On the morning of July 2, at 10:40am, Samsung Electro-Mechanics Co., Ltd. (KRX:009150) told investors and reporters it would build a global manufacturing hub for high-performance package substrates in Sejong. The fair-disclosure filing puts the expected investment at about ₩8tn, roughly $5.4bn at the ₩1,475.60 rate the company itself used in a filing three weeks later. The venue was a national briefing on the Chungcheong region's advanced-industry development vision.
₩8tn is a big number in a headline. Divide it by the period the filing gives and it stops being one.
The schedule in the disclosure runs from January 1, 2026 to December 31, 2040. That's fifteen years. Eight trillion won across fifteen years averages ₩533bn a year, or about $361m.
Now set that against what this company already does. Purchases of property, plant and equipment totalled ₩12.54tn across the eleven fiscal years from FY2015 through FY2025, an average of ₩1.14tn a year. The plan's annual average is 47% of that. In FY2025 alone the company spent ₩1,192.1bn. In the first half of 2026 it spent ₩843.2bn, up 89.6% year on year and equal to 70.7% of the entire prior-year figure in six months.
The obvious defence is that ₩8tn covers one site and one product line, not the whole company, so comparing it to total capex is unfair. That's true and it doesn't rescue the number. Package substrates are one of three segments, and the substrate business is the one management says is being transformed by AI server demand. If the transformative segment's flagship site absorbs ₩533bn a year on average, the transformation is being funded at a pace that would not move the consolidated capex line much.
There's also a timing issue nobody can resolve from the filing. Fifteen-year plans are not spent evenly. Most of the money could land in 2027 through 2030, which would make the early years far heavier than the average and the late years a rounding error. The disclosure doesn't say. It gives a start date, an end date and a total, and explicitly notes that scale and schedule may change with market conditions.
Item 4 of the disclosure asks for the date of the board resolution. The answer is a dash.
That's not an irregularity. Korean fair-disclosure filings under this form are for guidance that a company gives to the market, and guidance doesn't require board approval the way a capital commitment does. The information provider is listed as the IR group inside the planning team. The audience is listed as domestic and overseas investors and media.
But it changes how the number should be read. This is not ₩8tn that directors have approved and committed. It's a figure the company put into a government-hosted regional development event, alongside a policy narrative about building advanced industry in Chungcheong. Korean groups routinely produce long-dated regional investment totals in that setting, and the totals are real intentions rather than signed cheques.
Contrast it with the glass core venture approved the same day. That one has a board resolution date of July 2, four outside directors recorded as attending, a specific share count, a specific price of ₩319.1bn, a payment date of September 1 and a put-call agreement. When Samsung Electro-Mechanics commits capital, the paperwork looks different.
The gap between the plan and the price is where this gets uncomfortable. At the August 27 close of ₩1,385,000 across 74,693,696 shares, the market capitalisation is ₩103.5tn, about $70bn.
First-half net income was ₩581.0bn. Annualise it and the shares trade near 89 times. On FY2025's actual ₩731.0bn it's 141 times. Book value at June 30 was ₩10.53tn, so the price is 9.8 times book. The best net income in the eleven years of history available here is ₩993.5bn in FY2022, and the best year for operating income was FY2021 at ₩1,486.9bn.
To reach even 25 times earnings at today's price, this company would need roughly ₩4.1tn of annual net profit. That's about 5.6 times FY2025 and 3.6 times the current annualised run rate. It is more than four times anything the business has ever produced.
That's not automatically wrong. Q2 2026 delivered a 12.7% operating margin on record revenue, and if AI-driven substrate and MLCC demand carries the company toward a structurally different margin the earnings could get there. The point is what the capital plan implies. A company that genuinely expects to quadruple its profit pool from substrates would not fund the flagship site at roughly half its historical annual capex. It would be building faster than it ever has.
The strongest counter is that substrate economics don't scale with concrete. Package substrate value comes from layer count, line width and yield, not from floor space, and the incremental capital needed to move from a mid-range organic substrate to a large-body AI accelerator substrate is a fraction of what a memory fab costs. On that reading ₩8tn is plenty, and the constraint is engineering rather than money.
The second counter is that the ₩8tn figure is a floor stated conservatively at a public event, and that actual spending will run well ahead of it. The first-half capex number supports this. ₩843.2bn in six months, up nearly 90% year on year, is not the behaviour of a company holding back. Whatever the 2040 plan says, the money is already moving faster than the average implies.
The third is that valuation is a statement about the market's view of AI packaging generally, and no components company in this cycle trades on trailing earnings. That's true. It's also the argument that has preceded every components de-rating in this company's history, including the one that took operating margin from 15.4% in FY2021 to 7.4% in FY2023 on higher revenue.
The FY2026 results in January, and specifically the capital expenditure the company guides for 2027. That number is board-approved and gets disclosed. If it comes in at or above the roughly ₩1.7tn pace the first half implies, the ₩8tn Sejong figure is a floor and the substrate build is happening at speed. If 2027 guidance drops back toward ₩1.2tn, the fifteen-year average is closer to the real plan, and the market's ambition and management's are different.
The second thing to look for is a follow-up disclosure with an actual board resolution attached to a Sejong site. Korean rules require a separate filing for a new facility investment once a board approves one. Until that exists, the ₩8tn is a stated intention with a 2040 end date and a disclaimer, and the only capital this company has formally committed to next-generation substrates is the ₩319.1bn that goes into a glass core venture on September 1.
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.