TPG RE Finance Trust (TRTX) is the only CRE mREIT with office exposure below 5%. After the post-quarter $227.1M NYC office loan paid at par, effective office is under 2%. The Q1 2026 10-Q (filed April 28, 2026) shows a 100% performing $4.3B portfolio, a CECL benefit, and a fourth consecutive quarter of unchanged risk ratings. The stock trades at 0.754x book ($11.06) at an 11.5% dividend yield. Q1 ROE of 9.0% on a ≈10% cost of equity implies fair P/B near 0.90x ≈ $9.95.

What the filing says

Risk rating distribution unchanged four consecutive quarters: 92% risk-3, 4% risk-4, zero risk-5. CECL recognized a $0.3M benefit in Q1 (vs a $3.4M provision in Q1 2025). One non-accrual loan ($42M multifamily, 1% of book) was modified to extend maturity to May 31, 2026.

Distributable Earnings of $0.25/share covers the $0.24 common dividend by a penny. Book value $11.06 — flat over trailing twelve months while paying $0.96/year in dividends. Buyback deployed 35.6% of the $25M authorization at $8.07 (73% of book). 77.9% of borrowings are non-mark-to-market CLO debt; FL7 closed Q1 at tight spreads.

Series C Preferred (TRTX-PC, $201.25M outstanding, 6.25% coupon) is first callable June 14, 2026. Redeeming would reduce equity from $1.062B to ≈$861M, breaching the $1.0B Tangible Net Worth covenant. The call won't happen — preferred holders earn the 6.25% coupon indefinitely with 6.2x distributable-earnings coverage and no near-term capital event.

What the market thinks

CRE mREIT P/B (April 29, 2026 spot):

The cohort bifurcation is already in trailing returns and P/B spreads — clean books at 0.70–0.75x, migration books at 0.39–0.52x. A naive long-clean / short-migration pair captures roughly zero forward EV because mean-reversion of the depressed shorts cancels recovery in the longs.

Standalone TRTX: 16% discount to a fair P/B of 0.90x plus 11.5% yield = 25–35% total return over 12–18 months if convergence happens, 18–25% IRR depending on speed. Idiosyncratic Sharpe ≈0.30 (modest); levered to a market beta of 1.32 the total Sharpe lands near 0.80.

Why the gap exists

• The 11.5% headline yield is high because DE covers the dividend by one penny — the cushion is thin.
• $42M multifamily non-accrual is an unresolved overhang (May 31 maturity).
• TNW covenant has only $62M headroom (5.8% above floor).
• External management (TPG, 1.5% fee, ≈$64M termination cost) is a structural cohort discount.
• 70.6% of variance is idiosyncratic — just under the 75% target. The other 30% is SPY/XLRE beta with no edge.

What is not priced: sustained P/B premium for office leadership over the next 12–18 months. Peers are 12–24 months behind on office cleanup. Another KREF or SEVN provision quarter widens the spread.

Risks (ranked)

1. TNW covenant. $62M headroom. A $42M charge-off plus one risk-4 migration tests it. ≈10%.
2. $42M multifamily charge-off ≥$15M by Q3 2026. ≈35%.
3. Peer convergence. KREF or BRSP accelerates office disposals; window 4–8 quarters.
4. Sector beta drawdown. β_SPY = 1.32; cohort moves with the credit cycle.
5. Dividend coverage erodes. $0.01 cushion over $0.96 annual dividend.

Catalysts

What would change our mind

Bullish update: $42M multifamily resolves at par by May 31. KREF or SEVN takes another large provision Q2. TRTX pulls back to the $7.50 area (margin of safety improves). New originations stay multifamily/industrial.

Bearish update: $42M multifamily charge-off >$25M. TNW headroom drops below $40M. Office or life-science loan migrates to risk-5. BoA facility extension fails to execute June 6. Sponsor TPG begins divesting the management contract.

For the common: 16% multiple convergence + 11.5% yield against 30% sector beta and a thin TNW covenant cushion. RSI at 75 signals extended conditions; the risk/reward improves materially on a pullback.

For the preferred (TRTX-PC): a 6.25% coupon with 6x earnings coverage from a clean-book operator, structurally protected from being called by the issuer's own covenant math. Different risk/return shape than the common — a credit instrument with mREIT exposure, not a multiple-expansion bet.

Evidence