Back in ‘22, when loanDepot was losing more than $200M a quarter, management made a series of painful decisions to stop the bleeding. One of the biggest: shutting down its wholesale channel. Three years later, the channel is back — suggesting the company may have some ragrets.

Brokers told The Mortgage Scoop Monday morning that loanDepot products were once again available to price on Loansifter.

The return to wholesale fits the broader reset underway at loanDepot since founder Anthony Hsieh returned to day-to-day operations. The lender is rearchitecting its tech stack w/ Dom Marchetti, restructuring its consumer direct division &, in many ways, bringing the band back together 🎸.

In today’s edition of The Scoop, we dive into loanDepot’s wholesale return. Plus, we unpack Better’s unbelievable claim about 47-second underwriting & share the freshest data on purchase v refi trends.

(🙏 If you like what you’re reading, tell a fellow mortgage junkie to sign up here.)

Matt Mancasola has returned to loanDepot following wholesale stints at Nation’s & Homepoint. He’ll answer to Dan Peña, who leads loanDepot’s JVs & other partnerships. 

Many brokers enjoyed working w/ LD back in the day. Pricing & products were good, they said. But the central question is this: Just how heavily loanDepot will invest in a channel they shut down only 3 years ago amid heavy losses? Trust doesn’t come easily in the broker space & loanDepot hasn’t yet hired any notable account executives, which is foundational to growing volume, multiple wholesale executives told The Scoop.

“Why would good AEs go back to a shop that shut down wholesale?” asked one exec. “Matty is good but I think he’s got an uphill battle to fight. We shall see.”

loanDepot did not immediately respond to a request for comment.

Agency mortgage securitization loan purpose type distribution shifted notably in Q4 ‘25, w/ refinance mortgages picking up steam. Purchase money production accounted for 62.4% of new securitizations, down from 79.8% in Q3 ‘25 and 69.5% in Q4 ‘24.

Downward trends in the 10-year Treasury rate and mortgage spreads over the past 12 months are contributing to an uptick in refinance activity. Rate/term refinances represented 26.6% of new loans in Q4 ’25, compared to 9.2% in Q3 ‘25 and 19.0% in Q4 ‘24. Meanwhile, cash-out refinances declined slightly to 10.9% of new loans, compared to 11.1% in Q3 ‘25 and 11.4% in Q4 ‘24. 

The consistent level of cash-out refinance is supported by a continual churn of borrowers who seek liquidity regardless of rates. Of those lenders issuing at least $1 billion in Q4 ‘25 (excluding builders and housing finance agencies), Cornerstone Capital Bank had the highest percentage purchase of their total production (92.2% of issuance) followed by PrimeLending (85.7%) and NFM (81.4%).

Of those lenders issuing at least $1 billion in Q4 ‘25, UWM had the highest percentage rate/term refinance of their total production (53.2% of issuance) followed by Freedom Mortgage (46.0%) and Village Capital (38.5%). Of those lenders issuing at least $1 billion in Q4 ‘25, loanDepot had the highest percentage cash-out refinance of their total production (26.1% of issuance) followed by Wells Fargo (26.1%) and Rocket Mortgage (22.9%).

This report is brought to you by Milliman. Milliman provides strategic and quantitative consulting services across the mortgage market with expertise in origination, servicing and capital investment.

Source: Fannie Mae, Freddie Mac, and Ginnie Mae MBS Data Disclosure

Better.com has a new claim that’s ricocheting around the mortgage industry: underwriting a loan in 47 seconds. I legit can’t get shoes 👟 on my toddler that fast!

The company told CNBC that its mortgage engine, paired with OpenAI models, can compress what normally takes about 21 days into less than a minute. If true, it would be one of the most dramatic productivity breakthroughs the industry has ever seen. But people I’ve spoken to across the mortgage tech ecosystem are reacting with a mixture of curiosity and… skepticism.

Better didn’t immediately respond to my requests for comment, but here’s how Better says it all works: Users log into a ChatGPT Enterprise account, download the Tinman AI credit decision engine & connect their guidelines, pricing & CRM system to process, underwrite & fulfill mortgage applications. Throw a file in & watch the magic 🪄 happen. OpenAI’s models save a crazy amount of time by simultaneously running parallel workflows on appraisals, title reports, income, credit reports & other docs, according to Better.

“AI is now doing mortgages,” is how CEO Vishal Garg described it to CNBC. “Rocket, UWM, Pennymac, a bunch of guys that are large public companies, make their money by effectively charging a tax of 1.5% to underwrite mortgages. … That’s $20B that’s paid by the American public in a typical year.” 

Better says it has matched the underwriting criteria & pricing parameters w/ “over 45 different institutional buyers of mortgage and home equity loans totaling over 80% of the US mortgage market.”

The skepticism is this:

Theo Ellis, who runs AI pre-underwriting company Friday Harbor (full disclosure: Friday Harbor is an advertiser), said his tool runs hundreds of large language model calls per loan application.

“A reasoning model can take two to three minutes to run a single call. So is this 47 seconds a real claim? Are you really extracting data, categorizing using LLMs & agentic solutions in 47 seconds? Or did you just put a chatbot in front of a deterministic system that requires a not-messy file in order to get that kind of response?"

(🙏 If you like what you’re reading, tell a fellow mortgage junkie to sign up here.)