If New Clean Rooms Stay Empty Into 2028, Where Does Cash Stick?

ANALYSIS · MEMBERS CONTINUE · 5 min free / 14 min full · Data as at 30 Oct 2026

On 29 October 2026 Samsung Electronics printed ₩107.4 trillion of operating profit — the first ₩100 trillion quarter any tech company has reported — and then told the same room that next year's spending will not lift 2028 bits much. It holds ₩197.6 trillion of net cash. DRAM prices rose about 20% while DRAM bits grew only low-to-mid single digits. 2027 HBM is sold out. The rooms that would add wafers are still shells. This map asks who is paid if new clean rooms stay empty into 2028.

In short

• The question: if new clean rooms stay empty into 2028, where does the cash stick?
• The answer: Scenario B. The empty room, not the cheque, is the stuck step. Builders and toolmakers are paid on shells and deposits. Samsung is paid 55.0% on the bits that already shipped. Shareholders are in line for ₩90–110 trillion this year — 0.46–0.56× net cash (ours). 2028 bits wait on rooms that ₩197.6 trillion cannot hurry
• Why: today's Korea Pulse gave the ₩107.4 trillion quarter and the clean-room bind. Today's Stack shows the inverse: a $6.5 billion hall signed with no construction money. This morning Tokyo Electron billed ¥935.0 billion in the July–September quarter, 20.4% of it Korea (ours). This map is the empty room, not a stock
• The size of the gate: net cash is 2.34× this year's ₩84.4 trillion plant budget (ours). Only ₩9.6 trillion, 33%, of the ₩29 trillion of third-quarter capex showed up as extra plant on the books (ours). P5 still targets 2028. We do not invent a wafer-start
• What would break it: a named P5 wafer-start that actually adds 2028 bits (A), or tool lead times that push first wafers past 2028 while the shells sit (C)
• Nothing here is a recommendation. Arithmetic shown below the wall

The numbers that matter

Sources: Samsung newsroom and 3Q 2026 presentation, 29 Oct 2026; call as transcribed; Tokyo Electron tanshin and presentation, 30 Oct 2026; ZDNet Korea on P5. Ratios and dollars at ₩1,334.3 per dollar (Yonhap, 15:30 KST, 30 Oct) are ours.

Why is the empty clean room the stuck step?

An AI accelerator does not leave without high-bandwidth memory beside the logic die. HBM, explained is the free page for the stack. Korea still invoices a large share of those stacks. Yesterday Samsung told the room that 2027 industry spending will rise sharply and that this will not mean many more bits in 2028: the extra money buys space first, construction costs more than it used to, the lead time from fab to wafer is long, tools cost more and take longer, and bigger HBM dies with more metal layers shrink the bits each wafer makes. Then it stated the plan: secure clean rooms in advance, and stay flexible on when to put tools in. Korean trade press calls that shell first.

That is a different question from 28 October, when more than half of 2027–28 memory was already sold mapped SK hynix's LTA book, not an empty bay. It is not what Micron's price assumes about the clean-room stuck step. It is not what Samsung's price assumes about the wafer pool. It is the inverse of today's TensorWave hall: there a $6.5 billion lease is signed and the construction money is missing; here ₩197.6 trillion is on the balance sheet and the 2028 bit is still missing. A sold-out 2027 HBM book is a claim on this year's rooms. It is not a 2028 cube.

Three clocks around it are not this bind. The P&L already cleared: ₩107.4 trillion on the bits that shipped, 99.2% from chips (ours). The cheque followed that margin — ₩30 trillion of dividend, ₩12.5 trillion of buybacks, ₩90–110 trillion for the year. The coater-track at Tokyo Electron is already an invoice: ¥935.0 billion of sales, Korea ¥190.7 billion. None of those three turns an empty 2028 bay into a wafer. What is left is a room: paid for as a shell, not yet a line.

What does the map already assume?

Three answers compete: P5 and the rest of the ₩84.4 trillion convert into 2028 bits inside the year (A); cash sticks with builders, toolmakers and the existing-fab invoice, while 2028 bits wait on rooms (B); or tool lead times and HBM die size push first wafers past 2028 while the shells sit (C). A tip sheet would pick the stock. Analysis asks which answer yesterday's ₩197.6 trillion already needs.

Our read, in one line: the map already assumes Scenario B. Net cash is 2.34× this year's plant budget. Only a third of the third-quarter capex is extra plant on the books. Tokyo Electron is paid on tracks that already shipped; Samsung is paid ₩55 of each ₩100 (company margin) on bits that already shipped. P5 is still a 2028 target. This is not a company price. The scenarios and the arithmetic are below.

The free preview ends here.

MEMBERS CONTINUE HERE — who gets paid · three scenarios · takeaways