Samsung Electronics Co., Ltd. (KRX:005930) produced ₩145.4tn of operating cash in the six months to June, against ₩33.9tn in the same half of 2025. That number isn't the interesting one. It is the mechanical consequence of memory pricing and it was legible the moment the Q2 profit figure landed. The interesting part of the half-year report filed to DART on 13 August is what happened to the money afterwards, and it doesn't match the way this cycle is being narrated.
Purchases of property, plant and equipment came to ₩31.2tn (~$22bn at roughly ₩1,400/$, an approximation I'll come back to). Purchases of long-term financial instruments came to ₩32.4tn, against ₩0.6tn in the first half of 2025. Short-term financial instruments rose a further ₩26.3tn on a net basis, where a year earlier they had run down by ₩3.2tn. Net of ₩1.1tn of long-term disposals, something close to ₩57.5tn went into deposits and securities over six months. That is about ₩1.84 into financial assets for every ₩1.00 into plant.
The claim here is narrow and I want to state it plainly: Samsung's own capital allocation says it is treating 2026 as a price event, not a volume event. That is a more candid statement about the durability of this cycle than anything in the company's public guidance.
One boundary. kstock covered the ₩90–110tn shareholder-return guidance and the ₩15tn buyback separately, and this is not that piece. This is about the cash that was neither returned nor spent on capacity.
Operating cash flow rose 328% year on year. Property and equipment purchases rose 16%, from ₩26.8tn to ₩31.2tn.
Annualise the half and you get roughly ₩62tn of capex for 2026, against ₩47.5tn in FY2025, ₩51.4tn in FY2024 and ₩57.6tn in FY2023. So spending is up in absolute terms. It is up less than it was three years ago, in a year when the company is earning something like eight times what it earned in 2023.
The ratio is where it becomes obvious. In FY2025 Samsung put 56% of its operating cash flow into plant. In the first half of 2026 it put 21%. A memory maker that genuinely believed AI demand had reset the long-run level of the market would not be sitting at 21%. It would be pouring concrete, because in this industry the constraint is wafer starts and the lead time on a fab shell is measured in years, not quarters.
There is a benign reading and I don't want to skip past it. Fab construction cash lands late. Shells get built, tools get ordered, and the payments cluster when the tools arrive. Samsung has restarted work on the second Pyeongtaek complex, and that line targets 2028. On that reading the ₩57.5tn parked in instruments is simply pre-funding for a spending wave that hasn't hit the cash flow statement yet, and the 21% ratio is a timing artefact rather than a judgment.
That reading is plausible. It is also testable in about eleven weeks, which is the useful thing about it.
Inventories went from ₩52.6tn at the end of December to ₩71.4tn at the end of June. That's a 36% rise, ₩18.8tn of it, in a half where revenue nearly doubled to ₩305.4tn.
Working capital absorbed ₩42.4tn of cash over the six months. In the first half of 2025 the same line absorbed ₩3.3tn. The difference is larger than the entire capex programme.
Inventory building during a shortage is not automatically a warning. Two innocent explanations are available. HBM has long cycle times and a lot of value sits in process, so a stack that ships in October is on the June balance sheet at a high carrying value. And when average selling prices roughly double, the same physical wafers carry a much larger won figure without anything changing in the warehouse.
Both are probably true to some degree. Neither is verifiable from the half-year report, because Samsung doesn't break inventory into raw materials, work in process and finished goods in the summary statements, and doesn't publish unit volumes.
So the honest position is this. A 36% inventory build is the single line in the statement most capable of embarrassing the current earnings run, and there isn't enough disclosure to say whether it should. If the build is finished goods sitting in a channel that has stopped absorbing price increases, the reversal shows up as a margin event two quarters later. If it's HBM in process, it shows up as revenue. The filing doesn't say which.
Total assets went from ₩566.9tn to ₩759.5tn. Equity went from ₩436.3tn to ₩579.3tn. Cash went from ₩57.9tn to ₩92.9tn.
Adding ₩192.6tn of assets in half a year is what happens to a commodity producer at the top of a price cycle, and it has an awkward consequence for anyone valuing the company on book. The ₩579.3tn of equity now carries a very large slug of retained cycle profit. At ₩261,500 a share and a market value of ₩1,528.8tn, that's 2.64x book. On trailing earnings — FY2025 net income of ₩45.2tn, less the ₩13.3tn earned in the first half of 2025, plus the ₩118.8tn earned in the first half of 2026, so ₩150.7tn — it's about 10.1x. Annualise the half instead and you get roughly 6.4x.
A market doesn't put single-digit multiples on a business it expects to keep earning at this rate. The multiple is the market saying the denominator is temporary. Samsung's capital allocation, read on its own, says roughly the same thing.
That's the point of this piece. The two are agreeing. The disagreement is with the commentary.
The strongest objection is the timing one above, and it's a good objection. If Q3 capex jumps to ₩45tn or ₩50tn, the 21% ratio was a construction-schedule artefact and this reading was premature.
A second objection is that parking cash in instruments ahead of a very large distribution is ordinary treasury behaviour, not a statement about the cycle. Samsung paid ₩6.2tn of dividends and spent ₩13.2tn on treasury shares in the half, and has guided to far larger returns for the full year. Cash earmarked for that has to sit somewhere. This is fair, and it probably accounts for a meaningful share of the ₩57.5tn. It doesn't obviously account for the ₩32.4tn specifically in long-term instruments, which is a strange place to warehouse money you intend to distribute within months. I can't size the split from what's disclosed.
A third: capacity may be constrained by things money can't fix quickly. Power and water for Korean fabs are being handled at the level of government policy right now, which isn't a sign of abundant headroom. If Samsung can't get grid capacity, low capex isn't a choice at all.
Three specific things, in order of how much they would move the reading.
The Q3 report's property and equipment line. If nine-month capex comes in below about ₩50tn, the annualised run rate is still under ₩70tn and the timing defence weakens considerably.
The inventory line at 30 September. A second consecutive 30%-plus rise, with revenue growth decelerating alongside it, is a different picture from the one available today.
And the long-term financial instruments balance. If it keeps climbing while capex doesn't, the company is telling you, in the only language a balance sheet has, how long it thinks this lasts.