According to a report reposted by TrendForce, Tesla is reportedly putting KRW 1.1 trillion on the table to lock in Samsung Electro-Mechanics' FC-BGA substrate supply through 2033; NVIDIA is also said to be grabbing capacity. The item is a reposted report with 'reportedly'/'is said to be' wording, and whether the deal is actually implemented is unconfirmed. The source does not state the cause.
FC-BGA substrate (AI chip substrate)Tesla, Samsung Electro-Mechanics and 3 totalTrendForce Official (Market Data)Partnership/M&A
According to a news item reposted by TrendForce, NVIDIA-linked Groq is reportedly exploring 3D-stacked DRAM for its next-generation AI chips, potentially moving beyond SRAM as AI models grow larger. This is only a reportedly explored direction, i.e., at the rumor/planning stage, with no confirmed design change, production or product implementation; the report also raises the question of whether this could reshape AI inference chip design. The source does not state any adoption timetable or confirmatory details beyond the reported exploration.
3D-stacked DRAMSRAMGroq, NVIDIATrendForce Official (Market Data)
A background article from Semiconductor Engineering states that chip scaling is moving beyond the transistor. It notes that at SEMICON West 2026, advanced packaging, photonics, test, and AI are increasingly intertwined, with very different conversations ahead. The article provides no specific manufacturers, products, price data, or stated cause.
As reported by ChosunBiz and ZDNet Korea on October 20, BOS Semiconductors is partnering with LG Electronics and HANA Micron to develop and commercialize chiplet-based automotive semiconductors in South Korea, connecting domestic AI chip design with packaging and vehicle testing. The collaboration is at the planned/announced stage; the source does not state the cause, and no implementation completion or specific timeline is given.
chiplet-based automotive semiconductorsBOS Semiconductors, LG Electronics and 3 totalDigiTimes AsiaPartnership/M&A
Lam Research, a US chip equipment maker, is grooming Indian companies to become its suppliers rather than relying only on its global vendor base, a senior executive said, as the company deepens its presence in India, where the government now ties incentives to local sourcing under the Semicon 2.0 localization push.
Nvidia is expanding its chip business beyond technology giants into emerging AI markets such as new cloud providers, and has become a "central bank" helping these companies expand by backing customers and demand. As Nvidia-backed AI cloud companies come under pressure, concerns are growing over how long Nvidia can sustain this role. The source states no figures, dates, or cause.
US Commerce Secretary Howard Lutnick announced a new investment by industrial minerals company Sibelco to build a quartz plant in North Carolina, expanding semiconductor-related manufacturing, as the Trump administration pushes to bring more of the chip supply chain onto US soil. This is an announced planned expansion; the source does not state the cause, and no capacity figures or timeline are given.
Nvidia's lead over Huawei in China's AI accelerator market has narrowed by 80% in a year. According to IDC, Huawei shipped 572,000 Ascend chips in the first half of 2026, securing a 28% market share, just 6 percentage points behind Nvidia, compared with a 31pp gap a year earlier. This brings CEO Jensen Huang's warnings about losing the US$50 billion AI chip market in China closer to reality. The source does not state the cause of the market share change.
AI acceleratorAscend chipsNvidia, HuaweiDigiTimes Asia
Taiwan pump maker Walrus Pump indicates that liquid cooling is becoming mainstream in AI servers, and as chip heat rises, data-center customers are assessing pumps that can move 100 metric tons of water per hour. The company expects slower ordering momentum to limit its technology-pump revenue growth, while broader adoption could improve its earnings mix. The source does not state the cause of the slower ordering momentum.
AI cooling pumps (technology pumps, 100 metric tons of water per hour)Walrus PumpDigiTimes Asia
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
Texas Instruments' third quarter suggested the analog chip upturn has moved beyond restocking into a broad-based recovery, and the company's multi-year build-out of 12-inch wafer capacity is now lifting margins rather than absorbing cash. Revenue rose 28% year on year and 11% sequentially to US$6.06 billion for the quarter ended September 30, 2026.
On Oct. 21, Samsung Electro-Mechanics announced it had signed a strategic business partnership with Nvidia to supply semiconductor package substrates. The two companies will jointly invest to establish dedicated production facilities for artificial intelligence (AI) semiconductor package substrates at Samsung Electro-Mechanics' Sejong and Busan plants; the plan to expand package substrate production lines through customer-backed investment, under consideration since the first half of 2026, has taken shape through the Nvidia partnership. The products include flip-chip ball grid array (FC-BGA) substrates for AI computing chips and for network switches that enable high-speed connections between chips in data centers. Samsung Electro-Mechanics will supply next-generation substrates from its Korean production facilities under a multiyear arrangement beginning in 2026. Supply volumes are expected to grow at a double-digit annual rate starting in 2026, and by 2028 Nvidia-related production is expected to account for a significant share of the company's total capacity. The substrates will also embed passive components including multilayer ceramic capacitors (MLCCs) and silicon capacitors (Si-Caps), which will help control power noise generated during ultra-high-speed computing and maximize power efficiency. This is a planned/expected expansion; implementation remains to be carried out.
Song Jae-hyuk, chairman of the Korea Semiconductor Industry Association (KSIA) and president of Samsung Electronics, said at the 19th Semiconductor Day ceremony held at the Grand InterContinental Seoul Parnas in Gangnam, Seoul, on Oct. 21 that timely investment is the most important factor for the semiconductor industry. He called on the government to take proactive administrative measures to support companies in developing new technologies and investing at the right time through proactive infrastructure development, expedited permits and approvals, and bold, swift policy support. He noted that competition in semiconductor technology is intensifying between countries, with major countries fostering semiconductors as a national strategic industry, and that as the AI era gains momentum, advanced semiconductors offering high performance and high bandwidth are becoming increasingly important. He also pledged that the association will continue investing in advanced manufacturing facilities, next-generation technology development and talent cultivation. The article states South Korea achieved $10 billion in semiconductor exports in 1994, and cumulative shipments reached $340.3 billion through September 2026, roughly double the 2025 full-year exports of $173.4 billion. This year, 54 industry figures received government awards, with the Gold Tower Order of Industrial Service Merit going to Lee Jun-hyuk, chairman of Dongjin Semichem. This is a call/proposal, not an implemented policy.
semiconductorKorea Semiconductor Industry Association, Samsung Electronics and 3 totalThe ElecPolicy/Regulation
Related manufacturers
Korea Semiconductor Industry Association, Samsung Electronics, Dongjin Semichem
On 21 October 2026 (Wednesday), Intel announced that it had started installing production tools at its Fab 62 in Arizona, with the fab on track for mass production by late 2028; however, the company has not disclosed when it intends to begin high-volume production, the planned wafer processing capacity, the capital expenditures required to complete the facility, and it has not officially assigned Intel 14A to Fab 62 (Fab 62 remains unassigned, while Fab 52 has been designated for Intel 18A). Per Intel's manufacturing roadmap disclosed earlier this year, Fab 62 will be able to make chips on 18A, 18A-P, and 14A nodes. Intel also confirmed that it has integrated High-NA EUV lithography into high-volume manufacturing (HVM), used to produce select tiles for Panther Lake CPUs at its D1X fab in Oregon; Intel has not explicitly confirmed whether High-NA EUV scanners are currently used to produce the same circuits as Low-NA EUV machines. In addition, construction workers recently installed the first major structural truss at the first fab of the Ohio One complex in New Albany, Ohio, but Intel did not provide an updated construction schedule, disclose when the first fab is expected to begin operations, or reveal which process technology it will use or its estimated capacity. The source does not state the cause (this is a capacity expansion progress update, not a price or disruption event).
Leaker Jaykihn on X says Intel's Core Ultra 9 4995HX looks set to be the fastest Nova Lake chip for laptops, using an 8+16+4+2 layout: 8 P-cores, 16 E-cores, 4 LPE cores, and 2 Xe3 GPU cores, totaling 28 CPU cores—four more than Intel's current Arrow Lake-HX chips, with the extra cores coming from the LPE cluster Arrow Lake-HX lacked. The layout matches the previously reported 8P+16E+4LP ceiling, and the name fits the Core Ultra 4000 branding said to be extending to HX parts. The source explicitly notes Intel hasn't confirmed any of this, and clocks, cache, and power figures are still unknown. The same post lists the Core Ultra 9 4970H as 4+8+4+12, i.e., 16 CPU cores and a 12-core Xe3P iGPU—the same core count as Panther Lake's flagship Core Ultra X9 388H, so any gains should come from the new Coyote Cove P-cores, Arctic Wolf E-cores, and Xe3P graphics rather than extra cores; the 4995HX's tiny iGPU is less of a concern since HX chips almost always pair with a discrete GPU. Jaykihn also put the final 8+16 HX compute tile, built on TSMC N2P, at 14.8 x 6.6 mm, roughly 98 mm². Current reports point to the first Nova Lake-H and HX laptops showing up around CES 2027, though that timing should be treated as tentative.
Nova Lake-H/HX laptop CPUsXe3 GPU coresXe3P iGPUIntel, TSMCTechPowerUp
Altera announced an extension of its Agilex FPGA product line to defense systems. Defense-grade Agilex FPGAs and SoC FPGAs operate reliably across a temperature range of -55°C to +125°C, with leaded-ball and ruggedized packaging options designed to withstand temperature cycling, vibration, and shock in mission-critical aerospace and defense applications. The defense-grade devices will span Altera's entire Agilex portfolio, from cost-optimized Agilex 3 FPGAs to mid-range Agilex 5 and high-performance Agilex 7 and Agilex 9 Direct RF-Series FPGAs. The article also notes that Altera's HyperFlex architecture uses Hyper-Retiming and Hyper-Pipelining to enable higher clock frequencies and greater processing throughput, and that fabrication in the United States provides a domestic source of wafer production supporting supply-chain resilience for aerospace and defense programs with long service-life requirements. Initial shipments of defense-grade Agilex FPGAs are scheduled for Q1 2027 (planned status, not yet shipped). The source does not state the cause.
Synaptics introduced the Astra SR200 series of xMCUs, which run multimodal AI workloads entirely on-device, combining heterogeneous, AI-native compute and acceleration with integrated high-bandwidth custom DRAM scalable to hundreds of Mbytes. The devices are powered by Synaptics' Torq AI engine with Google Research's transformer-capable Coral NPU and deliver up to 400 GOPS of AI inference. The series comprises three pin-compatible product lines: the SR210 reaches 200 GOPS for always-on voice and sensing; the SR230 adds vision, display, and optional audio at 300 GOPS; and the SR250 provides 400 GOPS with dual vision and display paths for demanding edge applications. The devices also feature Arm Cortex-M85 and Cortex-M52 cores, integrated vision and audio capabilities, and hardware-rooted security, with a tiered architecture pairing up to 8 Mbytes of SRAM with extensible, high-bandwidth custom DRAM that Synaptics says runs workloads at one-third the power of typical system DRAM.
Microchip's radiation-tolerant GaN point-of-load converter POL-05-20S-T delivers up to 20 A of continuous current to power FPGAs, MCUs and other digital loads in space. It uses a synchronous buck topology with current-mode control, an input voltage range of 4.75 V to 13.2 V, and an adjustable output of 0.7 V to 5 V. Its electrical parameters are specified before and after irradiation up to 50 krad(Si), and its single-event effects (SEE) have been characterized up to a linear energy transfer (LET) of 43 MeV·cm²/mg; it operates at full load over −55°C to +100°C. Housed in a 1.51×1.51×0.45-in. surface-mount package, it is now available in production quantities. The source does not state the cause (not an event-driven report).
GaN point-of-load converterGaN power HEMTsMicrochipEDN
BrainChip's Sequora processor IP core runs generative AI and large language models (LLMs) directly on battery-powered edge devices, such as earbuds and sensor nodes. It pairs with the company's Akida neural processor IP to handle complementary AI workloads: Akida processes streaming sensor data including vibration, audio, radar, raw RF IQ data, and biological signals, while Sequora runs generative and sequential AI tasks, from on-device LLM inference to voice-to-voice interaction and AI agents. Sequora is optimized for state-space models (SSMs), which maintain a fixed-size state to represent input history, keeping memory requirements constant regardless of sequence length, an architecture suited to generative AI workloads on resource-constrained edge devices. It uses a unified multiply-accumulate (MAC) array for both signal processing and neural-network computation, eliminating the need for a separate DSP preprocessing stage, and its instruction-set architecture (ISA) allows new algorithms to be deployed through software updates. Sequora is available now as licensable semiconductor IP, delivered as process-portable, fully synthesizable RTL; the package includes verification IP, synthesis scripts, and an SDK with a compiler, C++ runtime library, and model zoo.
Sequora processor IP coreAkida neural processor IPBrainChipEDN
According to ZDNet citing industry sources, Groq, the AI chip company that licensed its inference technology to NVIDIA, is reportedly considering 3D-stacked DRAM for its next-generation chips, a potential shift away from its SRAM-centric design that could significantly increase on-chip memory capacity; this remains a reported plan/rumor, not a confirmed implementation. The report indicates Groq is exploring 3DS (3D Stacked Memory), which uses through-silicon vias (TSVs) to vertically connect DRAM dies; while both 3DS and HBM rely on TSVs, HBM focuses on high data transfer speeds while 3DS aims to increase memory capacity while retaining conventional DRAM interfaces. The existing NVIDIA Groq 3 LPX is a rack-scale AI inference accelerator designed for low-latency, large-context workloads on the Vera Rubin platform; according to NVIDIA, each Groq 3 LPU integrates 500MB of SRAM with 150TB/s of on-chip memory bandwidth, while a full LPX rack combines 256 LPUs to deliver 128GB of SRAM and 40PB/s of aggregate SRAM bandwidth; according to the company, each rack provides 12TB of DDR5 memory alongside 128GB of SRAM. According to 36Kr citing d-Matrix CEO Sid Sheth, SRAM capacity increases by only around 10–20% every two years while AI model sizes could double every six to nine months, making it increasingly difficult for SRAM alone to meet growing demands for memory capacity and bandwidth. On manufacturing, according to Seoul Economic Daily, Samsung Foundry produces NVIDIA's Groq 3 LPU using its 4nm process at its Pyeongtaek facility; Samsung executives said in June that the company was discussing foundry cooperation with NVIDIA on the next-generation Groq LPU, and TSMC Chairman C.C. Wei said during the company's April earnings call that TSMC was working with customers on next-generation LPU development. Separately, Cerebras, another AI chipmaker that relies heavily on SRAM, is also pursuing 3D-stacked DRAM; according to the company, its future CS-6 system will integrate stacked DRAM with wafer-scale SRAM and compute to significantly expand memory capacity while preserving fast data access.
SRAM3D-stacked DRAM (3DS)HBMGroq, NVIDIA and 6 totalTrendForce