Trio-Tech's AI Revenue Boom Runs on One Customer at a 16% Margin

Trio-Tech stock has roughly quadrupled on an AI chip-testing story that is easy to repeat and hard to square with the profit column. The company's third-quarter revenue jumped 124% to $16.5 million, the headline of a fiscal 2026 in which revenue nearly doubled. The story is real at the top line. The question the filings put on the table is whether it is a business worth roughly $100 million or a one-customer carry job dressed as a growth engine — because the same tripling of revenue produced almost no profit at all.

The best way to see it is to follow the revenue boom down to the margin line and then to the customer standing behind it.

The revenue boom that never reached the bottom line

Trio-Tech tests and burns in semiconductors for other companies — the reliability work that sits between chip fabrication and the finished device. For the nine months ended March 31, 2026, revenue came to $47.7 million, up 85% from the same stretch of the prior year, on the back of AI and automotive chip testing. That is a genuine acceleration, not paper round-tripping: revenue grows because customers send it devices to qualify and final-test. Note how abruptly the turn arrived — in fiscal 2025, the year before this surge, total revenue actually fell 14% to $36.5 million.

Now follow the same dollar down the income statement. Gross margin (the money left after the cost of the testing itself) was just 16% of revenue, down from 25% a year earlier. In the March quarter it ran at 16%, against 27% in the same quarter of fiscal 2025. The company's own explanation is honest and mechanical: the new volume is high-volume, low-margin final-test work, which dilutes the mix of the whole business.

The consequence is the number that should stop a reader cold. On $47.7 million of nine-month revenue, Trio-Tech earned just $62,000 of operating income. In the March quarter alone it reported a net loss attributable to common shareholders of $38,000. Restate that plainly: revenue grew 85%, and the company still could not reach seven figures of operating profit. Management calls the shift a mix effect, and so it is — but it is a mix effect with no margin leverage showing through after three quarters.

That is the gap between the price move and the economics. The stock is not expensive because of these earnings; it is expensive because the market is paying for earnings that the current operating model has not yet produced.

One customer is carrying the story

The concentration risk sits in a line the press releases do not headline. In the nine months ended March 31, 2026, a single customer accounted for 41.1% of revenue and 37.2% of receivables. The same dependence is not new — a single customer was 40.3% of revenue as far back as fiscal 2022 — but it matters more now because the entire growth surge is riding on one buyer's AI orders.

That buyer is not named in the filings. Management refers to a "leading AI chip manufacturer" and has announced a string of burn-in board orders supporting a next-generation AI GPU platform: roughly $5.3 million in the March quarter plus $2.5 million more, then $2.6 million in June and $3.8 million in July. Burn-in boards are the fixtures that stress-test chips at high temperature before they ship. Retail commentary has speculated the customer is a major GPU designer; that is speculation, and the company does not confirm it. What is on the record is the arithmetic: a low-margin, high-volume testing surge delivered by one customer that, if redirected or slowed, would take a large share of revenue with it.

There is a legitimate version of this bet. It is the story of a small, well-placed Asian test house that seized a seat in the AI supply chain, and whose margins could expand as Malaysian capacity (it leased an additional 104,000 square feet in Penang) fills and utilization rises. The counterweight is that the company has had three quarters to demonstrate that leverage and, so far, doubling-plus revenue has not produced it. The concentration and the thin margin are two red flags that compound each other: the revenue is narrow, and it is low-quality revenue.

Who finances the capacity

Growth like this costs money, and here the shareholder pays a second time. In April 2026, after the quarter closed, Trio-Tech raised roughly $10 million by selling 1,052,632 new shares at $9.50 each in a registered direct offering. On a shareholder base near 9 million before the deal, that is over a 10% dilution to fund capacity for AI and automotive demand. Cash stood at about $18.3 million at the end of March, before that raise.

Dilution to fund real capacity is defensible; it is the honest way a small operator builds out for a real order book. But it is also the tell of where this business actually is. A company whose revenue nearly doubled in a year did not generate enough internally to fund its own expansion, so it went to shareholders. Combine the two financings of the shareholder invoice — the thin profit and the new shares — and the buyer at current prices is paying a growth multiple for a business whose growth is one customer, sub-20% margin, and partly funded by diluting the very shareholders who are celebrating the revenue.

The price of the story

At a price near $10.50, the market values Trio-Tech at roughly $100 million. Against nine-month operating income of $62,000, that multiple is not meaningful as a profit ratio; it is a bet on the AI future. Insiders and institutions hold a large part of the shares, and roughly 900,000 shares are sold short against a float near 6.7 million — the market itself is visibly split on the story. The stock has been a spectacular winner, and that, not the fundamentals, is what the current price mostly measures.

For a beginner trying to decide whether this belongs on a watch list, the test is simple. Ask whether Trio-Tech has yet shown that its AI volume can be converted into durable margins. Three-quarters in, the answer is no: revenue up 85%, gross margin down nine points, operating income in the tens of thousands. The benign resolution is that high-volume final test matures into higher-value work and the Malaysia capacity raises utilization — that would justify revisiting it. The persistent-but-lawful case is that it remains a low-margin, one-customer test house whose revenue is real and whose profitability is not, leaving it priced on hope. There is no evidence of anything worse; the red flags here are concentration and margin quality, not fraud.

The number to wait on is the one already in the filings: whether the gross margin on this booming revenue ever climbs back toward what the industrial business earns, and whether the customer diversification follows the expansion. Until that margin line moves, Trio-Tech's $100 million story is a revenue story, not an earnings story — and a revenue story with a single customer behind it is the most expensive kind to own.