Hanmi Semiconductor Co., Ltd. (KRX:042700) closed at ₩230,000 on Friday, September 4. It rose 9.26% on the day. The stock had fallen for three straight sessions before that. Friday's move came with the broader Korean chip rally. It came two days after the delivery deadline on the company's only disclosed order. The company's market capitalisation was ₩21.9 trillion at Friday's close. That's roughly $15.6bn at an approximate ₩1,400 per dollar.
Hanmi makes thermocompression (TC) bonders. These machines stack memory dies into high bandwidth memory. HBM4 is the fourth generation of that memory. It's now ramping for Nvidia's next accelerators. I wrote on August 28 that the second-quarter operating margin of 51.9% was the thing to watch, because a margin that high invites competitors. This piece is about a different line in the same report. Hanmi booked a record ₩251.2bn of revenue in the quarter and ₩130.3bn of operating income. It also burned cash. Operating cash flow was negative ₩18.2bn in the quarter, after negative ₩38.4bn in the first quarter. Cash on the balance sheet fell from ₩276.2bn at the end of 2025 to ₩88.6bn at the end of June.
That's not a contradiction. It's what a capital-equipment maker looks like after a big batch of machines ships at quarter-end. The customer hasn't paid yet. But it does mean the record quarter exists on paper for now. The paper still has to turn into money.
The ₩187.6bn drop in cash over six months has three parts. Only one of them is about operations.
The first is the dividend. Hanmi paid ₩75.9bn of dividends in the first quarter, up from ₩68.3bn a year earlier. The company returned cash on the strength of 2025's results. Meanwhile, 2026 was still in its weak first quarter. Revenue was only ₩50.9bn.
The second is investment. Cash from investing activities was negative ₩69.1bn in the second quarter. The company filed in June that it was buying shares in another company. Korean press reported the target as SpaceX stock worth about ₩50bn. Property and equipment purchases were modest at ₩6.5bn for the quarter. So this wasn't capacity. It was a financial stake in a private US company, made from the balance sheet of a bonder maker.
The third is working capital, and that's the interesting one. Total current assets were ₩496.0bn at the end of June. Cash was ₩88.6bn and inventories were ₩143.9bn. That leaves about ₩263bn in other current assets, mostly trade receivables and contract assets. Three months earlier the same residual was about ₩83bn. Roughly ₩180bn of the quarter's ₩251bn of revenue was still owed to Hanmi at the end of June.
The half-year report explains the payment terms, and they're standard for the industry. Export customers pay 30% at purchase order and 60% before delivery. They pay the final 10% after acceptance testing. Domestic customers pay 30 days after sign-off. Under those terms, the receivables should clear within a quarter or two. If they do, third-quarter operating cash flow will look very different from the second.
The inventory line tells a cleaner story than the cash line.
Inventories peaked at ₩183.5bn in the first quarter of 2025. They held near ₩183.3bn at the end of that June and have come down since. They were ₩150.4bn at year-end. They rose to ₩173.0bn in March as the company built machines for the second quarter. They fell to ₩143.9bn by the end of June after those machines shipped.
That pattern matters because inventory at an equipment maker is machines waiting for a customer. A year ago Hanmi was sitting on bonders while SK hynix qualified a rival supplier. Now the inventory is moving out the door faster than it comes in. Whatever else the cash flow statement says, Hanmi is not stuck with product.
The geographic mix of sales confirms the shift. In the first half of 2026, exports were ₩228.5bn and domestic sales were ₩73.5bn. Two years ago the split ran the other way. In 2024, domestic sales were ₩328.3bn and exports were ₩230.6bn. The domestic line was SK hynix. The export line is Micron and the rest. Hanmi has rebuilt itself as an export business while its old anchor customer shopped elsewhere.
Here is the part that should bother anyone paying 99 times trailing earnings.
The half-year report has a section for orders on hand. It contains one entry: the HBM4 TC Bonder 4.5 Griffin contract with SK hynix. The contract was signed June 8. It was worth ₩44.2bn with delivery due September 2. The columns for quantity delivered and remaining balance are blank. The company adds a sentence saying its contracts could reveal customer investment plans. So it does not report orders beyond what it has already disclosed.
That's a legal choice, and a defensible one. But it means the disclosed backlog is 7.7% of last year's revenue, and it expired on Tuesday. Everything else supporting a ₩21.9 trillion valuation is orders investors are inferring. They infer them from press reports and Micron's capex. They also infer them from the fact that the second quarter happened at all.
Trailing twelve-month net income is ₩222.0bn, which puts the stock near 99x. Net income for fiscal 2025 was ₩214.0bn. To grow into that multiple, the second quarter has to repeat, and then some. The undisclosed order book is where that repeat lives or doesn't.
The bear case starts with the customer list. SK hynix qualified Hanwha Semitech as a second bonder supplier in 2025, and Hanmi's domestic sales collapsed as a result. The June order suggests SK hynix has come back for HBM4. DigiTimes reported in August that Hanmi was pulling ahead of Hanwha as HBM4 orders rebounded. That is good news, but it's also proof that this customer can and will switch. A ₩44.2bn order is a foot back in the door, not a long-term commitment.
The second risk is the one this piece is about. If the receivables from the second quarter don't clear on schedule, the cash position keeps falling. Hanmi has ₩134.7bn of total liabilities, almost all current. It paid ₩75.9bn in dividends in the spring. The company has ₩88.6bn of cash and just spent ₩50bn on a private-company stake. It has less margin for delay than it did a year ago.
The third is the margin itself. A 51.9% operating margin on a product where the customer has already qualified a competitor is not a stable state. Either Hanmi's pricing holds because HBM4 bonders are genuinely hard to replace. Or the next round of orders comes at lower prices. The half-year report doesn't say which.
The third-quarter report, due in mid-November, will show whether the ₩263bn of non-cash current assets converted. If operating cash flow turns strongly positive and cash rebuilds toward ₩200bn, the second quarter was real. In that case, the receivables were just timing. If cash stays below ₩100bn while revenue holds, something in the payment terms has changed.
Before that, watch DART for a new single-supply contract filing (단일판매ㆍ공급계약체결). Korea's exchange rules require companies to disclose contracts above a size threshold. Hanmi's last one was the June SK hynix order. Watch for a follow-on order from the same customer or a first disclosed contract with Micron. Either would be the first public evidence that the backlog behind the valuation exists.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.