thinner 2028 window. It rarely makes sense to break ground into a Sun Belt metro that overshot — Austin, Phoenix, Nashville, Dallas, Denver — that is still digesting record deliveries.

Why start a construction project now if rents are still soft? Because of the lead time. A ground-up multifamily project takes roughly 24 months to complete (garden-style averaged about 722 days per Yardi Matrix, September 2025), so a deal you break ground on in 2026 delivers into 2028 — not into today's soft market. Yardi's conservative forecast has asking-rent growth at roughly 0.5% in 2026, 1% in 2027, and 2.3% in 2028, so the recovery is timed to your delivery, not your groundbreaking. The 24-month clock is the entire reason the window is a now decision: by the time rents are visibly recovering, it is too late to start a building to catch it.

Which markets should I avoid building into in 2026? The Sun Belt metros that absorbed record deliveries in 2023–2024 and are still working through them — Austin, Phoenix, Nashville, Dallas, and Denver are the commonly cited overshoot markets (Matthews, March 2026). Those submarkets face a long absorption runway, which means concessions, slow lease-up, and compressed rent growth precisely when your new building would be trying to stabilize. The thesis here is the opposite: deliver into submarkets that never overbuilt, where the new pipeline is thin and absorption is steady.

If I build a small project, can I refinance it into an agency loan when it stabilizes? Maybe not — and this is the most overlooked risk in small ground-up. As of April 15, 2026, Freddie Mac retired the standalone Small Balance Loan program and folded it into Conventional Small, which carries a $2 million loan floor. A ground-up deal that stabilizes to a permanent loan below $2M cannot use Conventional Small at all. That sponsor needs a community-bank or credit-union perm, a debt-fund mini-perm, Fannie Mae Small, a sale, or a deal sized large enough to clear the $2M floor. Plan the permanent takeout before you size the construction loan, not after the building is finished.

Loans are for business purposes only and are not subject to TILA, RESPA, or HOEPA. Not for primary residences. Equal Housing Opportunity. All loans subject to underwriting approval. Rates and terms shown for illustration; actual rates depend on deal specifics. We do not lend to borrowers with credit below 600 or on owner-occupied properties.