Granite Point Mortgage Trust is an internally managed commercial mortgage REIT whose office-heavy book of floating-rate senior loans is shrinking faster than it is being replaced, and whose income now comes from carry on collateral it has already marked down.
The mechanism sits in the second quarter print, and the loss is the number that carries the story. GAAP net loss attributable to common stockholders reached $62 million. Book value per share fell from $7.05 at quarter end to $5.70. The allowance for credit losses grew to $165.8 million after a $47 million quarterly provision.
Management's own forecast concedes that unrestricted cash, now $58.5 million, can dip below the post-amendment covenant floor between the third and fourth quarters of 2026. The Miami Beach sale is the only committed source of the cushion, and the common and preferred dividends are the most likely casualty if that sale slips.
Can the JPMorgan refinance and the extended facility carry the balance sheet through the resolution of the five nonaccrual loans without another capital event?
Granite Point Mortgage Trust Inc. is an internally managed commercial real estate finance company, incorporated in Maryland and listed on the NYSE, that originates, invests in, and manages senior floating-rate commercial mortgage loans and other debt and debt-like investments. The company operates as a REIT, maintains the exclusion from registration under the Investment Company Act, and runs as a single reporting segment with a small internal team split between New York and Minnesota. Its stated objective is to preserve stockholders' capital while generating risk-adjusted returns primarily through dividends derived from current portfolio income.
The balance sheet is the story, and it is a balance sheet in runoff. The company held 38 senior loans at quarter end. The unpaid principal balance stood at $1.4 billion, and total commitments reached $1.5 billion. Office properties now represent 48.6% of the loan book, up from 43.6% at year-end, because non-office assets have paid down and been resolved faster than office exposure has. Two properties, the Miami Beach office building and one other office asset, sit in real estate owned with a combined carrying value of $90.7 million.
The funding stack is as central to the thesis as the assets. The company finances the book through two consolidated CRE CLOs, three bank repurchase facilities at Morgan Stanley, JPMorgan Chase, and Citibank, a secured credit facility, a mortgage loan payable, and a small slice of loan participations. In July the company refinanced the two legacy CLOs onto the JPMorgan facility, extending it to July 2028 with three one-year extension options and upsizing it to $651 million. The refinance lowered the cost on the $521 million of CLO debt by 38 basis points.
The strategic posture is one of managed runoff. First-half originations totaled $22.3 million, while principal repayments ran to $275.7 million. The weighted average risk rating has drifted to 3.2 from 2.9 a year earlier. Management is selling, resolving, and extending its way toward a smaller, cleaner book, and the question is whether the income produced by what remains can service the preferred dividend, the common dividend, and the covenant stack at the same time.
The product is a senior floating-rate commercial mortgage loan, typically underwritten to a low-to-mid 70s stabilized LTV over a three-year term with one or two borrower extension options. There is no technology moat in the conventional sense. The company's durable advantages are its Granite Point underwriting relationships, its ability to close loans that larger CMBS issuers no longer price, and a servicing operation run by a third-party servicer with an internal asset management team retaining all credit decisions.
The moat is narrow but real. Granite Point originated the $64.8 million California office loan at a high LTV. The coupon sits at S+5.50%, and the all-in yield at origination was S+5.65%. A structure like that reflects the depth of its relationship with the borrower and the flexibility of its balance sheet relative to the CMBS market. The company also embeds preferred equity, unsecured notes, and B-note style positions in the borrower entity of four loans, adding $12.7 million of non-controlling other investments and up to $7.7 million of further commitments, which increase the recovery pool on a downside and add residual upside if the asset stabilizes.
The real estate owned assets are the least attractive part of the product line. The Miami Beach property was classified held-for-sale in Q2, marked to $54.9 million, and carries an executed purchase and sale agreement that management cites as the centerpiece of its covenant mitigation plan. The second REO, an office asset carried at $35.8 million, is held-for-investment. Carrying both properties produces roughly $3 million of quarterly REO revenue against roughly $5 million of quarterly operating expense and depreciation, a structural drag that does not disappear until the sale closes.
The company's competitive position within commercial mortgage REITs is defined by its size and its funding stack rather than by asset quality. At $1.5 billion of commitments, it is smaller than the sector's largest operators, which gives it flexibility but also limits its cost of funds. The two CLOs, now refinanced, and the extended JPMorgan facility are the structural features that distinguish GPMT from its micro-cap peers. Those same features concentrate the funding relationship with a single lender in a way that the next amendment cycle exposes.
The second quarter print produced a GAAP net loss attributable to common stockholders of $62 million, or $1.29 per share. The loss is the number that carries the quarter. Interest income of $21.8 million covered only $17 million of interest expense. Net interest income came in at $4.8 million, down from $8 million a year earlier. The $47 million provision for credit losses, the $6.1 million Miami Beach impairment, and the operating expense line turned that narrow margin into a large loss.
Distributable loss, the non-GAAP measure the board uses to set the dividend, was $37.7 million, or $0.79 per share. That figure is the dividend-relevant number, and it excludes the unrealized provision. It does include $29.7 million of realized write-offs and $3.1 million of discount on participations sold. The quarter-on-quarter swing from a small distributable loss in Q1 to a $37.7 million loss in Q2 is driven almost entirely by the Chicago resolution and the associated write-offs, which confirm the scale of cash that can leave the book in a single quarter when a loan resolves below carrying value.
The portfolio is shrinking by roughly $120 million per quarter. In Q2, $129.8 million of unpaid principal balance left the book through repayments, resolutions, paydowns, and amortization, against $8 million of new fundings. The largest single item was the resolution of the $76 million Chicago retail and former office loan, a loan that had been impaired in the prior year. The full repayment of the Richmond office loan and the partial repayment on the Los Angeles office loan completed the runoff.
The allowance for credit losses is the load-bearing number. It reached $165.8 million at quarter end, a level equal to 11.4% of total loan commitments. Of that total, $130.2 million, or 78%, is allocated to specific loans. Five loans, with an aggregate principal of $252.9 million, are rated 5 and on nonaccrual, including two office properties, two hotels, and one multifamily. The weighted average exit capitalization rate on the collateral-dependent loans is 9.29%, and the weighted average discount rate is 11.72%, inputs that reflect the spread between the origination vintage and the current office market.
The next six months resolve three questions: whether the Miami Beach sale closes on schedule, whether the unrestricted cash covenant holds through the resolution of the remaining nonaccrual loans, and whether the common and preferred dividends survive the Q3 and Q4 board meetings. Management's revised rolling 12-month forecast, disclosed in the second quarter filing, concedes that unrestricted cash could fall below the post-amendment floor between the third and fourth quarters of 2026. That floor is the greater of $20 million and 5% of recourse indebtedness. The mitigation plan names two levers: closing the Miami Beach sale, which the filing estimates at roughly $55 million of fair value less costs to sell, and temporarily reducing or suspending both dividends to preserve cash.
The JPMorgan refinance is the most important single execution event of the period. The company moved $521 million of CLO debt onto the JPMorgan facility. It also extended the facility to July 2028, removing the near-term maturity wall and cutting the cost of that debt by 38 basis points. The trade-off is concentration: JPMorgan now carries the recourse repurchase borrowings plus the full CLO asset stack, giving it roughly 65% of total financing. The Morgan Stanley and Citibank facilities, both extended to late 2027, provide the remaining cushion, and the next amendment cycle is the first real stress test of that concentration.
The loan book's maturity profile forces a decision on every remaining asset. With 0.4 weighted average years to maturity, the company is left to extend, sell, or foreclose on each of the 38 loans within the next year. The four modifications disclosed in the trailing 12 months, covering the New York office loan, the Tempe hotel, a second New York office loan, and the Encino office loan, set the template: extend in exchange for an upsizing, a higher coupon, a higher residual return, or a preferred equity stake. The risk is that the modification menu, which has already been used on four of the largest positions, runs out before the portfolio does.
Dividend sustainability is the second execution risk. The board declared $0.05 per share in Q2, the same as Q1, and the same as both quarters of 2025. At a $1.45 average share price over the last two quarters, the dividend yield is roughly 14%. That sits well above the run-rate distributable earnings before realized gains and losses, which has been negative or near zero in each of the last four quarters. The preferred dividend of $0.4375 per share per quarter is fixed and cumulative, and a suspension of the common dividend is a capital allocation choice the board has not made yet.
The downside scenario starts with the Miami Beach sale. If the purchase and sale agreement slips, breaks, or closes at a price below the $54.9 million fair value, unrestricted cash drops toward or below the covenant floor without the offsetting inflow. The company's own disclosure concedes the temporary covenant breach between Q3 and Q4 of 2026, and the open question is whether the breach is a disclosure event or a default event. A default on the JPMorgan facility, now carrying the bulk of the book, would force a sale of collateral at exactly the moment office values are at their most liquid-discounted.
The second risk is the credit book itself, and it is the risk with the largest tail. The five nonaccrual loans total $252.9 million in aggregate. That is 18% of unpaid principal balance. The allowance covers only $165.8 million of total commitments. If the weighted average exit capitalization rate on the collateral-dependent loans drifts to 10.5%, or the discount rate to 13%, the implied fair value of the collateral drops by a meaningful margin. That move adds a further $20 to $25 million of unrealized provision before any realized write-off. The quarter already recorded $29.7 million of write-offs, which shows the scale of the realized-loss tail.
The third risk is funding concentration. JPMorgan's share of total financing rises to roughly two-thirds after the CLO refinance. The next amendment cycle in July 2028 is two years out, but the interim extension options, three one-year term extensions, are the only structural protection. If JPMorgan's appetite for the office-heavy book shrinks as the vintages mature, the company's cost of funds on the remaining recourse facilities rises, and the net interest income line, already at $4.8 million per quarter, compresses toward zero.
The fourth risk is the REO drag. The two office properties produce a net quarterly loss of roughly $1.8 million, and the Miami Beach asset is not generating meaningful new lease income. If the sale extends beyond Q4 2026, the carry cost, including depreciation, interest, and property taxes, continues to erode book value at roughly $2 million per quarter, and the tangible net worth covenant headroom shrinks with it. The counterweight is that the REO assets are already marked to the lower of cost or fair value less costs to sell, so the downside on the Miami Beach asset is largely booked, and the residual risk is the timing of the inflow rather than its size.
The valuation framework is book value per share, because the company's assets are marked to amortized cost net of CECL and the market is pricing the common equity as a residual claim on a shrinking, marked-down loan book. Book value per common share was $5.70 at quarter end. The market price over the last three months has ranged from $1.05 to $1.45. That is a discount of 75% to 85%. The market discount reflects three things: the realized write-off pipeline embedded in the allowance, the dividend the board may suspend, and the covenant headroom the company's own forecast concedes is thin.
The bear case is the scenario where nothing goes right. The Miami Beach sale closes at the fair value of $54.9 million. The five nonaccrual loans resolve at a low recovery, and the board suspends both dividends through 2027. The nonaccrual book, written off at a steep discount, produces a realized loss of roughly $88 million against the allowance already carried. The common equity, after the inflow and the write-off, settles near $240 million, or $5.00 per share. At a 70% discount for the dividend suspension, the bear-case value is roughly $1.50 per share.
The base case is the scenario where the plan works as management has described it. The Miami Beach sale closes at $54.9 million. The nonaccrual loans resolve at a mid-level recovery over the next four quarters. The board maintains the preferred dividend while halving the common dividend to $0.025 per share. The remaining book runs off at a 6% yield with no new originations. A realized write-off of $50 million leaves a residual allowance near $2.40 per share. At a 50% discount for the ongoing runoff, the base-case value is roughly $3.00 per share.
The bull case is the scenario where the office market stabilizes and the board regains optionality. The Miami Beach sale closes at $60 million. The nonaccrual loans resolve at a strong recovery, and the board reinstates the common dividend at $0.05 per share once the covenant breach clears. The company redeploys at least $150 million of the proceeds into new originations. The allowance, partially released as loans resolve above carrying value, supports a book value per share of roughly $6.20. At a 40% discount for the office concentration, the bull-case value is roughly $3.70 per share. The spread between bear and bull brackets the current market price at the bottom of the range, which is where the dividend yield is doing the work of supporting the equity.
The second quarter of 2026 is not a normal quarter for Granite Point Mortgage Trust; it is the quarter in which the company's own disclosure moved the story from credit quality to liquidity. The provision, the Miami Beach impairment, the portfolio runoff, and the post-quarter JPMorgan refinance all point to the same conclusion. The asset side of the balance sheet is shrinking, the funding side has been extended, and the income side is no longer covering the dividend stack. The common equity is real, at $274.5 million. But the book behind it is 48.6% office, and the average loan sits 0.4 years from maturity, with nearly a fifth on nonaccrual.
The counterargument to a more bearish read is the allowance. The reserve sits at $165.8 million, or 11.4% of commitments. Of that total, 78% is allocated to specific loans. That structure means the reserve is built to absorb the nonaccrual book resolving in the upper recovery range without a further mark on the remaining performing loans. The JPMorgan extension to 2028 removes the near-term funding cliff, and the 38-basis-point cost reduction on the CLO debt is a real, if small, improvement in the NII line. If the Miami Beach sale closes on schedule and the five nonaccrual loans resolve at the upper end of the range, the company exits 2026 with a smaller, cleaner book, a single extended facility, and a dividend it has chosen to keep paying.
The judgment is that the market price already incorporates most of the bad news. The 75% to 85% discount to book value is not a mispricing; it is the price of a shrinking, office-heavy, single-lender balance sheet with a dividend the board has not yet cut. The thesis variables that move the equity are the Miami Beach sale price, the recovery rate on the nonaccrual loans, the dividend decision at the Q3 board meeting, and the JPMorgan relationship through the amendment cycle. On the evidence in the second quarter filing, the most likely path is a base case. The sale closes near $55 million. The nonaccrual book resolves at a mid-level recovery, the common dividend is cut or suspended once, and the common equity settles in the $2.50 to $3.00 range over the next two quarters. That is a meaningful recovery from the current price, but it is not a return to book value, and the preferred stock, at a $25.00 liquidation preference against a tangible net worth covenant floor of $500 million, remains the structural constraint on any further common equity expansion. The risk is concentrated in the next two board meetings and the next closing statement; the capital is already marked, the funding is already extended, and what remains is the arithmetic of the runoff.