Borrowing from the bank to play the stock market

An electrical construction company has total assets of 37 billion dong. 21 billion of that has sat in a single stock for eight years.

A property company reports Q2 profit tripling. Revenue from selling homes: 400 million dong.

And at least four companies in this piece are borrowing thousands of billions from banks while simultaneously holding hundreds of billions of shares on the balance sheet, the most volatile asset a business can choose to own.

When a company abandons its own trade

In Q2 2026, domestic institutional money was the market's main support. The group bought a net ~20,000 billion dong in Q1, then came back for nearly 6,000 billion in May, exactly as foreign investors sold a net 19,600 billion, the highest in six months. The VN-Index touched 1,933 points in May before correcting hard. Plenty of businesses deployed right at that peak, and the consequences appear directly in the Q2 financial statements.

Among those institutions is a notable group: companies not in the financial business. A property company, a phone distributor, an electrical contractor. Here is the full list: what they hold, at what price, and how much they borrowed to do it.

NDN (Nha Da Nang): the most transparent portfolio, the lowest leverage

At 30 June 2026, NDN held a stock portfolio at cost of 659.4 billion dong, 48.6% of its total assets of 1,356.7 billion. That is the highest weighting on this list. The detail:

The difference: NDN has almost no bank debt. Liabilities at the end of Q2 were only 139 billion (from 58 billion at the start of the year), mostly dividends payable (50 billion) and amounts due to brokers VPS and Yuanta (nearly 40 billion), not financial borrowings. To buy stock, NDN drained nearly all its cash: from over 261 billion to 26.3 billion. So NDN's issue is not leverage but opportunity cost: all the company's cash is sitting in five tickers.

NTL (Lideco): borrowing to hold stocks, collateralised by its own deposits

A portfolio at cost of 597 billion dong, 29% of total assets of 2,060 billion:

Total impairment provision: 25 billion dong. Long-term borrowings rose from 14 billion to 165 billion in one quarter, and part of a 320 billion loan (for a social housing project) is collateralised by the very deposits NTL holds at that bank. The company pays interest to borrow back its own money, while using other money to buy TCH, DBC and HHS.

PET (Petrosetco): betting on a new shareholder's ecosystem, with borrowed money

The equity portfolio is around 236 billion at cost (the equity portion of a 333 billion trading securities book, the rest being certificates of deposit and funds):

The whole VIX, GEX, EIB, VPB group belongs to the Gelex ecosystem, while PET is simultaneously establishing three infrastructure joint ventures with Gelex, and PET's new large shareholder is connected to that group. Vietcap estimates the portfolio's unrealised loss at 45–50 billion dong.

In parallel, PET has 7,222 billion dong of short-term borrowings (up 167 billion since the start of the year), total liabilities above 11,100 billion, nearly 80% of total capital. In Q1 2026 PET's financial activities alone lost nearly 38 billion, with the company citing sharply higher interest costs to fund working capital for a new distribution model. By Q2 the company began shrinking short-term financial investments (from 4,664 billion to 2,945 billion), but the enormous short-term borrowings remain.

DGW (Digiworld): exiting at the right time, but swapping positions rather than cutting risk

At the end of Q1 2026, the portfolio was nearly 800 billion at cost, concentrated in TCB and VPB. In Q2 the company sold all TCB and VPB, cut HPG (from 227.2 billion to 100.6 billion), and put nearly 200 billion into KBC, bringing the portfolio down to around 300 billion.

Chairman Doan Hong Viet publicly admitted at the April 2026 AGM: in 2025 the company lost relatively heavily investing in FPT, and it would tighten equity investment, shifting toward structured products and fixed-income assets. That is a rare admission of error on this list.

But DGW's leverage remains high: short-term borrowings and finance leases above 3,214.5 billion out of total liabilities of 8,778.7 billion. Debt to equity at end-2025 was 0.83x, against a self-imposed ceiling of 1x, meaning there is still room to borrow more.

VE1 (VNECO 1): one ticker, eight years, more than half the company

VNECO 1's total assets are only about 37 billion dong. Of that, nearly 21 billion, more than half, sits in a single stock: MAS (Da Nang Airport Services). The position has "sat still" since 2018 and carries a provision. The company now has nine employees, recorded no revenue in the first half of 2026, and a net loss of nearly 1.7 billion.

CII: buying shares while the target company's executives are arrested

In early June 2026, CII and a subsidiary raised their holding in PC1 to over 23.7 million shares (5.77%), immediately after a series of PC1 senior executives were charged and detained in May 2026. PC1 stock fell from over 30,000 (February 2026) to around 17,400, nearly halving. CII's CEO Le Quoc Binh said this was purely a financial investment in PC1's energy project portfolio, not an attempt to take part in management.

CII also holds 20.48 million HUT shares, market value 354 billion, about 35 billion below cost (so cost ~389 billion).

On debt: liabilities at the end of Q2 2026 reached 12,744.7 billion (+22% since the start of the year). Long-term borrowings and finance leases over 4,020.6 billion, short-term borrowings around 3,091 billion, plus 3,257.5 billion of convertible bonds. Q2 profit fell more than 91% to 16.9 billion; provisions for long-term financial investment losses rose to nearly 364 billion.

L14 (Licogi 14): not disclosing what it holds

L14's securities investments were roughly flat: 86.59 billion at the end of Q2 2026 against 86.95 billion at the start of the year. But impairment provisions leapt from 15.26 billion to 25 billion, meaning nearly 29% of the portfolio has been provided against. Notably, L14's recent financial statements do not disclose which specific tickers it holds, unlike NTL, NDN and PET which all give detail. That in itself is a question.

L14's leverage is very low: short-term borrowings of only 9.36 billion, long-term 440 million. L14's problem is not debt. It is that 25 billion of provisions has wiped out almost all of the company's half-year profit (1.62 billion, down 88% year on year).

Four tests before a dong is allowed to leave the core business

This is the question Warren Buffett poses to Berkshire Hathaway every year: a dong of retained earnings, if not reinvested in the core business, must create at least a dong of market value. Otherwise retaining it destroys value, even if the company is profitable on paper.

One, against a bank deposit or a government bond. With the 10-year government bond around 5% a year, NTL's 597 billion placed safely produces about 30 billion a year, risk-free. In Q2, NTL's equity portfolio produced 16 billion of gains, alongside 25 billion of provisions. This is the margin of safety in its simplest form: the gap between a certain return and a risky one.

Two, against buying back your own stock. If management believes the company is undervalued, no investment is more certain than repurchasing its own shares at that price.

Three, an investment serving the core business: an acquisition, a supply chain link, adjacent expansion. This is the reason PET gives for investing in the Gelex ecosystem. But the company does not disclose the portfolio in enough detail for shareholders to separate strategy from speculation by another name.

Four, if none of the above is viable, repay debt or return the money to shareholders. Repaying debt produces a return exactly equal to the interest rate being paid, with no market risk. For PET with 7,222 billion of short-term borrowings, every dong used to pay that down is a dong of certain return, against a dong in VIX currently down 26%. For CII paying interest on more than 10,000 billion of debt and bonds, the same question applies to the PC1 purchase. Neither company publicly answers it.

And for a company with no debt that also cannot find a better use of capital, like NDN or L14, the right answer is not to quietly hold the money in a volatile portfolio. It is to pay a dividend or buy back stock if the company is undervalued. That is the honest admission: "we cannot find a better use for this capital than shareholders can themselves."

An unrealised gain is not a gain

The clearest case is NTL's investment in TCH. At the end of Q1, NTL held over 21.2 million TCH shares at a cost of 308 billion, with a market value of 369 billion, an unrealised gain of 61 billion. Three months later the company had added to the holding, but TCH had fallen nearly 40% since early May. The result: the unrealised gain shrank to 354 million, nearly all of it gone in one quarter.

This is one of the common psychological mistakes of Vietnamese investors: with no predetermined selling threshold, a paper gain is only ever a paper gain, until the day it disappears.

The one who answered correctly

Across this entire list, exactly one company did the opposite: Vinh Hoan (VHC). VHC's equity portfolio is down to 44.5 billion (DXS 21 billion, NLG 3.5 billion, the rest in fund certificates), with a provision for a 23% loss. At the same time VHC announced a buyback of 15 million treasury shares at a maximum of 63,000 a share, equal to 945 billion dong.

The same money, two fundamentally different choices: buy other people's shares, or buy back your own. VHC chose the second, the correct answer to test number two.

The QuietVC lesson

Four numbers worth checking before concluding a business is "playing the market": portfolio cost over total assets; concentration in a single ticker; whether the company is borrowing to fund those investments; and finally, whether the company ever considered repaying debt or paying a dividend instead of buying shares.

That last question is rarely answered in any financial statement. Which is exactly why it is the most worth asking.

Silent discipline. Compound value.