‘Refi’ meets AI: impacts for MBS investors

If any parts of the economy look truly immune to AI-related change, it is not the US mortgage industry.

High costs and inefficiencies make the mortgage refinancing market ripe for disruption. Indeed, some lenders are already beginning to leverage the potential of AI to lower the barriers to refinancing.

This has implications for investors in the US$9.3tn US agency mortgage-backed securities (MBS) market, for whom prepayment risk (reduced returns from borrowers paying off their mortgages earlier than expected) is more material than default risk.¹

Investors need to interrogate the potential impacts of AI adoption on specified pools of MBS that have bespoke characteristics. Given that certain specified pools will be more susceptible than others, we believe there are compelling opportunities for investors to differentiate themselves in the pursuit of prepayment risk protection.

How could AI disrupt the refinancing market?

For US homeowners looking at refinancing their loans to lower mortgage rates, today’s highly manual process may appear daunting – and prohibitively expensive.

The combined fixed and variable costs typically range between 2% and 6% of the loan value.² Based on a median US home sale price of US$400,000 and the assumption of 20% equity, refinancing costs could therefore range between US$6,400 and US$19,200.³ The process – involving application, verification, processing, underwriting and appraisal – also takes between 30 and 45 days.⁴

With the assistance of AI, mortgage lenders may be able to significantly lower the barriers to refinancing by reducing fixed costs and boosting borrower awareness through more targeted outreach. Applying AI should also allow lenders to free up servicer capacity and close more loans during times of increased refinancing volume.

There is precedent emerging within the consumer asset-backed securities (ABS) space, where AI-led underwriting is cutting loan application decisions from days to minutes.⁵ Online non-bank platforms, who are less encumbered by legacy systems and processes, appear particularly disposed to automating underwriting and approval processes. These so-called marketplace lenders account for a growing share of consumer ABS issuance (see chart below).

The direction of travel coincides with a rising financial incentive for borrowers – in terms of lower monthly repayments – to refinance at a time of rising average loan sizes and limits for conforming mortgages.

Source: Impax analysis of data provided courtesy of JP Morgan Chase & Co, Copyright 2026. Branch lending refers to asset-backed securities (ABS) pooling consumer loans, issued through branch networks by traditional financial institutions or specialised lenders. Marketplace (online) lending refers to ABS issued through online, technology-driven platforms/origination.

Why does this matter for MBS investors?

Should this trend gather momentum, an uptick in mortgage refinancing could materially change the landscape for agency MBS – pools of securitised residential mortgage loans that are issued and guaranteed by US government agencies including Fannie Mae and Freddie Mac.

Prepayments and the timing of principal returns are the primary risks when investing in agency MBS given their implicit government backing. When a mortgage within a mortgage-backed security is refinanced, the loan is fully repaid and comes through as a full prepayment to the investor.

Generally speaking, prepayment volatility is undesirable because it disrupts the timing of expected cashflows and typically forces reinvestment at lower rates.

For this reason, agency MBS investors will typically look to shield themselves from prepayment spikes. By purchasing specified pools – MBS that have specific characteristics or ‘stories’ that have historically shown slower prepayment rates – investors can seek protection against prepayment risk.

The most common (and arguably most effective) example of these stories are low loan balance pools. After all, borrowers with relatively small mortgage values will typically have lower incentives to refinance given the relatively fixed costs involved.

To illustrate this, let’s compare a US$600,000 mortgage with a US$200,000 mortgage, both refinancing from a 6.5% rate to 5.5%. If we assume a fixed refinancing cost of US$6,000 in both scenarios, the former would be in the money within 16 months. In contrast, it would take 47 months (almost four years) for it to make economic sense to refinance the smaller loan.⁶

However, it is entirely plausible that enhancements from AI adoption could significantly reduce refinancing costs – and payback periods – and so lower the effective barrier to refinancing small loans. While positive for borrowers, of course, it would erode the prepayment protection that low loan balance stories have historically provided.

Where could MBS investors seek prepayment protection?

There are two types of stories that offer more durability, in our view: borrowers with low credit scores and geographically concentrated pools.

The first are specified pools comprised of mortgage debt held by borrowers with FICO scores below 700 who face greater qualification challenges when applying for new loans. They also pay higher mortgage rates as a result of loan level price adjustments (LLPAs), which are fees applied to mortgages based on a borrower’s credit risk. Unless LLPAs are lowered, low FICO borrowers will still have more difficulty meeting underwriting requirements, and have less rate incentive if they do, even with enhanced AI capabilities.

The second are specified pools focused exclusively on mortgage debt in individual US states where state-specific regulations or geographic characteristics present hurdles to refinancing. In New York state, for example, a mortgage recording tax – which can exceed 2% of the loan value – is payable on most refinance transactions.

Another increasingly important differentiator in the AI era could be the servicer of the pool, who is responsible for collecting the borrower’s monthly payments and managing the loan over its life.

Servicers that adopt AI more effectively should be better able to target borrowers that would benefit from refinancing and proactively reach out to them. As a result, two identical pools could prepay at very different speeds depending on AI capabilities and how aggressively their servicer pursues refinancing

Where could this trend lead?

Implementing AI effectively and at scale will likely require significant investments from originators and servicers. Any transformational change in the US mortgage market will therefore take time.

Nonetheless, we believe the direction of travel is set, raising questions over the valuations of certain specified pools that have historically commanded a price premium.

It is possible that we will see a shift in demand towards other products, such as agency collateralised mortgage obligations (CMOs), which also bundle residential mortgages. CMOs can offer more structural protection against prepayment risk by redirecting prepayments to specific tranches, independent of the underlying borrowers’ behaviour or characteristics.

Given the size of the MBS market and the potential for AI-driven disruption, we believe that it is crucial for fixed income investors to be aware of the risks and opportunities emerging in this evolving landscape. Investors’ desire for prepayment protection is unlikely to be eroded as quickly as certain specified pools’ ability to offer it.

¹ Bloomberg data, 1 July 2026
² Freedom Mortgage, 2026: How Much Does It Cost to Refinance a Mortgage?
³ Federal Reserve Bank of St. Louis, May 2026: Median Sales Price of Houses Sold for the United States (MSPUS)
⁴ Rocket Mortgage, 2026: How long does it take to refinance a house?
⁵ Capgemini, 2026: AI-powered credit decisioning systems
⁶ Impax analysis, July 2026

References to specific securities are for illustrative purposes only and should not be considered as a recommendation to buy or sell. Nothing presented herein is intended to constitute investment advice and no investment decision should be made solely based on this information. Nothing presented should be construed as a recommendation to purchase or sell a particular type of security or follow any investment technique or strategy. Information presented herein reflects Impax Asset Management’s views at a particular time. Such views are subject to change at any point and Impax Asset Management shall not be obligated to provide any notice. Any forward-looking statements or forecasts are based on assumptions and actual results are expected to vary. While Impax Asset Management has used reasonable efforts to obtain information from reliable sources, we make no representations or warranties as to the accuracy, reliability or completeness of third-party information presented herein. No guarantee of investment performance is being provided and no inference to the contrary should be made.