LG Innotek Co., Ltd. (KRX: 011070) just printed the best quarter in its history and still kept only about four and a half won out of every hundred it billed. Revenue reached ₩5.53 trillion in Q2 2026, a record, up 40.5% year over year. Operating profit was ₩245.8 billion against ₩11.4 billion in the same quarter of 2025, which is the twenty-fold jump every headline led with. The year-over-year math is real. It's also flattered by a base quarter in which the company barely broke even and lost ₩8.7 billion at the net line.
The number I'd anchor on instead is the margin, because it's the one that has stubbornly refused to move for a decade, and it's the one the AI substrate story has to change for the shares to make sense at ₩611,000. This piece argues that LG Innotek's problem was never demand. It's that the customer who supplies the demand also sets the price.
Strip out the base effect and Q2 looks solid rather than transformative. Operating margin was 4.4%, down from 5.3% in Q1 2026 on essentially identical revenue (₩5.53 trillion both quarters). Gross profit was ₩531.0 billion, or 9.6% of sales, versus 10.1% the quarter before. So the company added no revenue sequentially and gave back margin.
That 9.6% is the number that tells you what kind of business this is. A components maker with pricing power runs gross margins in the twenties or thirties. Single digits is what you earn when a customer hands you a specification, a volume, and a price. Optical solutions, sold almost entirely to Apple, accounted for roughly 82% of Q2 revenue according to reporting on the quarter. LG Innotek is supplying the variable-aperture main camera module expected in the iPhone 18 Pro, which is a genuine technical win and a real content-per-phone increase. It does not appear to be a pricing win.
There's a second tell further down the statements. Operating profit of ₩245.8 billion produced only ₩97.9 billion of operating cash flow in the quarter, because inventories climbed to ₩2.49 trillion from ₩1.46 trillion a year earlier. That's a 70% inventory build against 40% revenue growth. Some of that gap is the normal pre-launch stocking a September iPhone forces on the supply chain, and Q3 and Q4 should absorb it. But the build is running ahead of sales, and a supplier who guesses high on an Apple ramp eats the difference.
Here's the sentence that made me want to write this. In FY2015 LG Innotek did ₩6.14 trillion of revenue at a 12.6% gross margin and a 3.6% operating margin. In FY2025 it did ₩21.90 trillion, 3.6 times as much, at an 8.0% gross margin and a 3.0% operating margin. The company more than tripled its top line over a decade and ended up less profitable on every won of it.
That's not a bad-execution story. Fixed-cost businesses are supposed to get more profitable as volume grows, and this one went the other way, which means the price per unit fell faster than the cost per unit did. That's what a single dominant customer looks like when you plot it over ten years.
The exception proves it. FY2021 and FY2022 were the good years, with operating margins of 8.5% and 6.5%, when sensor-shift stabilization and the multi-camera build-out gave LG Innotek something Apple couldn't easily second-source. Then the margin decayed straight back down: 4.0% in FY2023, 3.3% in FY2024, 3.0% in FY2025. Technical leads in this business are rented, not owned. The variable-aperture module is the newest lease, and I'd expect it to follow the same curve unless something structural changes.
Something structural is what management is proposing.
The bull case isn't cameras. It's package substrates, specifically FC-BGA for AI accelerators, where LG Innotek is developing ultra-large substrates with what it describes as big-tech partners and targets mass production of training and inference substrates starting next year. Management has pointed at roughly ₩1 trillion of operating profit from package solutions by 2031. KB Securities models FC-BGA revenue going from about ₩140 billion this year to ₩1.1 trillion in 2028 and ₩2.3 trillion by 2030. If any version of that lands, it changes the margin mix completely, because substrates are sold into a market with several buyers rather than one.
The company has said the right things about funding it too. It committed ₩600 billion at Gumi last year and has talked about spending more than ₩1 trillion this year to scale a new Vietnam plant, with additional FC-BGA lines at both sites under review.
Then you open the cash flow statement, and it hasn't happened. Purchases of property, plant and equipment were ₩189.3 billion in Q1 2026 and ₩134.9 billion in Q2, so ₩324.2 billion for the half. The comparable figure for the first half of 2025 was ₩310.0 billion. That's flat. Full-year capex has actually been falling hard: ₩1,799.2 billion in FY2023, then ₩879.0 billion, then ₩611.0 billion in FY2025. Property, plant and equipment on the balance sheet has shrunk from ₩4.86 trillion at end-2023 to ₩3.69 trillion at Q1 2026, which is what happens when depreciation outruns investment.
A company that intends to build a multi-trillion-won substrate business does not spend less on plant three years running. To hit a full-year figure above ₩1 trillion, the second half has to run more than double the first half's pace. Maybe it will. Substrate tools have long lead times and payments land lumpy, and Vietnam construction spending can arrive in a single quarter. But right now the plan lives in presentations and the harvest lives in the accounts, and only one of those is audited.
At ₩611,000 as of August 28, the market caps LG Innotek at ₩14.46 trillion on 23.67 million shares. Trailing four-quarter net income is about ₩663 billion, so roughly 21.8x earnings, against 2.27x the ₩6.37 trillion of book equity. For a business earning a 3% operating margin, that's expensive. Contract assemblers with this profit profile usually trade in the high single digits to low teens.
So the multiple is not paying for the camera business. It's paying for the substrate business that doesn't yet exist at scale. That's a defensible thing for a market to do, and it's also why the shares have been so violent. The page's 52-week range runs from ₩160,800 to ₩1,530,000, and the stock is currently about 60% below that high after a sell-off through the summer. Foreign investors have been net buyers into the decline. Nothing in the operating results explains a move of that size in either direction, which tells you the price is tracking belief about 2028 rather than earnings in 2026.
One more thing the multiple has to survive: LG Innotek loses a lot between operating profit and net profit. FY2025 operating income of ₩665.0 billion became pretax income of just ₩408.6 billion, with finance costs of ₩189.4 billion against ₩109.3 billion of finance income. Net income landed at ₩341.3 billion, about half the operating line. Any earnings-based multiple here is being applied to a number that a leveraged balance sheet has already taken a bite out of.
The strongest counterargument is that the margin history is the wrong history. LG Innotek's decade of thin returns was earned in a business it is now deliberately de-emphasizing, and judging the substrate opportunity by camera-module economics is a category error. If FC-BGA revenue reaches even the low end of what analysts model, the blended margin doesn't improve gradually, it steps up. The multiple would then look reasonable in hindsight.
The flat-capex point could also be a timing artifact rather than a signal. Substrate capacity is bought in large discrete blocks, and a company can spend nothing for two quarters and ₩700 billion in one.
And the camera side isn't dead. Content per iPhone is rising, the variable-aperture module is a first, and Q2's 40.5% revenue growth was not a fluke of accounting. Higher revenue at a fixed thin margin still produces more absolute profit, which is what the twenty-fold comparison actually captured.
The bear case doesn't need Apple to leave, though, which is the thing to keep in view. It only needs Apple to keep doing what it's done for ten years, while the substrate ramp slips a couple of years to the right.
The figure that settles this is capex, and it will be in the Q3 report. First-half purchases of property, plant and equipment came to ₩324.2 billion. If the third-quarter cash flow statement shows quarterly capex still running near ₩150 billion, the substrate ramp is not being funded on the schedule management has described, and the 21.8x multiple is resting on a plan that hasn't started. If it jumps toward ₩400 billion or more in the single quarter, the Vietnam and Gumi build is real and on time.
Two supporting checks in the same filing. Watch whether the optical solutions share of revenue falls below 80% as substrates grow, since that ratio is the cleanest measure of whether the mix is actually shifting. And watch inventories against the ₩2.49 trillion Q2 figure. Q3 is the iPhone build quarter, so that number should come down as product ships. If it doesn't, the build was a forecast error rather than a ramp.
The semiannual report for the period ended June 2026 was filed with DART on August 14, and the segment detail behind these figures sits there.
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.