TWO - NYSE

Good morning. My name is Jennifer, and I will be your conference facilitator. At this time, I'd like to welcome everyone to Two's Second Quarter 2025 Earnings Call. [Operator Instructions]. I would now like to turn the call over to Ms. Maggie Karr.

Good morning, everyone, and welcome to our call to discuss Two's second quarter 2025 financial results. With me on the call this morning are Bill Greenberg, our President and Chief Executive Officer; Nick Letica, our Chief Investment Officer; and William Dellal, our Chief Financial Officer. The earnings press release and presentation associated with today's call have been filed with the SEC and are available on the SEC's website as well as the Investor Relations page of our website at twoinv.com. In our earnings release and presentation, we have provided reconciliations of GAAP to non-GAAP financial measures, and we urge you to review this information in conjunction with today's call. As a reminder, our comments today will include forward-looking statements, which are subject to risks and uncertainties that may cause our results to differ materially from expectations. These are described on Page 2 of the presentation and in our Form 10-K and subsequent reports filed with the SEC. Except as may be required by law, does not update forward-looking statements and disclaims any obligation to do so. I will now turn the call over to Bill.

Thank you, Maggie. Good morning, everyone, and welcome to our second quarter earnings call. Please turn to Slide 3. Fixed income and equity markets proved resilient in the second quarter, rebounding from poor performance in early April as the uncertainty of fluctuating tariff and trade policies roiled markets, spiking the VIX index to a multiyear high. As the quarter progressed, the tariff tension eased and the macro environment recovered steadily, leading the S&P to a record high and a significant recovery in the performance of Agency RMBS spreads. We remain disciplined in our approach to risk, keeping our interest rate and spread exposures low across the curve. We utilized leverage judiciously and preserved ample liquidity, which allowed us to navigate these periods of heightened market volatility not seen since last October. For the second quarter, including the loss contingency accrual of $1.92 per share, we experienced a total economic return of negative 14.5% and minus 1.4% without the accrual. For the first half of the year, this results in a total economic return on book value of negative 10.3% and 2.9%, excluding the accrual. Please turn to Slide 4. The 10-year U.S. treasury rate ultimately settled near where it began the quarter, as you can see in Figure 1, but not before moving through a wide range from a low of 3.85% in early April to a high of 4.62% in late May. The spread between 10- year and 2-year U.S. treasuries widened to 51 basis points, creating a steeper curve that continues to support attractive opportunities for RMBS and MSR portfolios. In this environment, we believe returns are compelling, and we expect further strengthening of the supply-demand dynamic, potentially leading to spread tightening. Mortgage rates generally track the treasury rate environment, moving higher in April and May before stabilizing in June. The 30-year fixed rate mortgage rose from 6.6% at its low to a high near 6.9%, ending the quarter in the 6.7% to 6.8% range. While still high by recent COVID era standards, rates remained below their 2023 peak levels, which has helped housing activity remain reasonably well supported. The Federal Reserve maintained its cautious stance and left rates unchanged even in the face of increases in inflation and mounting political pressures. Several members of the FOMC have suggested 1 to 2 rate cuts likely occurring later this year, and the market similarly projects 50 to 75 basis points of cuts in the second half of 2025, as you can see in the blue line in Figure 2. If the Fed does indeed cut rates in the latter half of this year, we expect RMBS and MSR portfolios to respond positively. With the majority of our MSR portfolio still more than 300 basis points away from the refinancing window, we do not expect a few cuts in the front end of the yield curve to materially alter mortgage rates or prepayments. We are strengthening our direct-to-consumer originations platform at RoundPoint, consistent with the market opportunity in order to recapture loans in our portfolio that may refinance. Please turn to Slide 5. In the second quarter, we funded $48 million UPB in first liens, up from $29 million UPB in the first quarter. Although starting from a low base, this increase of 68% outpaced the overall trend in mortgage originations, which saw funded loans rising nationwide 16% quarter-over-quarter. We are encouraged by the growth in our first lien originations despite the fact that most of our portfolio does not have an economic incentive to move or refinance. Additionally, we continue to actively market second liens to our servicing customers to help them extract home equity most efficiently. We brokered $44 million UPB in second liens in the quarter, and we have begun originating second liens in our own name, which we can ultimately choose to hold, sell or securitize. This activity not only increases revenue and improves recapture rates, but we have also noticed significantly slower prepayments for our MSR borrowers who have second liens on top of their firsts. Please turn to Slide 6. I'd like to mention some of the really interesting things we are doing in technology in order to increase efficiencies, reduce costs and most importantly, create better homeowner experiences. We are not alone in seeing the large opportunity that AI technologies can bring to the servicing and origination businesses, and we are making significant investments in time and resources in order to achieve the benefits that these technologies promise. Our initial focus has been within our contact center, and we are currently implementing AI in many areas across the platform. We use human emulation bots to move data across applications and to perform other repetitive tasks. Image recognition utilizes OCR technologies to help perform data validation. Speech recognition applications allow us to perform comprehensive analysis and statistics on our customer service calls, and we are using generative AI technology to create automatic call summaries, which saves significant time for our contact center employees while improving accuracy. Conversational AI, which we are just beginning to explore, includes allowing customers to interact more fully with customized AI interfaces for simple situations and calling in live people for more complex problems. As we look towards the future, we are also actively evaluating the application of AI on the origination side to automate the application and fulfillment process. I'd like to say that AI is just the newest form of technology, and we know that this technology is integral to success in operating our business going forward. Looking ahead, we believe the combination of our investment portfolio and operating company allows us to be dynamic and responsive as opportunities emerge across the mortgage finance space. Given the strength of our platform and the depth of experience across our team, we are confident in our ability to navigate and lead through changing market cycles, creating long-term value for our stockholders, customers and business partners. With that, I'd like to hand the call over to William to discuss our financial results.

Thank you, Bill. Please turn to Slide 7. As Bill mentioned, in the quarter, we took a loss contingency accrual of $199.9 million or $1.92 per share related to the ongoing litigation from the termination of our management agreement with PRCM Advisers in 2020. On May 23, the court ruled that Two Harbors did not have grants to terminate its management agreement for costs. And so we, in consultation with our independent accountants and legal advisers, determined that the loss was not probable and estimable under ASC 450, which is the accounting standard that governs loss contingencies. The amount of the loss contingency accrual is the same $140 million that we initially reserved in 2020 related to the nonrenewal of the management agreement before reversing the accrual in connection with the subsequent termination for costs. The current accrual also includes an assumed statutory prejudgment interest at the simple rate of 9%. No other potential losses are probable or estimable at this time. We are waiting for a trial date to be set to resolve certain claims related to intellectual property and on the issues of potential damages for the contract termination. The parties have also agreed to participate in voluntary mediation. Including the accrual, the book value decreased to $12.14 per share, representing a negative 14.5% quarterly economic return. Excluding this accrual, our total quarterly economic return would have been negative 1.4%. Please turn to Slide 8. Including the loss contingency accrual, the company incurred a comprehensive loss of $221.8 million or $2.13 per share. Excluding the accrual, we would have incurred a comprehensive loss of $21.9 million or $0.21 per share. Net interest and servicing income, which is the sum of GAAP net interest expense and net servicing income before operating costs was higher in the second quarter by $3.1 million, driven by an increase in our Agency RMBS portfolio and higher float income on MSR. This was partially offset by lower servicing fee income from MSR portfolio runoff and slightly higher financing costs. Mark-to-market gains and losses were lower in the quarter by $93.4 million. As a reminder, this column represents the sum of investment securities, net gains and losses and change in OCI, net swap and other derivative gains and losses and net servicing asset gains and losses. In the second quarter, mark-to-market gains and losses were impacted by unfavorable market movements on MSR, swaps, TBAs and futures, partially offset by overall positive market movements on Agency RMBS. You can see the individual components of net interest and servicing income and mark-to-market gains and losses on Appendix Slide 22. Please turn to Slide 9. RMBS funding markets remained stable and available throughout the quarter with repurchase spreads at around SOFR plus 20 basis points. At quarter end, our weighted average days to maturity for our Agency RMBS repo was 60 days. We issued a baby bond in the second quarter with the intention of using the proceeds in part for the refinancing or repayment of our 6.25% senior notes due in 2026. In total, we issued $115 million aggregate principal amount of 9.38% senior notes that are due in 2030 for net proceeds of $110.8 million. We financed our MSR, including the MSR asset and related servicing advance obligations across 5 lenders with $1.8 billion of outstanding borrowings under bilateral facilities. We ended the quarter with a total of $837 million in unused MSR asset financing capacity. Our servicing advances are fully financed, and we have an additional $61 million in available capacity. I will now turn the call over to Nick.

Thank you, William. Please turn to Slide 10. Our portfolio on June 30 was $14.4 billion, including $11.4 billion in settled positions and $3 billion in TBAs. Our economic debt-to-equity increased to 7x, which includes the effect of the loss contingency accrual on our book value. We brought our debt-to-equity and mortgage spread risk down in early April in response to market volatility and spread widening. As volatility subsided, we brought our leverage and spread risk back up, and we feel comfortable with today's level given the attractive opportunities in both Agency RMBS and MSR spaces. As you can see in figures 2 and 3, we continue to manage our exposure to rates across the curve very closely. You can see more detail on our risk exposures on Appendix Slide 19. Please turn to Slide 11. Agency RMBS spreads to interest rate swaps widened meaningfully in April, tracking overall market volatility before retracing over the following 2 months. As shown in Figure 1, our preferred volatility gauge, 2-year options on 10-year swap rates peaked at 104 basis points in mid-April and declined to end the quarter at 94 basis points, 4 basis points lower than the end of the first quarter. Hedged Agency RMBS performance varied across the coupon stack with higher coupons outperforming lower longer duration coupons. Current coupon nominal spreads widened by 3 basis points to 171, while option-adjusted spreads finished 12 basis points wider at 81 basis points, reflecting the drop in implied volatility. As you can see in Figure 2, spreads across the curve, both nominally and on an option- adjusted basis, shifted up with larger pickups in OAS. At quarter end, considering the drop in implied rate volatility and that mortgage spread volatility had fallen to its lowest level in the post-COVID period, spreads versus swaps looked and continue to look very attractive on a historical and return potential basis. Please turn to Slide 12 to review our Agency RMBS portfolio. Figure 1 shows the hedge performance of TBAs and specified pools we own throughout this quarter. Higher coupons generally outperformed lower coupons and specified pools outperformed TBAs in the lower coupons we owned, while 6% and 6.5% TBAs outperformed specified pools. We have seen some strength in higher coupon dollar rolls, particularly in 6.5s, fueled by historically strong demand for CMO floaters with CMO issuance accounting for over 85% of the net issuance in 6s and 6.5s. On the margin, we shifted our exposure up in coupon in the quarter, which you can see in Appendix Slide 18. We also increased our exposure to mortgage derivatives, which positively contributed to our performance. Figure 2 on the bottom right shows our specified pool prepayment speeds by coupon, which in aggregate increased from 7.4% to 8.6% CPR. Most of the coupons in the stack experienced a mild absolute increase in speeds, resulting from a pickup in turnover rates from seasonal factors. Higher coupons, particularly TBAs, displayed larger speed increases due to lower mortgage rates in March into early April. We observed fast processing speeds for agency collateral, reducing the lag time from application to closing.,6.5% TBAs in particular, showed a large pickup in speeds, similar to the prepayment response we observed last fall when they briefly went in the money. Please turn to Slide 13. The MSR market remains very well supported with bank and nonbank servicers aggressively bidding for a declining amount of supply. As you can see in Figure 1, the volume of MSR available in the bulk market has continued to trend lower from the peak years of 2022 and 2023, with supply about 30% lower year-over-year. That said, we have still been able to find pockets of opportunity in the bulk market. Figure 2 is a chart that we periodically update for our earnings deck, which shows that with mortgage rates at their current level of around 6.75%, only 0.7% of our MSR portfolio is considered in the money. If rates were to drop to around 5%, the portion of our portfolio in the money would rise to only about 8%. Importantly, prepays have remained below our projections for the majority of our portfolio. Please turn to Slide 14, where we will discuss our MSR portfolio. Figure 1 is an overview of our portfolio at quarter end, further details of which can be found on Appendix Slide 25. In the second quarter, we purchased $6.4 billion UPB of MSR through 3 bulk purchases. The price multiple of MSR was unchanged quarter-over-quarter at 5.9x and 60-plus day delinquencies remained low at under 1%. Figure 2 compares CPRs across implied security coupons in our portfolio of MSR versus TBAs. Quarter-over-quarter, our MSR experienced a 1.6 percentage point pickup in prepayment rates to 5.8%. The increase was anticipated owing to stronger seasonal factors, though the speed was slower than model expectations. Overall, prepayment rates on our low coupon MSR are expected to remain very slow on a historical basis, which will remain a tailwind for our portfolio. Finally, please turn to Slide 15, our return potential and outlook slide. This is a forward-looking projection of our expected portfolio returns, which contemplates the effect of the loss contingency accrual on our portfolio. As you can see on this slide, the top half of this table is meant to show what returns we believe are available on the assets in our portfolio. We estimate that about 72% of our capital would be allocated to servicing with a static return projection of 11% to 14%. The remaining capital would be allocated to securities with a static return estimate of 12% to 17%. With our portfolio allocation shown in the top half of the table and after expenses, the static return estimate for our portfolio would be between 8.8% to 12.1% before applying any capital structure leverage to the portfolio. After giving effect to our unsecured notes and preferred stock, we believe that the potential static return on common equity falls in the range of 9.4% to 15.3% or a prospective quarterly static return per share of $0.28 to $0.46. Looking ahead, ongoing tariff threats and trade negotiations as well as geopolitical tensions will continue to weigh on the market. As always, we are mindful of the many sources of volatility that can impact our portfolio. However, we believe there is also opportunity in this environment. The resilience that markets demonstrated in the second quarter is a reminder of the global demand for investment, be it in equities or fixed income spread products like mortgage-backed securities. Spreads for Agency RMBS, particularly when hedged with interest rate swaps, remain historically wide and offer good relative value to other high-quality spread assets like corporate bonds. Supply and demand is balanced with demand diversified between money managers, banks, REITs and overseas buyers. Demand from depository institutions should increase if regulatory reform proceeds as anticipated. Our core strategy of low coupon MSR paired with Agency RMBS is well positioned to benefit from both stable prepayments and wide agency spreads. Additionally, RoundPoint's direct-to-consumer franchise enhances MSR returns through efficient recapture should we enter a faster prepay environment. Taken together, we are confident that our portfolio construction should drive attractive risk-adjusted returns across a range of market conditions. Thank you very much for joining us today, and now we will be happy to take any questions you might have.

[Operator Instructions]. We'll go first to Doug Harter with UBS.

So your leverage increased this quarter, I guess, as a result of the litigation reserve. Can you just talk about -- is that kind of the new level of leverage that we should be thinking about? Or are there still more portfolio actions to come to kind of bring it back to the prior leverage range?

Doug, this is Nick. Thank you for the question. So we ended the quarter at a leverage of 7x. And the range that we have discussed in these calls in the past has really been a range that we quote a range of about 5 to 8 as a leverage target. If you look at our leverage historically, even in that chart that's on that page, you will see that we've been at 7x in the not-too-distant past, right? So it really is -- it's very much within the range of the leverage that we operate in. The -- by quarter end and through the quarter, we really did like the risk in the market. We believe the mortgages versus swaps look quite attractive. And so we felt it was right from a portfolio management standpoint to increase our leverage through the quarter. And that 7x, as you say, is inclusive of the $200 million loss reserve that affects our book value. If you were to add that back into our capital, our leverage would drop to about 6.3%. But in any case, the leverage, we feel very comfortable where the leverage is at 7x and we'll govern portfolio as we always do. It will depend upon where spreads are in the market, what the opportunities -- where the opportunities are, and we will move the leverage around accordingly based upon market opportunities and our capital base.

Great. Appreciate that. And is there any way you could give us an update on economic return performance so far in July?

Doug, Quarter-to-date, through last Friday, we were up about 1.5%.

That's on economic return, just to be clear, right?

Yes, economic return on the new book value.

We'll go next to Bose George with KBW.

Can you remind us -- on Slide 15, you guys have the breakout of the expected returns. Actually, what are the main drivers, the differences between that and the EAD metric?

So the EAD is based on historical purchase yields of the assets, right? And as a result, I always like to say that it is -- that the EAD calculations are asynchronous among assets in the portfolio, right? Because it depends on the yield on which it was on the day that it was purchased, whereas the return outlook slide is all meant to reflect the forward-looking mark-to-market yields at current prices, right? We show a range of returns on Slide 15 to reflect things like potential fluctuations in prepayment speeds, potential fluctuations in funding spreads as well as potential fluctuations in leverage of the securities, right? But the main difference is the timing and the market prices at which the assets are utilized in the calculation of the yields.

Okay. That makes sense. And just given where the EAD has been trending sort of in the back half of the year, does it make sense that it probably continues to trend sort of below the economic return?

Bose, yes, I think that's the case because spreads have widened recently. And so the EAD of old securities that were purchased previously would not reflect that change in market value.

Okay. Yes, that makes sense. And then just one follow-up on Doug's question on the leverage. Does your view on leverage change once that capital that's reserved sort of goes out the door? I mean, is there -- because right now, as you noted, your sort of optical leverage is different from kind of your real leverage because you still have that cash. So just curious how that -- whether there's any change once some or part of that is paid out?

No. I mean, Bose, this is Nick. As a general rule, no, it doesn't really change our view on leverage. It just -- we will manage the portfolio according to the amount of capital that we have. And -- but as far as a general view about leverage, no, it doesn't. It doesn't change the way we look at the market.

Or the way that we manage the portfolio.

Right. It will -- it could potentially change. It's another factor in how we manage the composition of our assets. But as an overall rule, no, I do not think it will govern how we view leverage in total.

We'll go next to Trevor Cranston with Citizens JMP.

Bill, in your prepared comments, you mentioned the small amount of originations you guys have done in second liens. And I think you said that's a product you guys could hold or sell in the future. Obviously, it's a small number right now, but I was curious if you could comment on -- is that a product you guys are actually interested in building into the investment portfolio? Or in the near term, is that something that's more likely to just be sold off to third parties?

Yes. Thanks for the question. I think it's a question about risk and reward, right? So if the yields available on the securities are attractive, we will hold them as we create them. And if we feel that we can extract more value by selling them either in bulk or on flow or in a securitized form, then we'll do that. Really, it's just another set of tools in our tool belt to be able to use. And if it's an attractive asset class, we will take advantage of that. But to be able to have multiple outlets for the product is important to us. And so we'll look at all the opportunities on a real-time basis.

Got it. Okay. That makes sense. And then you guys also made a comment about some increased exposure to mortgage derivatives contributing to performance this quarter. Can you just elaborate a little bit on kind of where you guys have been active in the mortgage derivative space?

Sure. Trevor, this is Nick. Thank you for that question. At the beginning of the year, we did add another team member to focus on derivatives, which to date has mostly taken the form of growing our inverse IO exposure on the book. But over the quarter, I think we may have allocated about $50 million invested in that sector, but it's still under 5% of the securities capital. So it's still a small component of the book. It's just another sector of the market given the level of expertise we have in prepayments and managing risk across the mortgage -- agency mortgage sector, it's a sector which we felt have opportunity. And we have a very skilled group of people here to manage that part of the book. So it's something that we do believe we're going to continue to have a lot of focus on in the upcoming quarters. But as a total amount of risk, it's still a fairly small amount.

We'll go next to Harsh Hemnani with Green Street.

Can we talk about financing strategy and maybe the thought process behind moving part of the financing from repo to unsecured this quarter?

Harsh, it's William Dellal. The reason we did the unsecured baby bond was to start to prefinance the maturity of the convertible. And some of the warehouse lines that we used to use are now kind of warehouse repo. So that's why there is some change there. But basically, the big change is the issuance of the baby bond, which is to prefund part of the convert maturity.

We'll go next to Kenneth Lee with RBC Capital Markets.

Just wondering if you could just talk about thoughts around potential impact of a steepening yield curve on the portfolio and then in particular, just further expand upon the potential benefits there to the MSRs?

Ken, this is Nick. Thank you for the question. In general, as we have said, we do hedge across the yield curve in general. So we don't have a particularly strong view about the curve or our risks don't really reflect a particularly strong view across the curve. And if you look at the appendix slides that show our general curve exposure, you can see they're fairly small. So as a broad measure to our performance, we are hedged across the curve. That being said, steeper yield curves are usually good things for mortgage spreads because it does incent depository institutions frequently to go out on the curve and invest and particularly in one of their prime assets are mortgage-backed securities. And the history will tell you that when the Fed cuts and the yield curve steepens that you tend to see more participation by banks in our sector, and that drives spreads tighter. So there's a secondary effect of that. The actual -- the implications on MSR are again, built into our risks, MSR is an asset that is like an IO and that it's pre-payment sensitive. I think everybody knows that, but also a large component of the value of MSR is also the float income, right? Which is tied to the front end of the curve. So to the extent that the curve steepens, if front-end rates go down, that will take -- that will drive the price of MSR down. And in fact, that's one of the reasons why MSR these days, our low gross WACC still has as much -- it doesn't have much negative duration, but it does have negative duration. And a big component of that is the rates and the effect on float income. So the effect is that if you do get into a steeper yield curve environment on MSR, there really -- there are 2 counterbalancing effects. The first is that the float income will go down. The second is if the curve steepens, that does -- that also forward rates for longer forward rates typically go higher, which means that prepayment assumptions go lower. So there's a little bit of a balancing, but those 2 things will govern how the price action of MSR performs over time.

Got you. Very helpful there. And just on my follow-up here, just wondering if you could talk a little bit more about your risk appetite. I think you said that in the quarter, you like the risks in the markets. And looking at the risk exposure, it looks as if there's some quarter-over-quarter slight increase in rate exposure, maybe unchanged on spread. Just want to get a little bit more color around your appetite for risk in the current market?

Our exposure quarter-over-quarter is virtually unchanged. Our spread risk from quarter end to quarter end is pretty close if you measured by our spread risk on the appendix slide and on the main portion. Overall, I would say risk right now, we like the market. We like where mortgage spreads are. If you look at mortgages hedged with swap spreads, they are generous historically. And even on an OAS basis, they look pretty cheap relative to the amount of spread volatility that we've seen. As I said in my prepared remarks, spread volatility has declined to the lowest levels since the pre-COVID time period. We had a bit of a spike at -- in the early part of April when the market was quite volatile and the VIX hit that bit of a multiyear high. But the market settled right back down to where it was before that. And it's in this -- we're in a good place. I mean the realized rate volatility right now has been fairly low. Even once we got through the beginning part of the second quarter and for the third quarter so far, we've been in a pretty tight range of rates. We like where spreads are. And our MSR continues to perform quite well, and the market is extremely well supported right now in terms of demand.

We'll go next to Jason Weaver with JonesTrading.

In your prepared remarks, you mentioned something about maybe your outlook for the mortgage origination market and how that might affect the sort of MSR appetite, the opportunity set there and the competitive landscape going forward?

Yes. Well, you know -- I mean, we've been saying for a little bit now that our mortgage origination effort is still small. And as we said in the prepared remarks, only 0.7% of our portfolio is eligible for refinance from a rate and term perspective. And so we're trying to be mindful of the cost attached to growing that effort too large, too fast because that would create a drag on our earnings. But we're trying to do things within the business in order to be able to scale more quickly, originating second liens is part of that, and we can utilize having more loan officers making loans and making second liens and brokering second liens while there's not a lot of first lien activity, right? So we can share some of that. And then should rates fall later, we can transfer some of those things back to first liens right some seconds where the loan sizes are bigger and there will be more opportunity should rates fall and we get into an environment where more loans are eligible for refinance. So we're trying to balance those factors of costs and opportunity. And I think we'll watch the market going forward, and we'll be ready to react as rates move.

Got it. And I was wondering, can you provide maybe some more clarity or some bracketing around what you expect on time line for resolution of the PRCM litigation as well as if you have a ballpark for what the claims are for you against IP?

Well, look, unfortunately, I appreciate the question. I really do. Unfortunately, I can't really say a lot more than what William said in his prepared remarks. A trial date has not been set yet. And so we are waiting for the next stages to occur. When there is something to update you and the rest of the market on, we will, of course, do that as soon as we can.

We will go next to Rick Shane with JPMorgan.

I'd like to talk a little bit about the expense structure. Just servicing costs down sequentially, down year-over-year, comp and OpEx down sequentially. I assume that there's -- in the first quarter on the comp side, some sort of annual stuff that comes through, and that's what drives it. But particularly as you are expanding your investment in AI, can you help us understand what will be expense, what will be capitalized and how you see that driving your expense lines going forward?

This is William, Rick. A lot of what we're doing is going to be expensed rather than capitalized. The rules for capitalizing are quite strict. The material we're doing now is likely to be expensed, which is why our expense ratio is tended to stay constant or even creep up a little bit.

So when we think about expenses for the second half, the combined expense between servicing costs and comp benefits and OpEx was $45 million in the second quarter. Is that a decent run rate to build off of into the third, fourth and into 2026?

I think we're going to be in that ballpark, yes.

Okay. And in terms of the AI build-out, it's an interesting, I think, challenge for companies of your size balancing off the shelf versus bespoke internally developed solutions. Help us understand how that works in your business, please?

Look, I think it works the same way it works really almost everywhere. There's going to be -- there's lots of activity and lots of developments and lots of products being brought to market in real time and changing rapidly in this space that can help us do all kinds of efficiencies and cost savings and improve borrower experience. So it's changing very, very rapidly. I'd say it's more likely that we will be accessing some of those resources from other companies rather than building ourselves. But some simple things that we could do ourselves that are particularly bespoke for our needs, we are likely to do. But I'd say the bulk of it is probably going to be from third parties.

And third parties stuff that we take, some of it requires a little bit of customization to our particular needs, but the main bulk of it will be third party.

We'll go next to Jason Stewart with Janney.

A follow-up on Trevor's question on second liens and the balance sheet. As production ramps on the core business, is there any possibility or intent to retain POS? Or is the goal to continue to sell that?

I think there is a possibility for that. And I think that's one of the things that we're looking at as we grow the origination effort and the products that we offer and the outlets that we will access to do those things, whether it's selling it and some of it could involve retaining some of that stuff. So yes, I think that's one of the things that we're looking at.

Okay. And I assume that you look at the same way as you described second, meaning it's got to hit certain economic return hurdles, et cetera. Is there any other strategic thought behind shifting the balance sheet from primarily agency to private label credit?

Well, I think it would still be primarily agency. I don't think we would talk about expanding our non-agency exposure to a level that it would be primarily that. So as a minority interest in the portfolio of other kinds of asset class with predominantly similar risks, but some slightly different ones, I think it has a benefit in the portfolio. And so that's the way we're going to look at it.

At this time, there are no further questions. I'll turn the call back to Bill for any additional or closing remarks.

I want to thank everyone for joining us today. And as always, thanks for your interest in Two Harbors.