PNJ and the Eight-Year Promise Nobody Charged For

In twenty days, a listed Vietnamese company paid its customers more money than it took in from sales, three and a half times more. Not because the company went bankrupt. Because eight years earlier it had bundled a promise with every sale, and nobody, the company included, ever charged for that promise.

A month without precedent in seventeen years of listing

On 2 July 2026 the former director of the diamond grading laboratory under PNJ (P-Lab for short) was charged in a case alleging the smuggling of more than 28,000 diamonds, bought in India, consolidated in Hong Kong, flown into Vietnam illegally, with the allegation that the original GIA certification numbers were ground off, P-Lab codes engraved in their place, and new certificates issued.

The market reacted immediately. The stock hit its floor limit on 3 July. But the more notable reaction was at PNJ's own counters: customers began bringing diamond jewellery back to sell, at a rate that climbed day by day.

From 1 to 20 July, PNJ recorded revenue of 1,588 billion dong, while the value of goods bought back from customers reached 5,900 billion dong, 3.72 times revenue. On average the company paid out 295 billion dong a day to buy back what it had sold.

On 21 July PNJ had to change the policy: instead of paying immediately, sellers would be paid on a five-instalment schedule over 120 days (10% at once, then 20%, 25%, 25% and 20% at the 30, 60, 90 and 120-day marks), or were encouraged to exchange for other products instead. The buyback rate then fell by about half, to roughly 145 billion dong a day. But by the end of July the total bought back had reached about 7,062 billion dong.

PNJ closed July down 50.87% from the start of the month, from 63,100 dong to 31,000. Against the all-time high of 127,000 dong set in late January 2026 (equivalent to 84,667 after adjusting for the 2:1 bonus share issue in April), the fall was 63.4%. On 30 July PNJ published second-quarter results with an after-tax loss of 283 billion dong, the first quarterly loss since the company began disclosing results in 2008.

The question worth dwelling on is not "how far did the stock fall," but: what mechanism allowed a criminal case at a subsidiary to produce cash outflows of 3.72 times revenue in twenty days?

The nature of the business model: an eight-year-old put option

When a customer buys diamond-set jewellery at PNJ, they receive two things on the same invoice: the item, and an attached right, the right to bring the item back and sell it to PNJ at any time, at a fixed percentage of the original invoice value, regardless of what that stone is worth on the market at the moment of resale.

The buyback policy applies to invoices issued from 11 June 2018, with the following specific rates (as published and effective to 4 July 2026):

Loose diamonds:

Diamond-set jewellery:

Four features of this table determine the whole story of last month.

First, the strike price is fixed to nominal value, not to market price. A customer who bought a 100-million-dong diamond ring in 2021 has the right to sell it back for 80 million on any day, even if the stone's real market value today is only 50 million. Called by its proper financial name, this is an American put option, perpetual, free of charge, unhedged. PNJ's loss when a customer exercises is not the 20% discount. It is the gap between what the company must pay (80 million) and what it can actually recover by reselling that stone.

Second, the certificate is almost meaningless on the buyback side. On the sell side, the P-Lab certificate is the basis for a high price. On the buyback side the price depends only on the invoice, and the certificate's role is close to zero. This creates an important paradox: financially, the certification scandal did not add a single dong of obligation. The obligation was fixed on the day of sale, whether the diamond was genuine or not. But behaviourally, the scandal was the detonator: it created no new liability, it merely caused tens of thousands of people to decide at once to exercise a right that had lain dormant for years. In other words, this obligation existed in full on PNJ's economic balance sheet for eight years without ever appearing on its accounting balance sheet.

Third, there is an adverse selection built into the policy design. PNJ excludes from the buyback obligation exactly the large-stone category (above 10 mm), and that is the category whose world price is rising on scarce supply (the Argyle mine closing, De Beers cutting output). Conversely the obligation applies in full to the 3–8.6 mm range, exactly the range where global commercial diamond prices are down 15–30% from the 2022 peak. There is no evidence this was deliberate, but the economic result is a book of obligations that is systematically adversely selected.

Fourth, and most important for sizing the risk correctly: the gold leg and the diamond leg are two entirely different problems, even though both sit inside the single figure of 7,062 billion dong.

Of the 7,062 billion dong paid out, only the diamond leg creates a real economic loss. The gold leg creates only cash pressure. That distinction substantially narrows the true scale of risk against the crude reading of "7,000 billion lost."

Unpacking second-quarter profit: two provisions, two stories

PNJ recognised 1,233 billion dong of provisions arising in the second quarter of 2026 alone, split across two lines in two different places in the income statement:

That split matches the two-leg distinction above exactly: one is an obligation that was never on the balance sheet, the other is the impairment of an asset already sitting on it.

Strip both provisions out of the quarter and the picture changes completely:

On the core alone, the second quarter of 2026 grew 62% year on year, a world away from the 283-billion loss reported in the press.

The inventory provision alone (368 billion) pulled gross margin down about 4.3 percentage points, more than the entire 3.1-point decline the company attributed to "a rising share of 24K gold reducing gross margin." That explanation, on its own, does not account for the quarter's numbers.

There are two ways to read the 1,233-billion provision, and both have a basis.

Reading one, conservatism beyond the standard: very few companies provision for a risk that had not yet crystallised at the cut-off date (30 June). PNJ chose to absorb the full shock at once. This is how the company handled a different shock ten years ago (see below).

Reading two, earnings smoothing: the provision was booked before the actual buyback transactions were closed, entirely on management's estimate of recoverable value. The consequence is that the third quarter, which absorbs the July buyback peak, will look considerably less bad than it would have otherwise. One bad quarter traded for a less bad one.

There is a way to test between the two readings: back out the loss rate that the 865.5-billion provision itself implies.

Cross-check with market logic: the customer paid 100, PNJ pays back 80. The wholesale value of a piece of jewellery retailed at 100 is roughly 65–75 (retail margins on diamond jewellery usually run 30–50%). Subtract a further decline of about 25% in global commercial diamond prices since the original purchase, and recoverable value is around 49–56. The loss on every 80 paid out lands around 24–31, which is 30–39%.

Two independent calculations meet around 30%, and that figure lines up fairly closely with what has actually happened to international diamond prices:

PNJ is not facing a uniquely Vietnamese shock. The entire global natural diamond industry is going through a collapse in its scarcity premium, driven largely by the rise of lab-grown stones. That De Beers, the group that once shaped an entire market, had to be written down by 6.8 billion USD makes PNJ's 865-billion provision (about 33 million USD) look entirely unremarkable.

Quantifying the number the company has not disclosed

The most important question in the whole story, the total value of diamond jewellery in circulation that PNJ is still obliged to buy back on demand, has never been disclosed. It can be estimated two independent ways.

Method one, inferring the total obligation from observable flow:

If diamonds are 30–40% of that base, the diamond obligation lands around 6,000–19,000 billion dong.

Method two, bottom-up from cumulative retail revenue since the policy took effect (11 June 2018):

The two methods start from completely different data and converge on the same range: 7,000–17,000 billion dong, equal to 44–107% of PNJ's market capitalisation at the end of July (15,863 billion) and 50–122% of shareholders' equity.

From there, the unprovisioned potential loss can be estimated:

Three factors could soften that figure: the share of customers who actually exercise is usually far below the theoretical maximum once panic passes (the buyback rate halved immediately after the 120-day policy took effect); the loss only crystallises if the company is forced to liquidate, whereas PNJ can unset the stones, redesign and sell through its own 431 stores at retail prices, a structural advantage no small gold shop has; and the exchange rate (85–95%) is higher than the buyback rate (75–90%), so every customer who takes a product instead of cash converts an outflow into a sale.

Three factors work the other way: the 120-day deferral only moves the obligation, it does not erase it, and the final instalment falls around November 2026; global commercial diamond prices have still not found a floor below 1.5 carats; and if confidence does not recover, the cumulative exercise rate could be far higher than assumed.

The moat: measure it against the wrong benchmark and you reach the wrong conclusion

The conventional analysis would say PNJ's diamond moat has been destroyed, because the whole ability to sell diamonds at high margin rested on one proposition, that diamonds sold by PNJ are independently and credibly graded, and P-Lab existed precisely to vouch for that proposition. When the apparatus allegedly used to legitimise smuggled goods is the company's own seal of assurance, that is not an external risk striking the moat. That is the moat being hollowed from the core.

That is true, but only half true, because it is a self-referential measurement, comparing PNJ with itself before the event. An economic moat, by definition, must be measured relatively, against competitors.

Looked at that way, the picture is completely different. Between June and July 2026, more than 30 businesses trading gold and diamonds in Vietnam closed, stopped trading, or suspended their buyback service. In the five days after the case was announced, eight more shops announced temporary closures. Several once-known chains had to extend payment periods or shut entirely, leaving customers holding jewellery with nobody to sell it to.

The cause lies in the industry's shared structure: most diamond businesses in Vietnam commit to buying back at 80–95% of the original selling price, and the secondary market for diamonds depends almost entirely on the seller itself, since each business only buys back what it sold. It is a system of self-issuing "banks" of put options, with no exchange, no independent market maker, no insurance mechanism in between. When confidence collapses, all of them are run on at once, and whoever lacks the capital closes.

PNJ paid out 7,062 billion dong and is still standing.

The full chain of reasoning: the shock destroyed confidence in buyback commitments across the industry; more than 30 businesses proved that such commitments are worthless under real pressure; consumers learned that a commitment is only credible if the issuer is large enough; and barriers to entry jumped, because from now on anyone wanting to sell diamonds in Vietnam has to demonstrate the financial capacity to withstand a run. The only listed company, with audited public accounts, 13,912 billion dong of equity, no long-term debt, 431 stores, and a real-world stress test just passed, is PNJ.

Put briefly: the industry's absolute moat was destroyed; the relative moat of whoever is still standing widened. Before July 2026, a buyback commitment was merely a condition of entry, everybody had one. After July 2026 it became the only differentiator that still means anything.

Three limits to that argument deserve honesty. The industry is shrinking before it consolidates; the diamond market is "frozen," purchasing power is sharply down, so PNJ will take a larger share of a smaller pie, at least for the next 12–24 months. PNJ is the company named in the case itself; the other 30 died for lack of money, but PNJ is where the detonator sat, so it carries a brand disadvantage the other survivors do not. And the cost of acquiring that position was far from small: 7,062 billion dong paid out and a record quarterly loss.

Set side by side, PNJ's moat has a clear, measurable ceiling. The listed Hong Kong jeweller Chow Tai Fook earned a 32.3% gross margin in its latest financial year against PNJ's 22.0%, a 10.3-point gap, mostly from a higher share of high-margin designed pieces (Chow Tai Fook recently launched a collection that did more than 500 million Hong Kong dollars of retail sales in its first two months). That is both the practical ceiling of PNJ's current model and the room for improvement if the "multi-format lifestyle retail" strategy the company is pursuing (expanding into watches, accessories, partnerships with entertainment brands) works.

On the market's long-term room, gold jewellery consumption per head in Vietnam remains low for the region:

If Vietnam merely reached China's level, the market would grow 2.7 times, a genuine long-term growth thesis anchored in urbanisation and an expanding middle class. But it has to be read alongside the short-term data: Vietnamese gold jewellery consumption fell 20% in 2025 against 2024, as sharply rising gold prices pushed consumers from jewellery into bullion. The long-term room is real, but the short-term cycle is running the other way.

A licence tied to the company's own compliance record

There is a legal link that gets little attention, and it turns PNJ's compliance record into an economic variable rather than merely a governance story.

Decree 232/2025/ND-CP, effective 10 October 2025, abolishes the state monopoly on producing gold bars and moves to a licensing regime for qualifying companies and commercial banks. Licensed companies may also import and export gold bars and raw gold under annual quotas allocated by the size of their charter capital.

The licensing conditions under Article 11a cover four items, two of them notable:

Across the whole market, only three companies satisfy both the capital condition and the existing gold-bar trading licence: SJC, PNJ and DOJI.

Meanwhile PNJ's compliance record over the last 14 months contains three notable entries: a State Bank inspection conclusion published in May 2025 finding that PNJ's parent company had breached reporting requirements on gold bar trading, along with indications of breaches on product labelling and anti-money-laundering; an administrative penalty decision in December 2025 against the manufacturing subsidiary relating to anti-money-laundering; and another inspection conclusion published in January 2026, with the file passed to an inter-agency body.

Put the two facts together: PNJ's record of administrative penalties, easy to read as a mere governance detail, is in fact a binding legal condition on one of the largest strategic opportunities the company has had in a decade. That helps explain part of why PNJ chose to honour 100% of the buyback commitment rather than seek to delay or negotiate it down, and why the company volunteered a dense stream of disclosures to the stock exchange: a business that needs to prove it has "completed remedial measures" has a very strong incentive to honour every other commitment, however costly.

The same link also raises the real seriousness of the legal risk from the P-Lab case: any further administrative sanction at parent-company level relating to gold trading would threaten that compliance condition, and with it the whole opportunity to obtain a gold bar production licence.

The board, and three governance points worth noting

PNJ's board has 9 members, 3 independent, 4 women, above the minimum compliance standard. The chair, Cao Thi Ngoc Dung, founded the company and has chaired it since 2004. The current chief executive, Phan Quoc Cong, took office on 3 April 2026, about 13 weeks before the crisis, succeeding a predecessor who had held the role since 2018. Mr Cong comes from consumer goods and distribution, having founded a large domestic personal-care brand; notably he registered to buy 1 million PNJ shares in the middle of the crisis, from a prior position of owning none.

Three governance points for the record:

First, a conflict in the supervisory structure. One board member, who also sits on the Audit Committee, simultaneously headed the member unit responsible for grading, the very unit where the incident occurred, from early 2025, roughly half a year before the case. An audit committee exists to supervise group units independently; when a member of that committee also heads the unit to be supervised, independent control is structurally void, whatever the individual's competence or good faith. After the event the company set up a special oversight team including independent members, the right step, but one that should have been there from the start.

Second, internal control shows signs of lagging the pace of expansion. The 2025 annual report voluntarily published the results of an in-depth internal inspection that found 345 operational errors, an average of 4.61 per store; the share of staff meeting the risk-management knowledge standard rose from 68% (2024) to 91% (2025), meaning that in 2024 one in three staff fell short. Publishing those numbers at all is a rare level of transparency for the Vietnamese market, but the numbers themselves show a control system running behind the company's scale.

Third, a precedent from ten years ago, set out in detail below, involving a bank in which PNJ was a founding shareholder, where a former senior executive of that bank (related by family to PNJ's leadership) was found to have committed serious violations and ended up facing criminal proceedings. This is not the first time a unit in the orbit of the founding family has had its head involved in serious wrongdoing connected to precisely the role entrusted to them.

On the more positive side, management's capital allocation record is fairly solid: equity grew from 5,242 billion (2020) to 13,275 billion (2025), mostly from retained earnings (there was only one outside raise, 1,425 billion in 2022, deployed to exactly the four committed purposes with nothing left over). A rough estimate puts the incremental return on each dong of retained earnings over that period at about 26.6%, many times the opportunity cost of a deposit or a government bond. A cash dividend of 20% of par has been maintained for six consecutive years without interruption, including the hardest year.

One underappreciated point: the 2025 annual report, published in early April 2026, nearly three months before the event, listed "liquidity risk from buyback activity, particularly when gold prices rise quickly" as one of six material risks, and described the company building a measurement tool and running periodic stress tests on gold price movements. Management identified the risk mechanism correctly. But the stress test was built on the gold price variable, which moves gradually and allows reaction time, while the actual shock came from the confidence variable, which collapsed in a single morning and originated in their own subsidiary. The lesson is broader than PNJ: a risk map can get the mechanism right and still draw the trigger axis wrong.

Another crisis, ten years ago

This is not the first time PNJ has faced a major shock to confidence, and looking back gives a base rate worth comparing against.

In 2015 a bank in which PNJ was a founding shareholder (an initial contribution of 8 billion dong, at one point a 7.7% stake) was placed under special supervision by the State Bank, after a former senior executive of that bank, related by family to PNJ's leadership, was found to have committed serious violations and subsequently faced criminal proceedings. PNJ immediately provisioned 310 billion dong within 2015, cutting full-year after-tax profit by 37.5% to 160 billion; in the fourth quarter of 2015 alone, profit fell 81% to just 9.8 billion. By the end of 2016 the company had provisioned 100% of the investment's book value (395 billion).

Comparing the intensity of the two shocks as a share of the prior year's profit, the 2015 shock was actually larger: the 310-billion provision equalled 121% of 2014 profit, while this year's 1,233-billion diamond provision equals only 43.6% of 2025 profit.

From the profit trough of 160 billion dong in 2015, PNJ recovered to 1,194 billion in 2019, 7.46 times in four years, and compounded profit at 33.3% a year over the full decade to 2025.

That is the strongest base rate in the whole file, but three differences stop it transferring directly to the present situation. The 2015 investment was a financial holding outside the core business; losing it did not affect selling jewellery, whereas the diamond crisis strikes directly at the highest-margin segment. The 2015 shock did not trigger a wave of customers returning goods; the 2026 shock did. And most importantly, the 2015 provision was a closed loss, fully provisioned and done, whereas the diamond buyback obligation is still running with no known end point.

The historical base rate therefore supports confidence in the business's capacity to recover while the core model is intact, but it says nothing about how large the current obligation is or how long it will last.

The balance sheet: third-quarter projection and the liquidity pinch point

At 30 June 2026 PNJ held 4,002 billion dong in bank deposits against 3,989 billion of short-term borrowings (no long-term debt), a slight net cash position of 12.5 billion. Operating cash flow for the first half was positive 1,990 billion, a full reversal from negative 108 billion in the same period last year.

But July's own cash arithmetic is far more striking: total cash actually paid out during the month (5,900 billion paid immediately in the early phase, plus 10% of the amount shifted onto the deferred schedule) was around 6,016 billion dong, more than 50% above the entire deposit balance held at the start of the month. The gap of roughly 2,014 billion had to be covered by sales cash flow during the period or by new borrowing. In other words, the deposit balance at the end of July is almost certainly well below the 4,002 billion held at the start, or borrowings have risen substantially. This is an important data point not yet observable; it will only appear in the third-quarter report.

On inventory: adding the 7,062 billion bought back in July to net inventory of 15,149 billion at 30 June, less the estimated cost of goods sold, net inventory at the end of July could have reached about 20,311 billion dong, up roughly 51% from the end of the first quarter and 1.28 times the current market capitalisation. This is an accumulation of working capital on a scale with no precedent in the company's history. But unlike a bank facing a run, PNJ is exchanging cash for real assets (mostly gold with a transparent market price), not cash for receivables.

The 120-day deferral policy will also create an entirely new line on next period's balance sheet: payables to customers. On an illustrative scenario (a further 12,000 billion bought back in the third quarter, 10,000 billion of it on the deferred schedule), payables to customers outstanding at the end of the third quarter could be around 7,000 billion dong, pushing total liabilities to about 14,105 billion and the debt-to-assets ratio from 34% (at 30 June) to about 54%. A company known for having no long-term debt may appear in its third-quarter report with a debt ratio above 50%, not because it borrowed, but because it signed an IOU to its own customers.

In economic substance, the policy converted an off-balance-sheet obligation into an interest-free 120-day loan from tens of thousands of individual customers.

PNJ's three largest cash outflows of the year are likely to fall in the same window: the final instalment of the 120-day schedule (around November 2026), the maturity of most of the current short-term borrowings (late August to November 2026), and the peak of raw-material purchasing ahead of the Tet season. That is the sharpest liquidity pinch point identifiable in advance from available data. Supporting factors: the fourth quarter is usually the strongest selling quarter of the year, and the company still has unused credit lines at several banks.

A durability test on the whole picture: net inventory value would have to fall 91.8% to wipe out PNJ's current equity entirely. Even in a combined scenario, a 25% fall in the gold price plus the entire outstanding obligation exercised at the upper-bound loss rate, about 43% of current equity would remain. That helps separate three risks that need to be kept distinct: insolvency risk (low, on these numbers); liquidity risk (real, and nearly realised in July); and the risk of a permanent re-rating from a higher governance risk premium, the hardest to measure and probably what the market is reacting to most strongly.

Which scenario is the market pricing?

At 31,000 dong at the end of July, PNJ's market capitalisation was about 15,863 billion dong, a trailing twelve-month P/E of roughly 5.48 and a P/B of about 1.14, far below the company's own historical 15–18 times P/E range.

A reverse discounted calculation, assuming a 12% cost of capital and zero long-term growth, implies that the market is pricing perpetual earnings of about 1,904 billion dong a year, roughly 10% below 2024 profit and never growing again. For a business that compounded profit at 21.5% a year over the last five years, that is a fairly harsh assumption, though not an unreasonable one if the high-margin diamond segment shrinks permanently and the buyback commitment has to be curtailed.

Another view: enterprise value (market capitalisation less net cash) is currently only about 1.05 times net inventory value, meaning the market is valuing all 431 stores, two factories, more than eight thousand staff and a machine that generated 20–25% ROE for six years at essentially nothing, plus only about 5% on the book value of the gold and jewellery inventory.

Applying different methods, earnings capitalisation at various discount rates, multiples from a regional comparable set (Chow Tai Fook and the normalised Chinese jewellery benchmark some international analysts use), asset value less the estimated outstanding obligation, and a three-scenario growth model, produces a fairly wide range, which correctly reflects the genuine uncertainty:

The most notable thing in the whole valuation section is not any single number but the gap between the methods. The large difference between the optimistic and pessimistic cases does not come from disagreement about the core business model. It comes almost entirely from one variable: the true size of the remaining buyback obligation, a figure that to this day nobody, the company included, has confirmed.

The milestones that will answer the open question

A few specific data points, once published, will substantially narrow that uncertainty:

• The reviewed half-year report, which will show whether the independent auditor accepts the basis for the 1,233-billion provision, and will reveal for the first time the actual cash balance and borrowings after July's turmoil.
• The third-quarter financial statements, which will show the actual buyback value for the quarter (against July's 7,062 billion) and the size of the new payables-to-customers line.
• The progress of the diamond smuggling trial, which will clarify the actual scope of the investigation, still described as expanding.
• Any announcement on gold bar production licences for the three qualifying companies, which will be indirect evidence of how the regulator views whether PNJ has completed its remediation.
• Gross margin in coming quarters, which will show whether the enormous volume of repurchased goods is successfully re-commercialised through retail, or has to be liquidated at a higher loss.

The QuietVC lesson

PNJ is not a story about accounting fraud or a dishonest business. It is a story about a kind of risk the balance sheet was never designed to display: a commitment bundled in free of charge, accumulating quietly for years, becoming visible only when an outside event triggers it all at once.

This kind of risk is not confined to gold and gems. In Vietnam it appears under other names: guaranteed rental yields of 8–12% a year in condotel and resort property projects, commitments to buy back corporate bonds before maturity, cross-guarantees between companies inside the same family group, or simply an over-generous returns policy at a retail chain. What they share: none of them appears in a P/E, a P/B, or a debt-to-equity ratio, the most familiar measures a retail investor uses to judge whether a stock is "cheap."

A simple test when you meet a business with a commitment of this kind: ask whether the strike price is fixed or floats with the market; ask who decides when the right is exercised, the customer or the company; and ask whether the business charges for or hedges that commitment. If the answers are fixed, customer-decided, uncharged and unhedged, that business has written a naked put option, whatever the financial statements call it.

Against a 10-year government bond yield around 5% a year, the default benchmark for judging any risky investment in Vietnam, the right question is never "how much cheaper is this stock than its own history," but "is the risk premium on offer adequate for a risk that cannot be measured precisely." For PNJ, the answer depends almost entirely on one number that, as of this writing, nobody has published.

Silent discipline. Compound value.