According to ZDNet Korea, reported on October 26, Samsung Electronics and SK Hynix are targeting the second quarter of 2027 to begin volume shipments of eight-layer HBM4 memory to Nvidia, as the AI chipmaker reassesses its memory strategy for next-generation Rubin Ultra accelerators. This is a plan/target and actual shipments have not yet occurred.
8-layer HBM4 memorySamsung Electronics, SK Hynix and 3 totalDigiTimes AsiaCapacity Change
According to ZDNet, Samsung and SK hynix are preparing to ship 8-Hi HBM4 for Rubin Ultra in 2Q27, departing from the original 12-Hi HBM4E plan; 8-Hi HBM4 was initially absent from NVIDIA's plans, but talks gained traction in 3Q26. The source cites thermal challenges and lower yields in 12-Hi stacks, potentially easing supply constraints. Although both Samsung and SK hynix have completed NVIDIA's qualification for 12-Hi HBM4, their 8-Hi products require separate certification; a source said NVIDIA is expected to deploy both 8-Hi HBM4 and 8-Hi HBM4E in Rubin Ultra-based NVL rack-scale systems depending on rack configuration. TrendForce says 8-Hi configurations use fewer DRAM dies, lowering cost per stack and GPU bill of materials, and given tight HBM supply, 8-Hi setups are expected to be a priority for evaluation by many GPU and ASIC vendors in 2027; ZDNet reports 12-Hi HBM4 remains mainstream in 2026, with 8-Hi adoption expected to rise sharply from mid-2027. Economy Tribune, citing Daishin Securities, suggests Anthropic's in-house AI chip could also feature 8-Hi HBM4, raising the prospect of Samsung supplying both foundry services and memory. Separately, per Wccftech, NVIDIA trimmed Vera Rubin's HBM4 bandwidth specifications: an archived September 1 version listed the Vera Rubin NVL72 system with 20.7 TB of HBM4 capacity and 1,580 TB/s bandwidth, the Vera Rubin Superchip at 576 GB and 44 TB/s, and each Rubin GPU at 288 GB and 22 TB/s; NVIDIA has since lowered per-GPU HBM4 bandwidth to 19.2 TB/s and NVLink 6 throughput from the originally advertised 3.6 TB/s to 3 TB/s, cuts of roughly 13% and 17% respectively, and up to 26% for HGX Rubin NVL8 systems, with changes dating back to September. IT Home notes NVIDIA revised the Vera CPU memory specification from a fixed 1.5 TB of LPDDR5X per chip to 'up to 1.5 TB', and that lowering bandwidth requirements could allow more HBM4 products to qualify for Rubin GPUs, potentially boosting overall production output.
HBM4HBM4ELPDDR5XNVIDIA, Samsung and 4 totalTrendForceCapacity Change
Since the launch of TSMC's N5 5 nm process in 2020, the size of the SRAM 6T bitcell has remained around 0.021 µm², with logic transistor density increasing but memory not seeing similar advancements. With the A14 node, the 6T SRAM bitcell is expected to shrink significantly to 0.017 µm², a 19% reduction, with production set for 2028. TSMC is currently conducting pilot production to assess the viability of the node, enabling customers to evaluate how their designs will perform in TSMC's production environment; the progress of A14 production is reported as promising, with timelines advancing due to early successes. If pilot production continues smoothly without major yield or production issues, risk production is set to commence in 2027, followed by high-volume manufacturing in 2028; if no significant challenges arise next year, TSMC may even accelerate the production schedule by a few months.
According to an RT NEWS item reposted by TrendForce, Amazon said that AWS large-scale compute capacity is fully booked for 2026, most of 2027 is sold out, and a considerable share of 2028 is committed. CEO Andy Jassy also said he does not expect Memory and power shortages to end in 2027. The source does not state the cause of the fully booked capacity or the persisting shortages; this is a statement of capacity/demand booking status, not a realized price change or confirmed production cut.
large-scale compute capacityMemorypowerAmazon, AWSTrendForce Official (Market Data)Capacity Change
At an event this week, Micron announced that its next-generation 3D NAND (G11), which it plans to enter mass production in 2H 2027, is set to offer three significant advantages: doubling storage density (approximately twice the bit density of G9, with 3D TLC going from 21 Gb/mm² to 42 Gb/mm²), increasing I/O performance by up to 50% versus the company's current-generation NAND memory (interface speed from 3,600 MT/s to 5,400 MT/s), and reducing data transfer energy consumption per bit by more than 25%. G11 will adopt wafer-on-wafer bonding, manufacturing the NAND array on a new 3D NAND node and the peripheral circuits and I/O on a logic process before bonding them together, and will transition to more than 400 NAND layers; Micron has not disclosed the exact number of active wordline layers or memory decks. Micron executive Scott DeBoer described G11 as the world's most advanced NAND, ramping in the second half of next year. Micron also expects to disclose more G11 technical details in 2027 and, beyond G11, is developing high-mobility materials and denser cell architectures to continue XYZ scaling; specific architectural details, density targets, or production schedules for post-G11 generations have not been disclosed. These are the company's stated plans and expectations; mass production has not yet occurred.
3D NANDG11 3D NANDMicronTom's HardwareCapacity Change
In an early Friday announcement dated 23 October 2026, SK hynix said it had decided to abandon plans for a potential initial public offering (IPO) of Solidigm, its U.S.-based enterprise SSD subsidiary, citing unfavorable stock market conditions and semiconductor share price trends. At the same time, SK hynix confirmed it plans to build a new fab for Solidigm in the U.S. to address growing demand for enterprise SSDs. SK hynix stated that although an IPO could have helped finance a new fab in the U.S., its subsidiary's expansion remains a priority; the company has discontinued its review of both pre-IPO financing and an IPO and will continue to evaluate alternative funding mechanisms. The source states that the planned fab's capacity, investment requirements, and construction schedule have not yet been disclosed. Solidigm specializes in high-capacity data center SSDs based on floating gate 3D QLC NAND memory, and its existing manufacturing capacity is barely sufficient to address its projected needs.
enterprise SSD3D QLC NANDSolidigm, SK hynixTom's HardwareCapacity Change
According to Counterpoint Research, YMTC, known primarily as a NAND Flash manufacturer, has been producing about 5,000 wafers per month of 16 Gb LPDDR5X DRAM running at 7,500 MT/s at its Fab 2 since October 2025. The company is now preparing to begin volume production of LPDDR5X memory with speeds of 9,600 MT/s at its Fab 3, using Vertical Channel Transistor (VCT) 4F² cell structure technology. The volume production remains in the preparation stage and is a planned event.
Germany's Merck plans to expand the use of molybdenum dioxide dichloride (MoO₂Cl₂) precursors from NAND flash memory to DRAM and logic semiconductors. Bing Han, senior vice president of the Thin Film Materials business at Merck Electronics, said at a press conference at Sono Felice Samsung in Gangnam on Oct. 21 that as device structures shrink, conductivity issues with copper and tungsten will drive greater adoption of molybdenum. Merck has established a molybdenum precursor production facility in Eumseong, North Chungcheong Province, which will be integrated with its CHEMKEEPER Vapor 200 materials delivery systems made at its Ansan plant and offered as a turnkey solution. The expansion uses land near an existing plant of M Chemicals, which Merck acquired in 2023; the site previously produced hafnium-based high-k dielectric precursors, construction began in 2024, and the facility was prepared for operation within two years. Han said Merck is planning production capacity sufficient to meet all demand from Korean customers, but the company did not disclose specific production capacity or investment figures, and it plans to eventually expand support beyond Korea to customers across Asia. As background: Samsung Electronics began using molybdenum in its 200-layer-class NAND in 2024, and SK hynix has replaced some word-line materials with molybdenum in its 300-layer-class NAND; Lam Research, Tokyo Electron and Applied Materials are among the suppliers of molybdenum deposition equipment. Hein said Merck is reviewing production capacity with customers in preparation for semiconductor materials consumption potentially doubling within the next five years.
Daeduck Electronics said in a regulatory filing on Oct. 22 that it will invest an additional 615 billion won in semiconductor production equipment and ancillary facilities to expand semiconductor substrate production capacity. The amount represents 68.54% of its consolidated equity at the end of 2025, the investment will run through Dec. 31, 2030, and it will be financed through existing funds and external borrowing. The announcement comes three months after the company unveiled a 497 billion won investment in July; the three new facility investments disclosed in 2026 total 1.325 trillion won. In May, the company unveiled a 213 billion won investment to establish a flip-chip chip-scale package (FC-CSP) production plant at its B1 Center in Siheung, Gyeonggi Province; in July, it announced another 497 billion won investment to expand production of flip-chip ball grid array (FC-BGA), FC-CSP and memory substrates, with both projects scheduled for completion by the end of 2027. In July, it also acquired Newflex's Ansan factory for 52.8 billion won, an acquisition separate from the disclosed investments in new facilities. Daeduck Electronics has been negotiating long-term agreements (LTAs) with several global technology companies for FC-BGA, FC-CSP and memory substrates, and is expanding capacity using advance payments received from customers. In its filing, the company said the investment is intended to expand production equipment and ancillary facilities to meet growing demand in the semiconductor market.
TrendForce notes that Micron plans to begin mass producing its 1-delta (1δ) nm process in 2027, which is set to succeed its current mainstream 1-gamma (1γ) nm process. Beyond this transition, Micron intends to continue scaling its 6F² architecture before moving directly to 3D DRAM, bypassing the intermediate 4F² architecture; in contrast, South Korean and Chinese DRAM manufacturers still plan to adopt 4F² architecture, highlighting a growing divergence in process technology roadmaps among major suppliers. Micron is also developing memory solutions optimized for AI solutions, including vHBM and vEIM, to compete with SK hynix's 3D HBM and Samsung's zHBM. In NAND flash, Micron is preparing to ramp up production of its G11 generation in 2H27; by incorporating wafer bonding and stacking more than 400 layers, G11 is expected to deliver twice the bit density of G9, alongside a targeted 1.5-fold improvement in I/O speed. TrendForce believes these advances will enable Micron to close the technology gap with Kioxia and YMTC—both early adopters of hybrid bonding architectures—and strengthen its competitive position as NAND performance becomes increasingly critical to an AI-driven market. All of these remain plans/expectations, not yet realized mass production.
1-delta nm DRAM processvHBMvEIMMicronTrendForce Official (Market Data)Capacity Change
Nvidia and Samsung Electro-Mechanics will jointly invest in dedicated ABF substrate (package substrate) production lines in South Korea under a multi-year agreement that makes Samsung Electro-Mechanics a supplier of package substrates for Nvidia's AI processors and networking switches. The new lines will be set up at Samsung Electro-Mechanics' sites in Sejong and Busan, with deliveries due to start in 2026. Neither company has disclosed the investment amount or how the cost will be split. The agreement covers substrates for Nvidia's compute chipsets and for high-speed networking switches that connect chips inside data centres. Samsung Electro-Mechanics expects its supply volume to Nvidia to grow at a double-digit annual rate, and by 2028 Nvidia is projected to take what the company calls a significant portion of its total annual package substrate capacity. According to the company, the partnership will give Nvidia a stable supply of substrates for mass production of its AI chips and help ease supply chain bottlenecks in the AI infrastructure market, while giving Samsung Electro-Mechanics entry into the high-end AI substrate market, where it says technical barriers are exceptionally high. The agreement comes shortly after Samsung Electro-Mechanics signed a KRW 360.5 billion supply deal for memory semiconductor package substrates with an unnamed global customer. The expansion is planned and not yet in production.
According to Maeil Business Newspaper, citing industry sources, Samsung Electro-Mechanics is reportedly set to receive a combined KRW 3.7 trillion in investment from Tesla and NVIDIA to expand advanced chip substrate production; the investment is a reported commitment. The company has signed a long-term flip-chip ball grid array (FC-BGA) substrate supply agreement with Tesla and is expanding dedicated production lines at its Sejong facility. The substrates are expected to be used in AI chips for Tesla's autonomous electric vehicles and Optimus humanoid robots under a five-year agreement running from 2029 through 2033. Tesla will contribute approximately KRW 1.1 trillion toward the expansion in exchange for guaranteed substrate supplies; the funding forms part of Samsung Electro-Mechanics' KRW 4.27 trillion Sejong investment plan disclosed in September. Construction of the new FC-BGA substrate production facilities at the Sejong plant is expected to be completed by the end of 2027, with equipment installation and trial operations scheduled for the first half of 2028, mass production in the second half, and deliveries to Tesla expected to begin in 2029. According to TechNews, NVIDIA is also expanding global substrate capacity through supplier partnerships and aims to increase its total IC substrate capacity to 2.5 times its current level over the next three years: in Japan, NVIDIA has committed billions of dollars in capacity reservation prepayments to Ibiden, whose Ono plant aims to triple production capacity, with additional output expected to begin ramping in the first quarter of 2028; in South Korea, Samsung Electro-Mechanics is expanding its Sejong and Busan facilities, with NVIDIA expected to account for a significant share of its annual package substrate capacity by 2028 — while NVIDIA did not disclose an investment amount, Maeil Business Newspaper reports the commitment is believed to be worth approximately KRW 2.6 trillion; NVIDIA is also supporting Taiwan's Unimicron in expanding its KF substrate capacity through a capacity reservation agreement, with additional output expected from the second quarter of 2027 and full capacity targeted by mid-2028. Separately, Samsung Electro-Mechanics disclosed a KRW 360.5 billion contract on October 19 to supply memory semiconductor package substrates to an undisclosed global customer throughout 2027, with the customer not identified citing business confidentiality, according to Chosun Biz. The broader expansion comes as high-end FC-BGA production remains concentrated among a handful of suppliers, including Samsung Electro-Mechanics, Japan's Ibiden, and Taiwan's Unimicron; industry estimates cited by Maeil Business Newspaper put Ibiden's market share at more than 70%, with Samsung Electro-Mechanics working to narrow the gap.
FC-BGA substratememory semiconductor package substrateKF substrateSamsung Electro-Mechanics, Tesla and 5 totalTrendForceCapacity Change
Elon Musk said on October 20, during a live broadcast of SpaceX's Q3 2026 earnings call on X, that ASML and other critical semiconductor equipment suppliers have expedited shipments to the Terafab semiconductor manufacturing facility, a joint megaproject between Tesla and SpaceX, to support his aggressive timeline. He said the company is targeting volume chip production at Terafab's primary facility in Gibbons Creek, Texas, by the third or fourth quarter of 2028, emphasizing that the timeline is highly aspirational and carries 'big error bars'; SpaceX plans to manufacture its first experimental test chip at its research fabrication facility in Austin, Texas, around Q2 2027. Terafab will use traditional extreme ultraviolet (EUV) lithography systems utilizing Intel's 14A manufacturing node; Intel officially joined the Terafab venture in April 2026 to co-develop and refactor silicon fabrication technology. According to unofficial sources, Terafab's first EUV tool was reportedly slated for delivery in mid-2027, and it remains unclear if this is the expedited timeline Musk referred to. Other major suppliers include Applied Materials, Lam Research, and KLA Corporation, as well as Japan's Tokyo Electron; Tesla and SpaceX are also in talks with about 10 South Korean equipment makers, including Hanmi Semiconductor (advanced packaging equipment) and HPSP (high-pressure hydrogen annealing equipment). Musk also outlined AI compute goals: 2.5 to 3 GW of computing capacity active by the end of 2026, a total of 13 GW by the end of 2027, and plans to bring a 440,000-GPU Nvidia 'Vera Rubin' training cluster online by the end of Q2 2027. The event remains a plan/target and volume production has not been achieved.
EUV lithography systemslogic chipsmemoryASML, Tesla and 11 totalTom's HardwareCapacity Change
Japan's Fujifilm is expanding its extreme ultraviolet (EUV) photoresist production capacity to deepen cooperation with Samsung Electronics. Brian O'Donnellley, president of Fujifilm Electronic Materials USA, said the company is producing EUV photoresists in Pyeongtaek and, although it has not yet entered full-scale mass production, some EUV PR products have passed qualification tests at major customers. Fujifilm has expanded EUV PR production facilities in Shizuoka, Japan (its primary hub for advanced photoresist development and production) and Pyeongtaek, South Korea (near Samsung, selected for its most advanced products); EUV PR production and inspection equipment at both locations began operating in October 2025. The source does not state when full-scale mass production in Pyeongtaek will begin. The company cited artificial intelligence (AI) driving demand for materials used in processes at 2 nanometers and below, and expansion of advanced manufacturing processes increasing demand for ArF and EUV photoresists as the stated reason. Separately, Fujifilm is expanding its CMP slurry production facility near Samsung's Cheonan campus, planning to increase production capacity by approximately 30%, with additional facilities scheduled to begin operating in the first half of 2027; Fujifilm holds the largest global market share in CMP slurries for copper interconnects. The company also operates a facility in Carrollton, Texas near Samsung's Taylor plant and is considering various options to expand its manufacturing presence in South Korea, where it currently has two production facilities.
Samsung Electronics and SK Hynix are upgrading their NAND flash factories in China to make chips with more than 200 layers, Reuters reported on October 21, citing four people familiar with the plans. According to the report, the move would let the South Korean memory makers win back ground lost to Yangtze Memory Technologies Co (YMTC), even though US curbs on upgrading their China operations remain formally in place. This is a reported plan under way; no completion timeline is given and the US restrictions have not been lifted.
NAND flashSamsung Electronics, SK Hynix and 3 totalDigiTimes AsiaCapacity Change
Related manufacturers
Samsung Electronics, SK Hynix, Yangtze Memory Technologies Co (YMTC)
Chinese DRAM maker CXMT's next DRAM milestone is a plan to achieve 4F² server memory using hybrid bonding by end-2026. At a recent shareholder meeting, management pledged to enter what it called the 'uncharted territory' of extreme memory scaling, and stated it would use process innovation to overcome technical barriers. The milestone remains a stated plan; the source does not state any progress beyond the roadmap, the cause, or market impact.
DRAM4F² server memoryCXMTDigiTimes AsiaCapacity Change
According to a repost from TrendForce, Samsung has reportedly begun producing its 2nm Exynos 2700, with output potentially rising over 10% from the Exynos 2600 (no exact date given in the source). The chip is expected to be used in the Galaxy S27's standard and Plus models. This is a reported claim, and the output increase is a potential expectation rather than an achieved result; the report also raises the question of whether the chip could help revive Samsung's non-memory business.
2nm Exynos 2700SamsungTrendForce Official (Market Data)Capacity Change
According to Reuters, as AI-driven demand extends from DRAM to NAND and competition from China's YMTC intensifies, Samsung and SK hynix are planning upgrades at their Chinese NAND facilities: Samsung plans to shift its Xi'an plant from roughly 200-layer to 286-layer V9 NAND under an internal project dubbed "Monami"; SK hynix, already authorized to produce up to 240-layer NAND in Dalian, is reportedly planning to upgrade from its current 192-layer technology to 240 layers or more, and may seek U.S. approval for R&D on 300-plus-layer technology. TrendForce estimates China accounts for roughly 30–35% of Samsung's total NAND output and an even higher 40–45% for SK hynix. U.S. export restrictions remain a major hurdle: both chipmakers must obtain annual licenses to install U.S.-made equipment at their Chinese plants, and Washington indicated in 2025 that it would not approve technology upgrades or capacity expansions there; both companies confirmed to Reuters that upgrades were underway in close coordination with Washington but declined to disclose detailed production plans. On competition, YMTC's NAND lineup is dominated by 160- and 232-layer products, it is already producing 294-layer chips in limited volumes and plans to push beyond 300 layers. Separately, per Seoul Economic Daily, Samsung has shifted roughly 60% of its V-NAND capacity to V9 (286 layers), is ramping shipments to NVIDIA, has begun mass production and supply of V10 NAND, and is accelerating development of next-generation V11 targeting around 500 layers; per Bloter, SK hynix is stepping up its shift to 321-layer NAND, which became its largest NAND production segment in 2Q26, with roughly half of its domestic NAND capacity expected to transition to 321-layer products by year-end.
NAND flashSamsung, SK hynix and 3 totalTrendForceCapacity Change
According to Reuters, Elon Musk said during a SpaceX conference call that his semiconductor manufacturing venture, Terafab, is targeting volume production in 3Q28, though he acknowledged that either 3Q or 4Q remains possible, describing the timeline as "very aggressive" given that advanced logic chip fabs in the U.S. typically take three years or more to build. Before volume production, Terafab aims to produce its first functional test chip using SRAM circuits around May or June 2027 at its Austin research facility. Equipment supply remains a key challenge, particularly from ASML and U.S. suppliers such as Applied Materials and KLA; asked whether SpaceX had ordered EUV lithography systems from ASML, Musk did not answer directly but thanked ASML and other suppliers for accelerating equipment deliveries. Per Reuters, ASML expects demand to outstrip supply through 2028 despite plans to expand EUV production capacity. According to The Elec, Tesla and SpaceX are in talks with South Korean semiconductor materials, components, and equipment suppliers to establish Terafab's supply chain, with several companies signing NDAs; HPSP and Hanmi Semiconductor remain the only two Korean equipment suppliers publicly linked to Terafab, and Terafab is reportedly seeking prices even lower than those offered by Chinese competitors. On manufacturing partners, TSMC Chairman and CEO C.C. Wei confirmed ongoing discussions but stressed that any fab involving TSMC must be operated by the company, contrasting with Musk's October 7 remarks that his companies would build and operate Terafab, potentially allowing TSMC to sublease space without an operational role; meanwhile, per Reuters citing Bloomberg, Intel CEO Lip-Bu Tan said the company would remain involved in Terafab.
SRAM test chipadvanced logic chipsEUV lithography systemsTerafab, Tesla and 10 totalTrendForceCapacity Change
According to Chosun Biz, Samsung Electronics reportedly began wafer production for Exynos 2700 earlier this month, with output potentially increasing by more than 10% compared with Exynos 2600, in a bid to revive its non-memory semiconductor business. Exynos 2700 is expected to power the Galaxy S27 series in 2027, primarily in the standard and Plus models, with possible adoption in a new Pro variant; however, the flagship Ultra model is increasingly likely to rely exclusively on Qualcomm's Snapdragon processor, and the final chip configurations remain undecided. Exynos 2700 reportedly adopts Samsung Foundry's second-generation 2nm process, SF2P, which targets 12% higher performance, 25% lower power consumption and an 8% reduction in chip area compared with the first-generation SF2 used for Exynos 2600, and it is reportedly the first chip to enter mass production using SF2P; per The Bell, Samsung has shared production plans with key partners, although initial wafer starts remain limited. The report notes that the extent of the non-memory recovery will depend on Exynos 2700's mass-production yields and adoption across the Galaxy S27 lineup; Samsung Foundry's 4nm lines are reportedly operating at nearly full capacity while it continues seeking additional customers for its 5nm process. Additionally, per TrendForce, mobile DRAM prices are expected to continue rising in 4Q26, albeit at a slower pace, as suppliers shift capacity toward higher-margin server products, making the price competitiveness of Samsung's in-house processors increasingly important.