TRIO-TECH INTERNATIONAL (TRT) stock intrinsic value fair value estimate & owner earnings analysis

Trio-Tech International is a California corporation run from Singapore. It earns money three ways: testing and burn-in services for chip makers performed in its own laboratories in Singapore, Malaysia, Thailand and China; burn-in boards and test equipment that it designs and sells; and a smaller Industrial Electronics unit selling aerospace-related products, equipment and distributed components. What changed since our last review. The fiscal 2026 annual report (year to June 30, 2026) was filed after our previous estimate. Revenue rose 72% to $62.6 million from $36.5 million, almost all in testing: that segment grew 99% to $49.0 million after the Malaysian laboratory began final-test services for AI chips in the first quarter, and Industrial Electronics grew 15% to $13.6 million. That new work is low margin, with no capital investment and little risk taken. Gross margin fell to 16.6% from 25.1%, gross profit rose only $1.3 million to $10.4 million, and the company still reported a small operating loss of $0.2 million because administrative costs rose 19%. One customer was 38.7% of revenue (3.5% the year before) and the three largest were 57.6%. Testing backlog rose to $19.8 million from $6.7 million. Testing in China fell as some customers moved their work to other countries because of tariffs. The previous estimate was built on a different basis, so its figures are not comparable. How we measure earnings. We take operating cash flow, remove the swings in receivables, payables and inventory, subtract stock compensation as a real cost ($0.57 million in fiscal 2026) and subtract only the equipment spending needed to stand still, treating the part of capital spending that tracks revenue growth as growth spending. We then add back the small interest paid (the company's debt is only $0.4 million of bank loans) and handle that debt separately. Over the last ten fiscal years this cash flow ranged from $0.3 million (fiscal 2019) to $3.5 million (fiscal 2022), with a median of $1.78 million a year, about $0.17 per share on today's share count. We use the ten-year median because the decade spans a full chip cycle, including the 2019 to 2021 slump when revenue fell 23% from $42.4 million to $32.5 million, and because fiscal 2026's $2.3 million came from one new, thin-margin, single-customer stream that has not yet proven repeatable. Fiscal 2025 was $0.9 million. The recent figure is about 160% above the prior year; our first-year assumption is lower than that, for the lease reason below, so no reversal needs defending. Committed costs. Operating lease obligations jumped from $0.9 million to $5.5 million with the new Malaysian plant. Scheduled payments are $2.3 million in the coming year and $2.0 million in the next, against $1.2 million to $1.5 million a year recently. In every case we charge the excess, about $0.95 million in year one and $0.70 million in year two, and after that assume leases renew at the historical cost. Base case: the business earns the $1.78 million median for ten years ($0.83 million and $1.08 million in the first two years after the lease excess), then fades in a straight line to zero over years 11 to 25. Ten years of record support the level; nothing named supports it beyond that, so it fades rather than growing. Worth $16.8 million, or $1.59 per share. Bear case: the Malaysian AI stream proves temporary (a single customer with no disclosed contract term) and a chip downturn like 2019 to 2021 and fiscal 2025 follows. Cash flow is $0.5 million a year for two years (below the $0.6 million average of fiscal 2019 and 2025), which after the lease excess is a loss of $0.45 million and then $0.2 million, then $1.1 million a year for years 3 to 10 (the fiscal 2019 to 2021 average was $1.1 million), then a fade to zero by year 15. Worth $6.6 million, or $0.62 per share. Bull case: the Malaysian final-test ramp, the $19.8 million testing backlog, 58% growth in Singapore burn-in board sales and the new Malaysian capacity combine to hold the level of the three best years (about $3.0 million; we use $2.9 million) for ten years, then fade to zero over years 11 to 20. Worth $25.8 million, or $2.44 per share. What we left out: growth from AI testing demand, share gains, or margin recovery. None is contracted, and the +72% revenue year produced no operating profit, so there is nothing evidenced to credit above the historical level. Balance sheet. Cash is $21.4 million, short-term deposits $4.3 million and short-term investments $0.4 million, $26.1 million in all, helped by $9.5 million raised in a share issue during the year. We exclude $2.8 million of restricted deposits and hold back $5 million to run the business. Bank loans are $0.4 million with no meaningful maturity. That leaves net cash of $20.6 million, or $1.96 per share on about 10.55 million diluted shares (10.37 million outstanding plus in-the-money options). The company spent $3.5 million in December 2025 to acquire the remaining 50% of the Malaysian subsidiary, so we count all of its earnings. Per share, each case is business value plus $2.00 of surplus cash less $0.04 of loans: bear $0.62 + $1.96 = $2.58, base $1.59 + $1.96 = $3.55, bull $2.44 + $1.96 = $4.40. No case is stressed by its debt. Margin-of-safety price of $1.60. Confidence in the earnings level is low to moderate: ten years of results range from nearly nothing to $3.5 million, one customer is 38.7% of sales and operating profit is zero at record revenue. Confidence in the cash is much higher, because it is on an audited balance sheet with almost no debt. The worst plausible outcome is the bear case plus a longer slump. Five straight years of $1 million annual losses, worse than anything in ten years of record and not in any case, would still leave about $1.48 per share of net cash. If we are wrong that $5 million is enough working cash and $9 million is needed, net cash is $1.58 per share. If the AI work proves durable but stays thin-margin, the shares are worth roughly the base case. The price assumes a holding period of about five years. The company pays no dividend, has only about $1 million left on its buyback authorization and keeps issuing stock compensation, so the reward is retained cash: roughly $0.69 per share over five years in the base case. The $1.60 price is below the net cash per share and assigns no value to the testing business. On the base path the shares would be worth about $4.0 in five years, which implies roughly a 20% annual return, and the bear path implies roughly 12% a year. The price is set no higher because the cash is the only part we can verify. Could not be reconciled: the filing does not disclose the contract terms or duration for the largest customer, and why payables rose $5.3 million alongside the new work is explained only as higher procurement.