Sông Đà 4 (SD4): recovery narrative faces forensic and balance-sheet constraints despite >25% implied upside
Intrinsic value VND 2,403 vs market VND 1,900 — implied upside 26.5% (model confidence: low).
Business Overview
Công ty Cổ phần Sông Đà 4 operates in construction and building materials, listed on UPCOM with 10,300,000 shares outstanding. The company is majority-controlled by state-owned Tổng Công ty Sông Đà (65.0%), with a small free float and limited trading liquidity (average volume ~6 shares over 2 weeks), which amplifies execution and marketability risk.
Operating in Vietnam's construction segment, SD4 has seen contracting revenue decline sharply from VND 306.2 bn in 2023 to VND 117.0 bn in 2025 and persistent net losses narrowing from VND -135.5 bn (2023) to VND -5.4 bn (2025). Reported margins are mixed: a gross profit margin of 28.4% and EBIT margin of 12.3% contrast with a negative net profit margin (-4.6%), negative EPS (VND -522) and negative BVPS (VND -1,277). These outcomes sit alongside forensic red flags and a reported distress-zone Altman Z-Score per the forensic summary.
Investment Thesis
SD4's intrinsic valuation (EV/EBITDA mid-cycle) points to VND 2,403/share, implying 26.5% upside to the current UPCOM match price of VND 1,900. The valuation uses a calibrated fair EV/EBITDA of 8.25, below the sector median EV/EBITDA of 9.85, reflecting conservative cycle assumptions and the firm's recent weak operating scale. The company's trajectory toward near breakeven net profit in 2025 (VND -5.4 bn) supports a recovery scenario if revenue stabilises and EBITDA normalises toward mid-cycle levels.
Offsetting the recovery case are material forensic and balance-sheet concerns. Beneish M-Score dynamics and an Altman Z-Score in the distress zone point to elevated manipulation and bankruptcy risk; our model flags "manipulation_risk" and "negative_equity." Operating liquidity is constrained, trading is illiquid, and state ownership (65.0%) limits immediate foreign ownership demand despite available foreign room. Given the model's low confidence, the implied upside is not fully convincing for a high-conviction buy — it is large enough to merit selective accumulation for investors willing to accept governance and execution risk.
Key practical consideration: UPCOM-listed liquidity constraints and the large controlling shareholder mean exit slippage and block-trade risks are meaningful. Any investment should be sized with an allowance for potential further equity dilution, asset restructurings, or prolonged recovery timelines.
Valuation Commentary
Intrinsic value derived from an EV/EBITDA mid-cycle approach: mid-cycle EBITDA was scaled to a fair EV/EBITDA multiple and calibrated with isotonic recalibration to avoid outliers.
- Fair EV/EBITDA applied: 8.25 (own_history) versus sector EV/EBITDA of 9.85.
- Model uses a multi-year median EBITDA series (7 years) and a calibration step that reduced the raw intrinsic value from VND 4,232.2 to VND 2,403.
- Sanity flags lowered model confidence (illiquid market, upside capping, manipulation risk, negative equity).
- Recent improvement in net losses to VND -5.4 bn in 2025 supports a scenario where mid-cycle EBITDA recovers toward model inputs.
The VND 2,403 target implies 26.5% upside but model confidence is low. That upside exceeds our >25% threshold for a buy only when confidence is medium/high; because confidence is low and multiple forensic flags exist, our conviction is limited. The valuation should be treated as conditional on earnings normalization and resolution of balance-sheet and governance risks.
Bull vs Bear
- Intrinsic value VND 2,403 implies 26.5% upside to VND 1,900, offering cushion if execution proceeds.
- Net losses improved year-on-year to VND -5.4 bn in 2025 from VND -135.5 bn in 2023, suggesting operating recovery momentum.
- Gross margin of 28.4% and EBIT margin of 12.3% indicate project-level economics are adequate when scale returns.
- Forensic red flags: Beneish M-Score -0.0597 (93rd percentile among peers) and a year-over-year deterioration of +3.39 in Beneish metrics point to aggressive accounting risk.
- Altman Z-Score in the distress zone (forensic summary) and negative BVPS (VND -1,277) imply material balance-sheet distress and bankruptcy risk.
- Illiquid trading (avg volume 2w = 6) and large state ownership (65.0%) raise execution and marketability risk; model flagged "illiquid" and "illiquid_upside_capped."
Sector Context
The Vietnamese construction and building materials sector trades with a wide range of multiples; sector EV/EBITDA median is 9.85. Peer comparative upside median is modest (median upside ~9.6%), while a subset of peers shows >30% model upside but often with similarly low confidence. UPCOM-listed construction names typically suffer higher liquidity premia and governance scrutiny, and SOE-linked players (like SD4 with 65.0% state ownership) are subject to state-directed balance-sheet and payout dynamics.
Regulatory and accounting considerations in Vietnam matter: VAS differences (recognition of construction contracts, work-in-progress and land use rights treatment) can distort short-term profitability and working capital. Banks and contractors also face SBV credit controls and potential VAMC/backwarded receivable solutions, which can change recovery paths materially and quickly when state-backed interventions occur.
Risk Factors
- Forensic/manipulation risk: Beneish M-Score -0.0597 places SD4 in the 93rd percentile among peers; year-over-year change in the M-Score is +3.39—this is a high-risk signal for earnings quality.
- Bankruptcy/distress risk: Altman Z-Score reported in the forensic summary sits in the distress zone; negative BVPS (VND -1,277) compounds solvency concerns.
- Liquidity and marketability: UPCOM listing with average 2-week volume of 6 shares and limited free float limits ability to enter/exit positions at model prices.
- State control and related-party dynamics: 65.0% ownership by Tổng Công ty Sông Đà concentrates control and could influence cash distributions, asset transfers or restructuring timing.
- Model and data risk: valuation confidence is low and the model was isotonic-calibrated, reducing the raw intrinsic value from VND 4,232.2 to VND 2,403; upside is sensitive to EBITDA normalization assumptions.
- Execution risk on backlog/revenue: revenues fell from VND 306.2 bn (2023) to VND 117.0 bn (2025); a failure to restore scale would erode the EV/EBITDA recovery case.
Catalysts
- Evidence of sustained revenue stabilization or contract wins that reverse the 3-year revenue decline (VND 306.2 bn → VND 117.0 bn).
- Transparent remediation of forensic concerns (audit disclosures, improved disclosure around accounting policies) which would reduce perceived manipulation risk.
- State-led restructuring or capital injection from the majority shareholder that improves solvency and restores investor confidence.
- Improved trading liquidity or an uplisting/event drawing more institutional interest.
Forensic Assessment
Forensic flags are the dominant concern. Beneish M-Score is -0.0597 and placed SD4 in the 93rd percentile among Vietnamese peers; the M-Score trend worsened year-over-year by +3.39, indicating increased likelihood of aggressive accounting. The Altman Z-Score is in the distress zone according to the forensic summary, raising bankruptcy risk. Positive counterpoints: an Earnings Quality Score of 58.4/100 with an accrual component at 100/100 and a Piotroski F-Score of 5/9 provide limited comfort that some underlying operational metrics remain intact. Overall, forensic signals are elevated and materially lower confidence in reported results until clearer remediation actions are visible.
Track Record
Our model has a 12-year record on this coverage set with a hit rate of 36.4% (model directional calls matched next-year price movement in 36.4% of years). Historical average upside on calls was large (avg_upside_pct 183.99083333333337), but the low hit rate indicates outcomes are binary: occasional large payoffs offset frequent misses. Given the low historical hit rate, model outputs for SD4 should be treated as directional scenario inputs rather than high-confidence forecasts.
Written by a language model on 2026-08-10 from this page’s own model outputs and financial statements, and may quote figures from that date. Descriptive analysis, not investment advice — no buy, sell or hold recommendation is given or implied.