Barkat Capital: What the extensions and overdues really say about the book

This time the question is the asset side, not the funding side

The main article centered on funding. This follow-up isolates a different question inside the same 2025 package: what do the extension and aging tables actually say about credit quality. By year-end 2025, Barkat had loans of NIS 321.1 million whose contractual maturity had arrived and was extended by agreement, plus another NIS 63.9 million that were already 90 days to one year past due. Together that is NIS 385.0 million, or about 37.3% of the loan book.

This is where the collateral argument has to be tested for real. On one hand, management presents a book that is fully collateralized, about 95% backed by first-ranking mortgages, with an average LTV of roughly 57%. On the other hand, the same evidence set shows that part of the stress has already moved beyond theory and into enforcement: there is a NIS 63.8 million loan that missed maturity, is already in legal and execution proceedings, and still carries no specific provision.

That is the core read. The collateral is not cosmetic. It clearly explains why provisions remain low. But at year-end 2025 it looks more like a mechanism that buys time and supports expected recoveries than proof that the pressure is already closed in cash.

Where the pressure actually sits

The aging table is unusually revealing. This is not a book with a few routine slips around quarter-end. It shows two explicit stress pockets: loans that had to be extended, and loans that are already deep in delinquency.

What matters is the shape of the delinquency. There is nothing in the up-to-90-day bucket, and nothing in the one-to-two-year bucket. The entire overdue balance sits directly in the 90-days-to-one-year range. In other words, when delinquency appears here, it is not showing up as a fresh technical delay. It is already in the zone where the real question becomes repayment, another extension, or enforcement.

At the same time, nearly one-third of the book has already passed contractual maturity and been formally extended. That is not automatically the same thing as default. The company itself explains that in construction-lending projects, timetable delays can lead to maturity extensions, usually for a few months and up to a year, sometimes with an added fee or higher interest. But at this scale, extensions stop being an operational detail and become a quality metric. The question is no longer whether extensions are allowed in the model. The question is how much of the book now needs them just to stay alive.

What the provision line says, and what it does not

The first number that supports the containment narrative is the total allowance balance. It rose to NIS 2.364 million at year-end 2025 from NIS 1.636 million at year-end 2024, an increase of about 44.5%. Stage 3 moved from NIS 36 thousand to NIS 923 thousand, so the company is recognizing some deterioration. But the recognition is still small relative to the size of the stressed buckets.

The most unusual data point is the mismatch between aging and provisioning. All of the specific allowance, NIS 871 thousand, sits inside the matured-and-extended bucket. The 90-days-to-one-year overdue bucket, worth NIS 63.9 million, carries no specific allowance at all, only NIS 36 thousand of general allowance.

The detailed note explains why. The largest loan inside that bucket is Loan C, with a carrying balance of NIS 63.8 million including accrued interest. It was not repaid on time. The company has already initiated collection proceedings, obtained a partial court ruling, and after the borrowers failed to pay by the required date, continued into execution proceedings. Even so, the company concluded that no specific provision was required because the present value of expected cash flows, based on the collateral in hand, still exceeds the carrying balance.

That distinction matters. At the accounting level, the collateral is already doing real work: it prevents a deep delinquency from flowing automatically into an immediate P&L hit. But the same fact pattern also says that collateral does not make the event quiet. It simply allows management to argue that the final loss is still not there.

Against that, the company did move one step further on two other loans. Loan A received a specific provision of about NIS 613 thousand, and Loan B a specific provision of NIS 258 thousand. In both cases the trigger was similar: a material engineering delay and a decline in project profitability. So Barkat is not ignoring every sign of stress. It is distinguishing between cases where multiple recovery scenarios and collateral values still leave expected cash flows below book, and a case where management still believes the collateral covers the debt.

One more point matters here. Loans A, B and C all come with relatively strong paper protections: first-ranking mortgages without a capped amount, plus cross-collateral arrangements between some of the loans. The reasonable conclusion is therefore not that the collateral is weak or fictional. The more reasonable conclusion is that the collateral gives the company a cushion, but that cushion does not eliminate the fact that a project can stall, collections can drag, and time itself can become part of the economic loss.

Where the average hides the tail

Slide 20 of the investor presentation gives a reassuring snapshot: 100% of the book is backed by collateral, about 95% by first-ranking mortgages, and the average LTV is roughly 57%. That read matters, but it is incomplete. A comfortable average does not tell you what is happening at the sharp end of the portfolio, which is exactly where extensions and provisions are born.

The top-10 borrowers table shows that tail clearly. The largest attributed exposure is NIS 60.5 million at 95% LTV. After that come exposures of NIS 31.9 million at 83% LTV, another NIS 29.6 million at 80% LTV, and NIS 19.1 million at 82% LTV. There are also more conservative large exposures, such as NIS 56.1 million at 58% LTV and NIS 15.0 million at 45% LTV. So the 57% portfolio average is real, but it is not a ceiling on risk. It simply smooths over pockets where the collateral buffer is already much thinner.

There is another layer of complexity here. Some rows in the table are measured by project absorption capacity rather than LTV, and some borrowers are presented on an equal pro-rata basis when they are jointly and severally liable on the same loan. The company also emphasizes that a meaningful share of projects is funded with non-recourse structures, so its economic exposure to certain borrowers is smaller than the gross attributed exposure in the table. All of that is true. But even after those qualifiers, the table still says something simple: in the visible upper part of the book there are meaningful risk pockets where collateral value is not far above debt.

That is exactly why extensions and overdues matter. When average LTV is 57%, it is easy to assume that collateral by itself resolves almost any problem. Once you look at the larger exposures in the tail, the real question becomes how much time the project can lose, or how much value it can shed, before that cushion starts to disappear.

What this says about Barkat's book

The conclusion is not that the extension and overdue tables expose a weak book hiding behind collateral. That would be too aggressive relative to what the company does show: strong collateral packages, mostly first-ranking liens, sometimes cross-collateralization, and an allowance balance that is still low in absolute terms. But the conclusion is also not that the collateral has already closed the issue.

At year-end 2025, the collateral looks like a mechanism that allows loss recognition to stay deferred as long as management believes realizable value is still there, not like proof that the stress has already been resolved. Those are not the same thing. When 37.3% of the book sits either in extension or in delinquency, and the main deep-overdue exposure is already in enforcement without a specific provision, the market should spend less time on the slogan that the book is fully collateralized, and more time on whether that collateral can actually be turned into cash within a reasonable time frame and without material value leakage.

Bottom line: for now the collateral contains the loss reading, but it does not yet clear the stress signal inside the book. For that reading to strengthen, three things need to happen: the NIS 63.8 million overdue exposure needs to be resolved without a meaningful hit, the extension bucket needs to start declining rather than merely rolling forward, and the large high-LTV exposures need to stay out of another round of specific provisioning.