ONE — One United Properties S.A.

Deep-value analysis — ONE (One United Properties S.A., BVB: ONE)

Date: September 18, 2026 · Reference price: RON … · Market cap: RON 3,285.1 mil · 105,123,422 shares

REFRESH REGIME. This analysis resumes the report from 2026-08-27 (rapoarte/deep/2026-08-27-deep-ONE.md), triggered by the mechanical triage deep_delta.py on the grounds “1 falsifier triggered: dso”. The conclusion of the first chapter below is that the falsifier didn’t break because the world changed, but because it measured something other than the claim it was supposed to test — and when measured correctly, the signal is stronger than three weeks ago, not weaker.

Inherited primary sources: Audited 2025 Annual Report (Deloitte Audit SRL, opinion 2026-03-27), H1 2026 financial report (authorized 2026-08-25), BVB current reports. New sources for this resumption: the 2026-09-11 Capital Markets Day release, the 2026-08-27 release on abandoning the One Tower sale, market data as of 2026-09-18, data-pack-ONE-20260918.md, mc-ONE-20260918.json, trian-ONE.md.

Executive summary

What I carry over unchanged. The core thesis from 2026-08-27 remains intact and I’m not re-deriving it: One United Properties is Romania’s best premium developer, with a 538.9 thousand sqm landbank and a GDV of EUR 2.7 bn that can’t be replicated quickly, but the profit it reports doesn’t turn into cash. In 2022–2025 the group reported cumulative net profit of RON 1,751.8 mil and generated cumulative operating cash flow of minus RON 104.9 mil. The RON 1,856.8 mil gap sits on the balance sheet. Unchanged: the business and moat analysis, the management and capital-allocation profile, the Thorndike profile (3.0/5), the structure of the accounting red flags, and the three pre-mortem scenarios. Also unchanged is the FCF bridge: normalized owner earnings of RON 240 mil, ex-revaluation gains, the 2023–2025 average raised 3.5% for the incremental rental NOI.

What I re-derived, and why. Four things have moved since 2026-08-27, and only one of them is a price move:

1. The dso falsifier was withdrawn as unverifiable. The metric in teza_deep.py reads the yfinance label Accounts Receivable — which for ONE contains only the invoiced trade receivables (RON 90.8 mil at 2026-06-30). The RON 1,190.3 mil of IFRS 15 contract assets, i.e. exactly the subject of the claim, sit in a separate label, Other Receivables, which the metric doesn’t see. The falsifier reported 52.8 days; the real figure, on the full receivables pool, is 745 days. The claim wasn’t disproven — it was never measured.
2. Measured correctly, the deterioration continued. Total receivables rose from RON 1,057.6 mil (2025-06-30) to RON 1,281.1 mil (2026-06-30), …, while TTM revenue fell to RON 863.4 mil. More important: the pool’s composition worsened — the invoiced (legally due) portion fell 43.7%, from 161.2 to RON 90.8 mil, and the unbilled portion, conditional on cadastral registration, grew 32.8%, from 896.4 to RON 1,190.3 mil. The share of contract assets in the total rose from 84.8% to 92.9%.
3. The One Tower sale failed (2026-08-27). The company abandoned the disposal of its 71.46% stake in One United Tower SRL, having failed to identify a buyer at the EUR 114 mil minimum price approved by the shareholders’ meeting. The only visible non-debt cash source on the short-term horizon thus disappears.
4. Capital Markets Day, 2026-09-11 — the “ONE 2030” strategy. Management announced spinning off the rental assets into a separate vehicle, One Income Properties, targeting over 300 thousand sqm and over EUR 1 bn of value, a dividend yield of 7…, and NAV growth of 3… per year. This is the most important new element of this resumption and management’s first structural response to the discount versus NAV.

Two technical errors corrected versus the delta package. The Monte Carlo simulation in mc-ONE-20260917.json, on which the triage was built, is unusable: it was run with --oe … (the absolute value in RON, in a parameter expressed in millions) and with 110.5 mil shares instead of 105.123 mil. The result — a margin-of-safety median of … — is arithmetically impossible and should have been auto-rejected. I re-ran it correctly: mc-ONE-20260918.json. The correct share count is independently confirmed by the market cap: … / … = 105,123,422.

Valuation. Five triangulated models give a range of RON … – 48.68/share, with a median at RON …, i.e. … versus the RON … price. Monte Carlo over 20,000 scenarios, with my assumptions: median …, undervaluation probability …%, probability of a margin of safety above 30% only …%. The improvement versus 2026-08-27 (… median, …% P(undervalued)) is entirely the effect of the 9.4% price drop, from … to RON …. Median intrinsic value is unchanged: RON … then, RON … now. The business hasn’t become more valuable; the stock has become cheaper.

Verdict: WATCH, DON’T BUY at RON …. The same verdict as in August, with an entry threshold lowered from 26 to … — a change of opinion I declare as such and justify in the verdict chapter: at … the margin of safety versus my own median would be only 7.5%, which is inconsistent with an earnings quality rated WEAK. At … the margin is 22% and, simultaneously, P/B falls to 0.76x — below the 0.85x threshold I require of an asset whose normalized return (7.3%) is below the cost of capital (…%).