Two Samsungs, one 2026
This cohort’s Samsung Foundry briefing covers a company still working to close a yield gap with TSMC, with profitability targeted only for 2027 [4]. Samsung’s memory division, covered here, is having an entirely different kind of year: aggressive, sustained price increases in a market where it holds real pricing power. Both are true simultaneously, inside one parent company, and conflating them into a single “how is Samsung doing” narrative flattens two genuinely different stories into one that describes neither accurately.
The pricing numbers
Samsung raised overall memory prices by up to 60% since September 2025, then raised DRAM contract prices specifically by roughly 60-70% in Q1 2026, followed by a further ~30% quarter-on-quarter increase in Q2 2026 [1]. Two consecutive quarters of double-digit-to-triple-digit percentage price increases is an extraordinary run for any commodity-adjacent product category, and it reflects genuine supply tightness rather than pure pricing opportunism — DRAM spot prices across the industry were reported at nearly triple their year-earlier level by the same period [3].
Where Samsung sits in HBM specifically
Within the three-company HBM market (SK hynix, Samsung, Micron), Samsung is generally covered as the second-largest supplier by share, behind SK hynix’s dominant position on Nvidia’s Rubin platform specifically [2]. That positioning matters for how a reader should weigh Samsung’s memory business against SK hynix’s: Samsung is a strong, profitable participant in the HBM supercycle without holding the single-largest-supplier position this cohort’s SK hynix briefing describes.
Why the split matters for anyone valuing Samsung as a stock
An investor or reader evaluating “Samsung” as a single entity risks averaging together a memory business with real, current pricing power and a foundry business still working toward profitability — producing a blended picture that overstates the foundry’s near-term prospects and understates how strong the memory division’s current position actually is. The more useful approach, and the one this cohort’s separate Samsung Foundry and Samsung Memory briefings are structured around, is to evaluate the two as what they actually are: a highly profitable, well- positioned memory business, and a foundry business making real technical progress but still proving its commercial case, sharing a parent company and very little else in terms of near-term trajectory.
| Samsung, split by business, 2026 | Position |
|---|---|
| Memory (DRAM/HBM) | Aggressive price increases; #2 HBM supplier behind SK hynix |
| Foundry (2nm logic) | 55-60% yield estimate; profitability targeted ~2027 |
| Shared parent | Samsung Electronics |
The honest takeaway
Samsung’s memory division is one of the clearest beneficiaries of the 2026 memory supercycle covered elsewhere in this cohort — genuinely strong, genuinely profitable, and rarely given credit as its own story because it sits inside a conglomerate whose foundry news tends to dominate AI-hardware coverage instead. Readers who want an accurate read on Samsung’s actual 2026 performance need to look at the two businesses separately, not average them into one number that represents neither.
The capital-allocation question this creates
A memory division generating this much pricing power, sitting inside the same conglomerate as a foundry division still years from profitability, creates a genuine internal capital-allocation question that outside observers rarely get visibility into. Every won of profit the memory division generates is a won that could, in principle, be redirected toward accelerating Samsung Foundry’s yield-improvement program and its Taylor, Texas ramp — or could instead be reinvested back into expanding memory capacity further, compounding the division’s own current advantage instead. Samsung’s public disclosures do not break out that allocation decision in enough detail for an outside reader to track precisely, but the scale of the memory division’s current pricing power means the decision is consequential regardless: it will meaningfully shape how quickly Samsung Foundry can close its gap with TSMC, independent of anything Samsung Foundry’s own engineering team does on its own.
A comparison worth holding in mind
SK hynix, by contrast, is a much more purely memory-focused company with no equivalent foundry business competing for the same capital. That structural difference means SK hynix’s memory profits flow overwhelmingly back into memory-specific capacity and technology, while Samsung’s equivalent profits are, at least potentially, in competition with an entirely different business unit’s capital needs. Whether that makes Samsung’s memory division a less aggressive long-term competitor than SK hynix, precisely because its profits have somewhere else to go inside the same company, is a genuinely open strategic question — and one this briefing raises rather than resolves, since the internal allocation data needed to answer it with confidence is not public.
