A 16% Dividend, a Falling Stock, and the Machine Behind It: Redwood Trust's $8 Billion-Per-Quarter Securitization Engine

A new mortgage securitization, Sequoia Mortgage Trust 2026-11, appeared on the market pipeline this week. The headline is a standard entry — one more deal from one more private-label mortgage trust — and for most investors it would pass unnoticed.

But look at the parent company, Redwood Trust (NYSE: RWT), and you find something worth pausing for. The stock trades at $4.48, down nearly 19% year-to-date, yielding a dividend that prints at over 16%. The business behind the curtain — a machine that securitizes prime jumbo mortgages at a pace of more than $8 billion per quarter — is running harder than ever.

The question for an investor is not about this deal. These are institutional securities you can't buy on a retail brokerage account. The question is whether the pattern this deal represents reveals something useful about the company that prints them, and whether a dividend that looks like a steal is actually a warning signal.

How a securitization works — and why Redwood needs them to keep flowing

When a bank originates a jumbo mortgage — a loan too large for Fannie Mae or Freddie Mac to buy — it has two choices. It can hold the loan on its balance sheet and absorb interest-rate risk, or it can sell the loan to someone else.

Redwood Trust is that someone else. Through its Sequoia platform — operating since 1997 — Redwood buys prime jumbo loans from banks and originators, bundles hundreds of them into a trust like SEMT 2026-11, and sells the resulting mortgage-backed securities to institutional investors. Redwood earns the spread between what it pays for the loans and what it receives from selling the securities.

The latest Sequoia deal appears on CreditFlow's pipeline with Redwood as sponsor and Wells Fargo Securities and another arranger helping to structure and distribute it. It follows a well-established pattern: the 2026-9 deal just before it was $737 million of fixed-rate jumbo mortgages with average borrower FICO scores near 778 and loan-to-value ratios around 73% The 2026-8 deal was $552 million with similar quality These are not subprime loans. These are first-lien mortgages to high-credit borrowers, typically on 30-year terms.

Redwood priced 20 securitizations through the first half of 2026 — including nine through Sequoia alone It locked $5.6 billion in loans on the Sequoia platform in the second quarter, and its total mortgage banking production exceeded $8 billion for the second consecutive quarter

The gain-on-sale margin — the profit Redwood makes on each transaction — was 92 basis points on Sequoia loans in Q2 That sounds small, but on $8 billion of volume it translates to real cash. The mortgage banking platforms generated $40 million in GAAP net income in the quarter, delivering a 33% annualized return on capital

This is not a speculative operation. Redwood has executed 166 securitizations over nearly three decades, with approximately $81 billion in cumulative issuance It has relationships with over 70% of the top 50 U.S. banks The machine works, and it's working at scale.

The problem the 16% yield hides

Now here is where the story splits.

Redwood Trust carries two businesses in one stock. The first — Sequoia, plus its home-equity platform Aspire and its construction-lending platform CoreVest — is the machine described above. It's profitable, growing, and generating real cash.

The second is the legacy investment portfolio — older mortgage and real-estate holdings acquired in earlier years, held on the balance sheet, and marked to market. These investments are losing value. In Q2 2026, legacy investments produced a $14 million loss on Redwood's adjusted earnings measure In Q1, it was a $13.1 million loss on a GAAP basis The losses come from interest-rate sensitivity and mark-to-market declines on fixed-income positions purchased when rates were lower.

Because Redwood is a public company, both businesses hit the same bottom line. Consolidated GAAP net income for Q2 was a loss of $3 million Revenue of $56 million missed analyst expectations of $94 million by 40% Book value per share has fallen from $7.36 at the end of 2025 to $6.90 at the end of June 2026

The market reacted. The stock has fallen from a 52-week high of $6.97 to its current $4.48.

And that's how you get to a 16% dividend yield. Redwood pays $0.18 per share quarterly At $4.48, that prints as 16.1%. But the dividend exceeds the adjusted earnings per share of $0.15 for the quarter Redwood is paying out more than it is earning, and the shortfall is being funded from the book value of the company.

That is not a sustainable dividend. It's book-value liquidation dressed as income.

Is management fixing the problem?

Redwood's management has acknowledged the drag. They're actively reducing exposure to the legacy portfolio — cutting its allocation from 15% of total capital to 12% in Q2, with a stated target of below 5% by year-end They estimate that shifting $100 million from legacy investments to the operating platforms would improve the return on equity by 200 to 400 basis points

There are also the joint ventures. Redwood announced a partnership with Castlelake in April to purchase up to $8 billion of prime jumbo loans, and has total joint-venture purchasing power approaching $18 billion These partnerships spread risk and reduce the capital Redwood needs to deploy from its own balance sheet.

The operating platforms are also getting more efficient. General and administrative expenses fell 23% quarter-over-quarter in Q2. Redwood is using AI-driven automation that has already saved over 23,000 annualized hours, and direct expenses dropped to 64 basis points of volume in the first half of 2026, down from 88 basis points in 2025.

The sequence makes sense: shrink the drag, scale the profitable engine, control costs, and redirect capital to where it earns 33% instead of losing value. If it plays out, the consolidated earnings problem resolves and the dividend becomes supported.

But "if it plays out" is the entire investment here. The legacy portfolio hasn't disappeared yet, and mark-to-market losses continue to erase book value faster than the operating engine can replace it. The dividend remains unsupported.

What this means for the investor

Redwood Trust is not a dividend stock. At least, it is not one right now. The 16% yield does not reflect durable income — it reflects a stock that has been punished for legacy losses, with a dividend that has not yet been adjusted downward to match reality.

The Sequoia securitization machine is real, it's profitable, and it's one of the longest-tenured platforms in non-agency mortgage finance. The broader non-agency RMBS market is expanding — issuance has risen 22% year-over-year through mid-2026, and institutional investors are paying 50 to 150 basis points more for non-agency tranches than for equivalent corporate bonds The structural tailwinds exist.

But the structural tailwinds belong to the operating platforms. The stock investor holds the whole company — the profitable machine and the legacy drag together.

The setup makes sense only if you believe management will execute on the transition: legacy allocation below 5%, the operating platforms continuing to deliver 33% returns, and the dividend eventually stabilizing as a share of actual earnings rather than book value. That could take one or two more quarters of legacy write-downs and pain.

Or the dividend could be cut before the transition completes, which would likely push the stock lower and reset the yield to something less seductive and more honest.

Neither outcome is catastrophic for a patient investor who understands the difference between yield from earnings and yield from erosion. But confusing the two — buying a 16% yield because it looks like income and not realizing it looks that way precisely because the underlying value is declining — is exactly the mistake that turns a good business into a bad investment.

The securitization machine behind SEMT 2026-11 is worth watching. The stock may eventually be worth owning. Right now, the math doesn't support the yield, and that matters more than any headline.