Invest1 distinct publisher3 min readUpdated
Jun Young-hyun credits only about 30 percent of Samsung's record quarter to Samsung, and treats first-to-market HBM4 as a starting line. That is a pricing warning, not modesty.
The Investor · Invest desk

Compiled by The InvestorSomething wrong?How this is made
Jun Young-hyun, vice chairman and head of Samsung Electronics' Device Solutions division [1], used a recent management meeting to tell his chip executives not to grow complacent about the company's recovery in high-bandwidth memory, calling complacency the biggest risk facing the business, according to industry sources cited by Seoul Economic Daily [2]. He was sharpest, those sources say, about colleagues who treated catching SK hynix as the win, reportedly asking: "Since when did Samsung work toward the goal of beating SK hynix?" [3]
The framing matters because of where Samsung started. Through the first half of 2025, SK hynix held 62 percent of global HBM shipments, Micron 21 percent and Samsung 17 percent [4], which put the self-described memory king third in a three-horse race, 45 points behind the leader and four points behind Micron [5], after ceding HBM3 and HBM3E to SK hynix [6]. Jun returned to Samsung in early 2024 and took over DS that May with "restoring fundamental competitiveness" as the stated priority, arguing that leadership cannot be reclaimed by following competitors' technology [7]. Samsung then became the first in the industry to commercialize HBM4, shipping commercial parts to customers in February at speeds up to 13 gigabits per second [8].
Recovering roughly 45 points of share, however, restores a position rather than defending one. SK hynix said in June that it had already sampled 12-layer HBM4E to major customers at up to 16 Gbps [9], about 23 percent faster than Samsung's shipping HBM4, albeit in the following half-generation [11]. Micron is mass-producing 12-layer HBM4 for Nvidia's Vera Rubin platform, sampling 16-layer HBM4, and working with customers on customized HBM4E with mass production expanding in 2027 [10]. Three vendors qualified or sampling inside the same generation window is the condition under which HBM stops being a scarce input and starts being a negotiated one. A person familiar with Samsung says Jun's message is that one good product is not a recovery, and that the gap has to be built again at HBM5, 6 and 7 [21].
His arithmetic on the quarter says the same thing more bluntly. Samsung posted record revenue of 171.5 trillion won and operating profit of 89.5 trillion won, with DS contributing 127.5 trillion won of revenue and 89.2 trillion won of operating profit [12][13]. That is 99.7 percent of group operating profit from one division on 74.3 percent of revenue, leaving about 0.3 trillion won for everything else Samsung does [14], at a DS operating margin near 70 percent against 52.2 percent for the group [15]. Jun told executives that "looked at coldly, our own capabilities account for only about 30 percent of the current results," attributing the rest to the AI-driven super-cycle, memory shortage and rising prices [16]. On his own split, roughly 26.9 trillion won of that operating profit is Samsung's work and roughly 62.6 trillion won is the cycle's [17]. Samsung's own earnings commentary pointed to industry-wide price increases on server memory demand [18], and Counterpoint Research concluded that surging memory prices were decisive in the result [19].
Watch the qualification calendar rather than the speed sheets: whether Samsung's 13 Gbps HBM4 holds price once 16 Gbps HBM4E parts are qualified at the same customers [8][9], and whether that near-70 percent DS margin survives a second and third qualified supplier [15][10]. Jun made a similar warning late last year when optimism was spreading [20]; the tell will be whether Samsung's next capacity and pricing decisions look like a company that believes him.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Jun Young-hyun is vice chairman and head of Samsung Electronics' Device Solutions (DS) Division.
Through the first half of 2025, SK hynix held a 62% share of global HBM shipments, Micron ranked third at 21%, and Samsung trailed at 17%.
Samsung had ceded leadership to SK hynix in the HBM3 and HBM3E markets, denting its standing as the "memory king".
Jun returned to Samsung as head of the Future Business Planning Office in early 2024 and took over the DS Division that May, setting "restoring fundamental competitiveness" as the top priority and arguing that "you cannot reclaim market leadership by following the technology and products of competitors."
Samsung began mass production of HBM4 and started shipping commercial products to customers in February, becoming the first in the industry to commercialize the chips; its HBM4 delivers speeds of up to 13 gigabits per second.
SK hynix said in June that it had supplied major customers with samples of 12-layer HBM4E chips running at up to 16 Gbps.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One outlet, hard numbers around anonymous quotes
The quantified spine of the story - H1 2025 HBM shares, HBM4/HBM4E specifications and layer counts, group and DS revenue and profit - is specific and internally consistent, and two attributions (Samsung's earnings announcement, Counterpoint Research) are named. But the narrative core, every Jun remark including the 30/70 attribution, rests on unnamed 'industry sources' and a 'person familiar with Samsung' in a single publisher with no on-record Samsung confirmation and no corroborating outlet in the cluster.
HBM4 generation is shipping, Samsung's share is not yet
Adoption is concrete at the product level: Samsung in commercial HBM4 shipment since February, Micron in 12-layer HBM4 mass production for a named Nvidia platform, SK hynix sampling HBM4E to major customers. Adoption of Samsung specifically remains thin on the only quantified measure in the cluster - 17% of H1 2025 HBM shipments, last of three - and no customer names, volumes or qualification wins are disclosed for Samsung's HBM4.
Publisher framing runs ahead of the numbers it prints
Mildly overstated overall, and the overstatement is the publisher's rather than the executive's. The piece opens by saying Samsung 'overtook rivals in the HBM market' with HBM4, but the only share data it supplies has Samsung last at 17%, and the rival milestones it cites are a faster HBM4E sample and a competitor already in volume HBM4 production for Nvidia. Pulling the gap back toward zero, the featured executive voice is deflationary - crediting ~30% of a record quarter to Samsung's own capability - and Samsung's own disclosure plus Counterpoint both attribute the results to memory pricing rather than share gains.
Company-adjacent anonymous briefing with an expectations-setting function
The reporting channel is visibly interested: unnamed 'industry sources' and a 'person familiar with Samsung' relay internal remarks to a Korean business outlet, and the message itself serves an expectations-management purpose - crediting most of a record quarter to a cycle lowers the bar for future quarters while framing HBM4 leadership as a work in progress. Counterpoint Research is a commercial research firm cited approvingly for the same cycle interpretation. The sources do not disclose who briefed the outlet or why, so the direction of interest is inferable but not stated.
Low-moderate: single publisher, anonymous core, undated milestones
Confidence is limited by structure rather than implausibility. One publisher supplies every fact; the decisive quotes are anonymous and second-hand; the milestone timings ('in February', 'in June', 'late last year') carry no years; and the reported quarterly figures are unusually large with no second source in the cluster to check them against. The quantified share and product data are consistent and specific enough to support the competitive reading with moderate confidence.
invest
Samsung's $72B payout and SK Hynix's $28.7B buyback bet the memory cycle holds1 distinct publisher
invest
Korea's leverage curbs moved the chip bet into margin loans rather than out of it1 distinct publisher
invest
SK's chairman warns of worst-ever memory imbalance while a U.S. fab stays under review1 distinct publisher
invest
Binance lists CXMT perps days after an $8.6B IPO, and the venue line blurs further1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
en.sedaily.com
1 article · August 15, 2026