Lenders including Rocket Mortgage, United Wholesale Mortgage, and Better.com are deploying artificial intelligence across their origination pipelines, compressing application-to-approval timelines from weeks into minutes. Better.com says it can process a refinance in as little as two minutes. United Wholesale Mortgage claims initial approvals now take fifteen minutes. Rocket Mortgage reports application-to-rate-lock times as short as thirty minutes and is targeting ten minutes as its next benchmark.
The speed upgrade is bigger than it sounds
Rocket Mortgage has put more than $500 million into AI and automation over the past six years, signaling this is a long-term infrastructure build rather than a feature launch. The result is a pipeline that can handle surges in demand without the manual bottlenecks that historically caused lenders to throttle volume during rate rallies.
Morgan Stanley strategists, including Jay Bacow, published a report on August 7, 2026 estimating that AI-driven efficiency could push refinance participation rates from roughly 33% of eligible borrowers during a rate dip to around 60%. If rates reached 5.5%, that uptake level could translate into approximately $2.1 trillion in market activity.
What faster prepayments mean for bond investors
The US mortgage-backed securities market is roughly $9 trillion in size, and it runs on a set of assumptions about how long borrowers will hold their loans. When homeowners prepay mortgages by refinancing, they return principal to MBS investors ahead of schedule. In a falling-rate environment, that is the worst possible timing: investors get their money back exactly when reinvestment opportunities offer lower yields.
Morgan Stanley’s strategists framed the implication directly: AI adoption could cause traditional 30-year fixed-rate mortgages to behave more like floating-rate instruments from an investor’s perspective. Baseline estimates from Morgan Stanley put the necessary spread adjustment at 10 to 20 basis points. On a $9 trillion market, even a modest spread widening represents a significant repricing of risk across institutional portfolios that hold agency MBS, including pension funds, insurance companies, and the Federal Reserve’s own balance sheet.
KBW published a separate analysis in July 2026 noting that the AI efficiency gains would primarily benefit larger lenders by improving operational throughput rather than disrupting the overall market structure. The big players with the capital to invest in AI infrastructure, the Rockets and UWMs of the industry, are positioned to capture outsized share of the refinancing volume surge while smaller originators lag behind.
A two-sided ledger
For MBS investors, faster prepayments shorten duration, which forces reinvestment at lower rates precisely when investors can least afford it. Wider spreads would partially offset this by delivering higher current income, but they also imply mark-to-market losses on existing holdings as the market reprices. If MBS spreads widen materially and persistently, the cost of new mortgage origination rises, which puts upward pressure on the rates that borrowers actually pay. Efficiency at the application layer does not automatically translate into cheaper credit if the bond market reprices the underlying risk.