Construction Loans: Draw Schedules and Interest Reserves — A Practitioner's Guide

Key Takeaways

Why Construction Loans Behave Differently

A construction loan is not a stabilized real estate loan with a different label. Its sizing constraints, draw mechanics, interest accrual, and risk profile are all materially different from permanent debt. Get the differences wrong and the deal looks underwriteable when it isn't — or, more commonly, looks more expensive than it actually is because the interest reserve is being sized off the wrong formula.

The structural differences come down to five facts. Construction loans are short-term (typically 24–36 months, including a lease-up tail). They are interest-only — principal isn't amortized during construction. They are floating-rate, typically priced as SOFR plus 275–400 basis points in May 2026, per the NAIOP Sentiment Index and MBA's CREF Forecast. They are funded in draws against construction progress, not advanced in full at closing. And they are sized to stabilized loan-to-cost and projected debt yield at stabilization, not to in-place NOI — because there is no in-place NOI during construction.

The article walks the five facts through one running deal example: a $50M multifamily ground-up in a Sun Belt metro, 30-month construction period, $50M loan at SOFR plus 325 bps. The interest reserve, S-curve, and circular reference between capitalized interest and total project cost get carried through end to end.

Anatomy of a Construction Draw Schedule

A construction loan funds in periodic draws against construction progress. The initial draw at closing typically covers land takedown plus pre-closing soft costs (design, permits, due diligence, financing fees). Subsequent monthly draws fund hard costs (materials, labor) plus continuing soft costs (architect, owner's rep, construction management fees, capitalized interest itself) against documented progress — usually an AIA G702/G703 contractor's application for payment, plus owner-controlled soft cost lines.

Three mechanics matter to the underwriter:

• Retainage. Lenders typically hold back 10% of each hard-cost draw, released at substantial completion. The retainage protects the lender against subcontractor liens and incomplete work; from the sponsor's perspective, it's working capital tied up for 18–30 months. On a $30M hard cost budget, $3M of retainage sits with the lender (or in escrow) until the certificate of occupancy issues.

• Lien waivers. Each draw requires updated unconditional lien waivers from the GC and material trade contractors. Lien-waiver failures — a subcontractor refusing to sign because of a fee dispute, or a supplier filing a mechanic's lien — can pause draws and trigger covenant defaults. The institutional convention is monthly compliance review by the lender's construction consultant.

• Lender inspections. The construction lender funds an independent third-party inspector (Construction Management Consultant or CMC) who reviews each draw against the schedule and certifies physical progress. The CMC's report is what the lender's credit committee relies on; the sponsor pays for it (typically $50K–$150K annualized for the full construction period).

The S-Curve and Why Front-Loading Matters

Construction draws don't fund linearly. The cash flow curve follows the classic S-shape: a slow build during sitework, demolition, and foundation; a steep climb during vertical construction (framing, MEP rough-in, envelope); and a tapered finish through interior finishes, FF&E, and stabilization. For a typical 30-month multifamily construction:

The S-curve matters because interest accrues only on the drawn balance, not on the loan commitment. A $50M loan that's drawn linearly to 100% by month 30 has an average outstanding balance of 50% over the period — producing one number for the interest reserve. The same $50M loan drawn on the S-curve above has an average outstanding balance closer to 62% — producing a materially different number. Front-loading the curve (heavy early draws for site acquisition, demolition, foundation) further raises the average. Sponsors who use the lazy 50% assumption systematically understate the interest reserve required.

Interest Reserve Sizing, the Right Way

Two methods are in active use. The simple-average method takes a rough estimate of average outstanding balance, multiplies by rate and tenor:

For a $50M loan at 7.75% all-in over 30 months at 50% average outstanding, the simple-average formula gives: (0.50 × $50M × 0.0775) × (30 ÷ 12) = $4.84M. This is the number that appears in many pre-LOI memos.

The draw-by-draw method projects the actual S-curve of draws month by month, computes interest on the actual outstanding balance each month, and sums. For the same loan at a 62% true average: ($31M average outstanding × 0.0775 = $2.40M annually; over 30 months = $6.00M).

The gap — $4.84M vs $6.00M — is $1.16M, or 24% understatement on a $4.84M reserve. On a $50M project, that's 2.3% of total project cost that has to come from somewhere when the construction loan runs short. Practitioners who used the lazy formula in 2021–2022 are now finding the gap exposed in real-time as construction periods extend and floating-rate exposure compounds.

Adventures in CRE's construction draw interest calculator is one of the best free models on the open web and uses the draw-by-draw method correctly. PropertyMetrics's interest reserve walkthrough uses the simple-average formula and explicitly flags it as approximate — though many readers miss the caveat.

Worked Example: $50M Multifamily Ground-Up

A 220-unit garden-style multifamily ground-up in Phoenix. Total project cost $62M: $12M land, $42M hard costs, $4M soft costs (excluding capitalized interest), $4M anticipated capitalized interest. Construction debt is a $50M facility at SOFR + 325 bps (May 2026 SOFR is approximately 4.55% per the New York Fed SOFR data), giving an all-in cost of 7.80%. The remaining $12M is sponsor equity. The construction period is 30 months, including a 6-month lease-up tail.

The draw schedule, simplified to six-month buckets:

The total interest accrued over the 30-month construction period is approximately $5.15M, with an effective average outstanding balance of $26.4M / $50M = 53%. (The example uses smoothed buckets; a real monthly schedule with a steeper front-loaded curve typically produces an average closer to 60–65% as discussed above.) For comparison, the simple-average formula at 50% would give $4.84M — understating by $310K, or 6%.

On a real institutional underwriting, the interest reserve is sized to a slightly higher figure than the central-estimate interest accrual to account for: (a) construction delays, which extend the period at peak outstanding balance; (b) SOFR moves during the construction period (rate caps on the construction loan typically have meaningful basis between cap strike and current SOFR); (c) cost overruns on hard cost that extend draws. A 15–20% contingency on top of the central interest estimate is the institutional norm, bringing the budgeted reserve in this example to roughly $6.0M.

The Circular Reference Problem

Construction loan interest accrues on the drawn balance, which includes the interest reserve, which is drawn monthly to pay the interest, which increases the drawn balance, which increases the interest. This is a circular reference, and it has to be resolved by iterative calculation.

In Excel, this means turning on iterative calculation (File → Options → Formulas → Enable iterative calculation, set to 100 iterations, max change 0.001). The model then computes the interest reserve that solves the circular relationship: the reserve covers all interest accrued including the interest accrued on the reserve itself.

A common shortcut: assume the interest reserve is a fixed input rather than a computed output. Set it at, say, $5.5M, and the model doesn't have a circular reference. The math works but the answer is wrong — the reserve doesn't actually cover all the interest unless you happen to guess right. Practitioner-grade models expose the circular reference explicitly and resolve it with iterative calc.

The other approach — and the institutional convention on most LP-quality models — is to compute the interest reserve as an explicit output of the draw schedule, sized to cover the central-case interest accrual plus contingency. The circular reference is acknowledged but bypassed by treating the reserve as a sources-and-uses line item that the model sizes once at acquisition and doesn't iterate on. This sacrifices some precision for transparency — you can see the reserve and the interest accrual on the same page.

When the Reserve Burns Through

Construction loans assume the interest reserve covers all interest accrued during construction. When the reserve runs out before substantial completion, the lender has three options, none of which the sponsor wants to live through:

• Borrower-funded interest. The sponsor writes a check from equity each month to cover accrued interest until construction completes. This is the standard remedy in mid-stage burn-through; the sponsor's equity check effectively grows by the shortfall. On a 6-month delay in the worked example, sponsor-funded interest might be $400–$500K of unplanned equity.

• Reserve replenishment from contingency. If the budget has remaining hard-cost or soft-cost contingency, the lender will permit a rebalance — moving unused contingency into the interest reserve. This requires lender consent and amends the loan agreement. Works when contingency is intact; rarely available at the back of construction when contingency has already been spent.

• Loan default / forbearance. When neither sponsor equity nor contingency is available, the lender enters workout. Practical resolutions include extending the maturity (with sponsor capital replenishment as condition), partial deed-in-lieu, or in the worst case, foreclosure during construction. The post-2022 vintage of value-add construction deals on bridge debt has produced a meaningful workout pipeline; the Trepp CRE CLO delinquency data tracks the trend.

Tying It to Sources and Uses

The interest reserve sits in sources and uses as a use of loan proceeds — the lender funds the reserve out of the construction loan itself. The relationship matters for LTC sizing: if total project cost is $62M and the construction loan is $50M, the LTC is 80.6%. But if the interest reserve is undersized at $4.84M rather than the true $6.0M, the project actually needs $63.2M of total cost, dropping the effective LTC to 79.1% and forcing $1.16M of additional sponsor equity.

The same dynamic affects the development spread and yield-on-cost. Stabilized NOI of $4.5M at a 6.5% exit cap implies a $69.2M stabilized value; against a $62M cost basis, the development spread is 0.95% (the gap between yield-on-cost at stabilization and the going-in cap at exit). An additional $1.16M of cost from undersized interest reserve pulls the development spread to 0.92% — small in isolation but consequential as it compounds with other "small" cost overruns through the construction period.

The 2026 Rate Environment

Construction debt pricing in May 2026, per NAIOP's Sentiment Index and MBA's CREF Forecast:

• SOFR approximately 4.55% per New York Fed daily reference rates
• Construction spreads running SOFR + 275–400 bps for institutional sponsors with stabilized rent comps; SOFR + 400–600 bps for first-time or less-capitalized sponsors
• All-in coupons roughly 7.30–8.55% on bank construction debt; debt funds at SOFR + 600+ bps (~10.5%+ all-in) for non-recourse balance-sheet construction
• Rate caps are now a standard construction loan requirement; a 30-month cap at a 1% strike above SOFR runs ~75–125 bps of upfront premium on a $50M loan

The implication for interest reserves: at 7.80% all-in (vs 4.30% in early 2021), the central interest accrual on the worked example is roughly 80% larger than it would have been five years ago. Reserves are 30–50% larger in dollar terms for equivalent project economics. Sponsors who built models in 2020–2021 with static rate assumptions and haven't refreshed the reserve calculation for the current environment are routinely misallocating $1–2M of project cost on a typical mid-market multifamily ground-up.

Six Mistakes Practitioners Make

• Using 50% average outstanding without checking the draw curve. Vertical construction front-loads more than 50%; horizontal site work back-loads more than 50%. The right number is the average of the actual S-curve, not the convention.

• Treating SOFR as a static input. Construction loans are floating-rate. Sponsors who modeled SOFR at 4.5% for the duration in early 2024 and didn't size a rate cap discovered material increases in interest accrual in late 2024 / early 2025 when SOFR moved. Model the rate cap as a real cost and the rate exposure as a real risk; use the Chatham Financial rate-cap pricing curves for current cap costs.

• Forgetting that retainage is working capital tied up. 10% retainage on $42M of hard cost is $4.2M of cash that the sponsor (or lender) finances during construction. If the loan funds retainage through draws, the financing cost on the retainage is part of capitalized interest; if the lender holds retainage in escrow, the sponsor's working capital position bears it directly.

• Sizing the construction loan to in-place NOI. There is no in-place NOI during construction. Construction loans are sized to a combination of LTC (typically 65–80%) and projected stabilized debt yield (typically 8.0%+ for the lender to take the construction risk). Anchoring LTV analyses against in-place NOI confuses the construction loan sizing with the permanent debt sizing.

• Underestimating soft cost burn-in. Soft costs (architect, owner's rep, financing fees, legal, permits, marketing, insurance) accrue continuously through construction and lease-up — not on the same S-curve as hard costs. Soft cost contingency at 5–10% of soft cost line items is the institutional convention.

• Treating the construction loan as the takeout. The construction loan matures at substantial completion or earlier; the deal needs a permanent loan or sale before that maturity. Sponsors who don't have a credible take-out plan in the underwriting fund themselves into a refinancing gap when the construction loan matures and permanent debt isn't available on the assumed terms. The 2026 vintage is more sensitive to this than 2018–2021 vintages because take-out debt costs more.

How to Model It in Excel

A practitioner-grade construction loan model has four tabs or sections:

• 1. Sources and uses. All-in project cost broken into land, hard costs, soft costs, capitalized interest, and contingency. Sources are equity, construction loan, and any subordinated debt or mezzanine. Reconciles to total cost. Capitalized interest is computed elsewhere and flows in.

• 2. Draw schedule. Month-by-month draws split by hard cost line, soft cost line, retainage release, and capitalized interest. Cumulative loan balance is the running sum. Average monthly balance is the input to the interest accrual.

• 3. Interest accrual. Monthly accrual = (prior-month balance + half of current-month draw) × (annual rate ÷ 12). Sum the monthly accruals to get total capitalized interest. The circular reference between the interest reserve (in sources & uses) and the capitalized interest (a draw line) is resolved with iterative calculation enabled.

• 4. Sizing constraints. The loan is the lesser of (a) LTC % × total project cost, (b) debt yield % × stabilized NOI ÷ projected exit cap, and (c) absolute commitment cap. The binding constraint is usually LTC for stabilized cash-flowing assets and debt yield for value-add or development.

Do It in Apers

Related Articles

• Cap Rate Calculator and Formula — the exit-yield metric that drives takeout sizing.
• IRR Calculator and Formula for Real Estate — the development IRR these reserves feed into.
• Cash-on-Cash Return: Levered vs Unlevered — how the levered return changes during stabilization.
• IRR Sensitivity Analysis and Stress Testing — how to stress-test the development assumptions.

FAQ