Samsung’s 60–70% LTA target is a mortgage on 2030 capacity

What the 10 + 10 actually is

There is more detail behind the line than the headline. Responding to a question from Meritz Securities, Samsung said it had signed long-term DRAM supply agreements with 10 major customers and was negotiating with 10 more, some nearing completion, and that demand is spreading beyond hyperscalers, server OEMs and SoC vendors into mobile, automotive and networking (@antfeedapp, live transcription). That broadening is the underrated part: it means the shortage is being hedged by device makers, not only by five cloud companies, and it makes the 60–70% number a portfolio rather than a single negotiation.

Every memory maker counts differently

The three companies are all “mostly sold out” and none of the percentages are comparable (@MojoTricks): SK hynix has agreements covering more than 50% of 2027 and 2028 capacity with 14 customers, typically on five-year terms; Samsung targets 60–70% of production capacity through 2030; Micron frames coverage as more than 35% of revenue through 2030. Different denominators, different years, different definitions of “covered.” Quoting the three numbers side by side as if they measured the same thing is how a squeeze gets overstated.

Micron is also the only one of the three that publishes the uncomfortable details. Its 26 take-or-pay agreements carry about $150B of remaining obligations counted at minimum volume and minimum price, with roughly $32B of customer commitments, mostly cash (@Schulz_Research). For contrast, Amkor’s 10-Q lists about $3.2B of customer advance payments for packaging through 2027 — packaging buyers are prepaying too.

Volume is locked. Price is not.

This is the sentence that keeps the arrangement from being a cartel: “the price inside those still moves with the market within an agreed range, so they lock in the volume more than the margin” (@tengyanAI). Micron’s version has explicit floors; SK hynix’s has bands and deposits; Samsung has not quantified either. If spot DRAM falls, contract prices drift toward the floor, and the supplier keeps the volume while giving up the upside. That is a good outcome for a buyer and a fine outcome for a supplier that has already funded the fab.

The disclosure gap is real and worth watching. As @notanotherquant points out, Micron tells investors what the minimum-priced deals are worth, that customers pay even if they do not take the memory, how much cash has already arrived and when it is paid back. SK hynix said it has a price range and deposits, and not much else; its full Q3 report is due by 16 November. Posts in the same thread claim Samsung has already collected roughly a quarter of the deposits agreed under its contracts, and @JungleReportFi corrected an earlier, looser version of the same claim; we found no filing that confirms the collected share, so treat it as unverified.

Why a buyer signs at the top of a cycle

Because the cost of being short is now measured in schedule, not in dollars. A GPU rack that ships a quarter late misses a rental window that was priced before the memory bill went up. Deposits are exactly how a supplier separates real demand from panic demand: “Deposits put a cost on over-asking, which is why they are the number to watch,” as Schulz Research puts it. And as @TechBizBrief notes, coverage is a demand anchor, not revenue already earned — the commercial test is delivering the promised mix at a profitable cost as new capacity ramps.

Where it breaks

The bear case has dates, and none of them are in 2027: Micron’s ID1 starts wafer output mid-2027, with its Hiroshima and Singapore expansions and a second Idaho fab following in 2H28; SK hynix opens Yongin Phase 1 cleanrooms in early 2027; Samsung has said P5 runs at full scale from 2028; CXMT is reported to be aiming at 500k wafers a month by end-2028, mostly in commodity DRAM. TrendForce already has NAND loosening in 2H27. If something cracks, it cracks in commodity DRAM and NAND first, well before HBM — and it cracks where the contracts are thinnest and China is not a party to any of them (@Schulz_Research).

What we could not verify

Samsung has not published the dollar value of its agreements, the size of its deposits, or the price mechanism inside them; the 60–70% figure is a target described as “increasingly achievable,” not a signed number. Micron’s and SK hynix’s figures come from their own disclosures and calls, not from filings we could read line by line.

Sources

• @jukan05, 29 Oct 2026 01:58 UTC — Samsung: LTAs with 10 customers, 10 in negotiation, 60-70% through 2030
• @antfeedapp — live transcript: Samsung response to Meritz on LTA demand widening
• @MojoTricks — the three coverage numbers and their different denominators
• @Schulz_Research — Micron take-or-pay terms, Amkor prepayments, the dated bear case
• @tengyanAI — LTAs lock volume more than margin
• @notanotherquant — what Micron discloses and SK hynix does not
• @JungleReportFi — correction: strategic agreements are take-or-pay, ~$32B deposits behind them
• @TechBizBrief — coverage is a demand anchor, not revenue earned
• Samsung Electronics — Q3 2026 results
• SK hynix — Q3 2026 business results

FAQ

Does an LTA guarantee Samsung a price?

No. The common structure is a volume commitment with prices floating inside an agreed band, sometimes with floors. Volume is what is being bought.

Are the customers prepaying?

Yes, partly. Micron reports about $32B of commitments, mostly cash deposits; Samsung says it has collected roughly a quarter of the deposits agreed. That money funds capex before the bits ship.

Is 60–70% a lot?

For a business that historically sold mostly on quarterly negotiation, yes. It converts a spot market into a book of orders, which is why the equity story is being repriced around durability.

What should I watch next?

SK hynix’s full Q3 report by 16 November, Samsung’s January allocation of its return plan, and any contract reopening or deposit refund. Those are the first places a change of regime would show.