Ask ten investors why DSCR rates sit 150 to 250 basis points above agency mortgage rates and most will say "risk." That is half true and mostly unhelpful. The real answer is that a DSCR loan is priced backward from a bond: nearly every DSCR loan originated today is destined for a non-QM securitization or a whole-loan sale to an insurance company or asset manager, and the rate on your term sheet is the exit price of that bond minus the margin everyone in the chain needs to make. Understanding that chain explains why quotes move when Treasuries have not, why a prepayment penalty buys you 40 basis points, and why two lenders can be 62.5 basis points apart on the identical file. The mechanics of the loan itself — how the ratio is calculated and what the appraiser's rent schedule does to it — are covered in how DSCR loans work; this piece covers the machine behind the rate sheet.
The pipeline: originator, aggregator, bond
A DSCR loan passes through three hands. The originator — the shop quoting you — funds the loan on a warehouse line, holds it for 15 to 45 days, and sells it. The buyer is usually an aggregator: a large non-QM platform or a REIT that pools a few hundred million dollars of loans from dozens of originators. The aggregator then either securitizes the pool — issuing rated bonds against it in an RMBS structure — or sells whole loans directly to insurance companies hungry for yield. Each hand takes a slice. If the AAA tranche of a new non-QM deal clears at a yield near 5.7% and the blended cost across the full capital stack lands around 6.3%, the aggregator needs roughly 75 to 100 basis points to cover credit enhancement, deal costs, and profit, and the originator needs 100 to 150 to cover operations and warehouse carry. Stack it up and the borrower-facing coupon lands in the high-7s to mid-8s — which is exactly where the market sits, as tracked in the Q3 2026 rate update.
Spread, not Treasuries, is the number that matters
Investors watch the 10-year Treasury and are then baffled when their quote moves the other way. The missing variable is spread: the premium bond buyers demand over the risk-free rate to hold non-QM credit. Non-QM AAA spreads have traded in a band of roughly 110 to 200 basis points over Treasuries in recent years, and that band does the work. When spreads tighten 30 basis points because insurance demand for whole loans is strong, DSCR rate sheets improve within a week or two even if Treasuries went nowhere. When a volatile week widens spreads, lenders reprice overnight — always faster on the way up. This is why rate-shopping quotes that are more than a few days old is a mistake, and why the honest answer to "where will DSCR rates be in six months" depends as much on securitization demand as on the Fed. The practical takeaway: lock when the deal pencils, and treat any quote older than five business days as expired.
Why your loan's attributes are worth real basis points
Securitization is also why the rate sheet is a grid rather than a number. Bond buyers pay measurably more for pools with specific characteristics, and that premium flows backward to you as pricing adjustments. The cleanest example is the prepayment penalty: a 5-year step-down penalty makes the bond's cash flows more predictable, so pools of penalty-protected loans price better, so accepting a 5-4-3-2-1 structure instead of a 3-year typically saves 25 to 40 basis points — the full menu is in the prepayment penalty structures guide. Same logic on the other grid cells: a 1.25+ DSCR is worth roughly 25 basis points over a 1.00 to 1.10 file, 65% LTV prices 25 to 50 inside 80%, and a 760 FICO can sit 75 to 125 basis points inside a 680. None of this is your lender being generous or greedy — it is the bond market's preference schedule passed through to the rate sheet. Paying discount points to buy the rate down interacts with the same math, and the break-even analysis is laid out in the points and buydown guide.
Why lenders disagree by 62.5 basis points on the same file
Take one file — 75% LTV purchase, 1.15 DSCR, 740 FICO, single-family rental in Tampa — and shop it to six lenders. Quotes will commonly span 7.50% to 8.125%. The spread is structural, not random. Shops differ in their exit: an originator selling to an aggregator with a strong insurance bid can price inside a shop that securitizes its own paper and carries the deal costs. They differ in warehouse economics — a lender paying SOFR plus 175 on its line has more room than one paying SOFR plus 250. And they differ in appetite: a platform that needs Florida exposure for a pool it is assembling will price Florida collateral aggressively for a month and then go flat. This is the capital-markets version of the tiering covered in top-tier versus specialty DSCR lenders: the biggest non-QM platforms usually win on clean files because their exit is cheapest, while specialty shops earn their wider coupon on the files the big platforms will not buy.
What the bond market will not take — and what that means for you
The securitization exit also explains most hard program floors. Rated deals need loans that fit the rating agencies' models, so attributes the models penalize become originator overlays: loan amounts under $100,000 to $125,000 get small-balance add-ons or outright declines because fixed servicing costs eat thin coupons; rural property, condotels, and unusual collateral get excluded because they comp poorly in a liquidation model; sub-0.75 ratios land in no-ratio buckets that only a handful of aggregators buy. When a lender tells you "we can't do that," the honest translation is usually "nobody will buy that from us." This is worth knowing in low-price-point markets — a $95,000 single-family in Cleveland can carry a beautiful 1.35 ratio and still struggle to find a lender, purely because of where the servicing math breaks. The workaround is scale: five such properties in a blanket structure clear the minimum-balance floor easily.
Reading a repricing cycle in real time
Put the machine together and you can read the market like an originator does. Treasuries rally but your quote does not improve: spreads widened, or the lender is holding margin ahead of a securitization pricing next week. Every lender suddenly wants 5-year prepay structures: bond buyers are paying up for call protection. A lender that was aggressive last month goes quiet: their aggregator filled its pool, and the bid moved to someone else — which is exactly why the lender directory matters more in this market than in agency lending, where every exit is the same two GSEs. Rate sheets in non-QM are living documents; the shop that was 40 basis points wide on your last deal may be the best bid on your next one, and re-shopping every transaction is not disloyalty, it is how this market is designed to be used.
The floor under DSCR rates
One last implication. DSCR rates cannot converge to agency rates no matter how benign the credit environment gets, because the structure will not allow it: agency loans carry a government guarantee that removes credit risk from the bond, while non-QM investors must be paid for taking that risk plus the origination chain's margin. The long-run floor for DSCR pricing is agency plus roughly 100 to 150 basis points in the tightest plausible spread environment. Anyone quoting materially inside that is either pricing a loss leader or quoting a different product. For investors, the planning assumption is simple: underwrite acquisitions at today's coupon, refinance if the spread cycle gives you a gift, and remember that the same securitization machine that sets your rate also guarantees the product's availability — the bid for DSCR paper is deep, institutional, and growing, which is why the product survived rate cycles that killed lesser loan types.