News Posts matching #HBM

Return to Keyword Browsing

SEMI Projects 300 mm Memory Equipment Investment to Surpass $50 Billion in 2026

Worldwide 300 mm fab equipment investment in the memory sector is projected to surpass $50 billion for the first time in 2026, rising 29% to $52 billion before increasing another 11% to $57 billion in 2027, SEMI reported today in its latest 300 mm Fab Outlook. The growth reflects continued AI-driven demand for advanced memory, supported by rising investment in AI infrastructure, data centers, and next-generation computing systems.

Looking further out, global 300 mm fab equipment spending in the memory sector is expected to grow at a 19% CAGR from 2024 to 2029. Worldwide 300 mm memory capacity is also projected to increase, reaching 4.1 million wafers per month in 2026 and 4.2 million wafers per month in 2027.

DRAM Triarchy Hit with Class-Action Lawsuit Alleging Memory Price-fixing and Cartel Behavior

A class-action lawsuit has been filed in the U.S. alleging that the three leading DRAM manufacturers, namely Micron Technology, Samsung Electronics, and SK Hynix, are engaged in price-fixing and causing artificial scarcities to keep prices of computer memory high. Consumer DDR5 PC memory saw a near 700% inflation in price in a span of just 1 year, caused mainly by DRAM manufacturers allocating their production capacity to HBM stacks used by AI compute GPUs; and booking orders years into the future, which would keep a large percentage of their manufacturing capacity occupied with HBM production going as far into the future as 2030.

The lawsuit calls for PC memory manufacturers to restore consumer DRAM production and availability. Among the plenty of examples the lawsuit can cite, it mentions Apple's recent decision to increase the prices of all its products, including Macs and iPhones, citing escalating memory prices. The lawsuit alleges collusion and cartel-like behavior by Samsung, Micron, and SK Hynix, to coordinate pricing. It also cites evidence of such activity in the past, specifically in the year 2000, where the U.S. Department of Justice fined these companies hundreds of millions of dollars, besides prison sentences for several company executives.

Qualcomm HBC Gen 1 Achieves 133 TB/s Bandwidth by Stacking LPDDR Memory

Yesterday, Qualcomm announced its High Bandwidth Compute (HBC) product, a memory-compute solution hybrid designed to replace traditional High Bandwidth Memory (HBM) and offer greater performance, efficiency, and throughput. HBC utilizes LPDDR memory, stacking them across multiple layers in a 3D vertical space and interconnecting them with through-silicon vias (TSVs). This approach provides better energy efficiency than traditional HBM, which stacks regular DDR in vertical layers, as the low-power DDR (LPDDR) chips consume less power while delivering similar bandwidth and capacity. Qualcomm has successfully stacked multiple LPDDR layers on top of each other, with the base of the HBC being a compute die that performs near-memory computation and offloads some processing from the main processor.

This technique is similar to what HBM4 memory currently does, with the base die being a logic die to better integrate compute solutions underneath for packet tracing and data preparation for entering and exiting HBM. However, Qualcomm aims to achieve higher efficiency and more optimized data movement, with a 6x increase in bandwidth per watt compared to the current HBM specification, likely HBM4. HBC Gen 1 achieved 133 TB/s of bandwidth on the AI250 accelerator card, a 18x increase over the LPDDR5X currently used in the AI200 card from Qualcomm. The company also claims there is a viable roadmap for using HBC across many Qualcomm AI accelerators, with HBC Gen 1 shipping with the AI250 AI accelerator in mid-2027. The next-generation HBC Gen 2 is expected to yield further significant gains in bandwidth.

Qualcomm Unveils Data Center Roadmap with New Dragonfly Portfolio

Qualcomm Technologies, Inc. (NASDAQ: QCOM) today announced at its Investor Day, new data center solutions, including the Qualcomm Dragonfly C1000 CPU, Qualcomm High Bandwidth Compute (HBC), Qualcomm Dragonfly AI300 inference accelerator, and connectivity products, together with custom silicon solutions, all engineered to maximize performance per watt and token throughput at lower total cost of ownership. The new platforms highlight Qualcomm Technologies' growing role in building full-stack data center infrastructure optimized for AI, spanning agentic and data-center-class CPUs, AI inference accelerators, high-performance connectivity, and at scale custom silicon solutions. The Qualcomm Dragonfly AI300 joins the previously announced Qualcomm Dragonfly AI200 and AI250 in its data center solutions portfolio with an annual cadence AI accelerator roadmap.

"Agentic AI is driving a significant increase in demand for AI inference in the data center. As these become the dominant workloads, infrastructure has to deliver much higher performance at lower power and cost," said Cristiano Amon, President and CEO, Qualcomm Incorporated. "That plays directly to Qualcomm's strengths, and we're well positioned for this shift. With Qualcomm Dragonfly, we're bringing our high-performance, low-power computing into the data center, with multi-year, multi-generation agreements with leading customers."

China's LineShine Debuts as World's Fastest 2 ExaFLOP Supercomputer

According to the June update of the TOP500 list, which tracks 500 of the world's most powerful supercomputers, China has taken the number one spot with its LineShine supercomputer. This system is capable of running 2.198 ExaFLOPS when measured on HPL, making it the world's first system to exceed 2 ExaFLOPS in dual-precision FP64 data processing for sustained workloads. It has a theoretical peak performance of 2.736 ExaFLOPS, meaning the system can sustain about 80% of its peak performance. This setup is powered by a massive network of traditional CPUs, with no GPUs or other accelerators contributing to the high FP64 figure. While traditional data centers have shifted toward 8-bit FP8 data formats for AI workloads, scientific HPC systems and supercomputers continue to use the traditional FP64 data. The Chinese LineShine is the first-ever system to sustain over 2 ExaFLOPS of compute capability, surpassing U.S.-based El Capitan, Aurora, and Frontier.

The LineShine comprises 13.79 million ARMv9 cores, spread across 20,480 nodes in a non-uniform memory access (NUMA) configuration. Each node contains two LX2 CPUs, with 304 cores and 32 GB of HBM memory. The CPU uses HBM for fast memory access, then offloads to system DDR memory, which has about 256 GB per CPU. The CPU is split across two dies, each divided into four NUMA domains. Each NUMA domain contains 38 cores running at 1.55 GHz and has 4 GB of HBM assigned to it. Four of these NUMA domains make up a die, and each die is assigned 128 GB of DDR memory. This is likely DDR5, but we haven't confirmed those details yet.

Micron and Anthropic Announce Strategic Agreement to Scale Next-Generation AI Infrastructure

Micron Technology, Inc. (Nasdaq: MU) today announced a strategic agreement with Anthropic that spans memory and storage AI architecture design, supply and demand, enterprise adoption of Claude across Micron and a strategic investment in Anthropic's Series H funding round. The agreement directly links the demands of frontier AI models to how infrastructure is designed, supplied, and deployed at scale.
"The AI revolution has permanently elevated the role of memory and storage solutions from the data center to the edge," said Sumit Sadana, executive vice president and chief business officer, Micron. "Micron's strategic collaboration with Anthropic brings together the industry-leading capabilities of both companies to innovate and scale next-generation AI infrastructure."

SK hynix Ships Samples of 12-Layer Next-Gen HBM4E

SK hynix Inc. announced today that it has shipped samples of HBM4E, a next-generation DRAM for AI, to major customers. "The company was able to deliver samples of the 12-stack HBM4E on schedule thanks to its advanced HBM development and production expertise for HBM," said SK hynix, adding that "We will work closely with partners for mass production in a timely manner." The 12-layer HBM4E shows improvements in both performance and power efficiency. The product features a maximum data processing speed of 16 Gbps per pin and power efficiency that is up more than 20 percent from previous models. These enhancements improve data processing capabilities for AI training and inference.

The HBM4E reduces data transfer latency through its latest interface and design optimization while maintaining stable operation in high-bandwidth environments. This enables customers to increase efficiency in processing data for AI datacenters and large-scale computing systems. SK hynix utilizes Advanced MR-MUF technology for HBM4E products to achieve a 48 GB capacity in a 12-layer stack while ensuring structural stability. In particular, the company has also improved heat resistance by 17 percent, compared to the preceding HBM4, enabling stable operation of memory chips in high-performance computing environments.

Samsung Electronics Begins Shipment of Industry-First HBM4E Samples

Samsung Electronics, a global leader in advanced memory technology, today announced that it has begun shipping the industry's first 12-layer HBM4E samples to major global customers, further strengthening its leadership in the next-generation HBM market.

Following the industry's first mass production and commercial shipment of its industry-leading HBM4 earlier this year, Samsung now extends its HBM roadmap with the introduction of HBM4E samples, addressing the rapidly evolving demands of AI computing and hyperscale infrastructure.

Agentic AI Drives Structural Expansion in Memory Demand, Global Memory Market Projected to Reach US$1.28 Trillion by 2027

The shift in AI development from large-scale model training toward inference-centric Agentic AI applications is driving a structural expansion in memory demand, according to TrendForce's latest findings on the memory industry. With the resulting supply deficit unlikely to be resolved in the short term, prices are set to rise further.

TrendForce has therefore significantly raised its global memory market forecasts, increasing its 2026 estimate from US$551.6 billion in the previous report to $889.3 billion. Meanwhile, the 2027 forecast has been revised upward from $842.7 billion to more than $1.28 trillion, representing annual growth of approximately 44%.

SK hynix Unveils iHBM Thermal Solution to Boost AI Performance

SK hynix Inc. announced today the launch of the iHBM solution that embeds integrated cooling elements (ICEs) within the high-bandwidth memory (HBM) package for next-generation HBM products. Heat management has become a critical challenge as HBM technology advances with higher stacking and faster speeds to cater to the surging demand for AI data processing. The efficient management of power density in the Die-to-Die Physical Layer (D2D PHY)—the interface connecting HBM and GPU—has emerged as a key factor defining the competitiveness of next-generation HBM.

With the iHBM solution, the company has taken a structural approach to addressing the heat management issue. Existing HBM products rely on an indirect cooling method that draws heat away through the core die. In contrast, the iHBM solution places ICEs directly in the D2D PHY area where heat concentration is the highest, creating an additional heat dissipation path. This latest heat management solution helps reduce thermal resistance by 30% and enables chips to operate stably even in high-temperature and high-pressure conditions.

Leaked Intel Crescent Island PCB Surfaces with Xe3P GPU, 160 GB LPDDR5X and 16-Pin Power

A PCB leak from YuuKi_AnS has given us the first look at Intel's Crescent Island PCIe accelerator and a few things stand out immediately. The Xe3P GPU die is visibly larger than the Xe2-based BMG-G31 in the current Intel Arc lineup. Xe3P is the architecture following the current Xe3. While Xe3 is heading to client products under the Arc C-Series branding, Xe3P is being positioned as a more scalable design that covers client iGPUs all the way up to data center inference. Memory is the other talking point confirming previous rumors by leaker Jaykihn from April. Rather than HBM, Intel is going with LPDDR5X, the PCB showing 20 memory sites in total, 12 on the front and 8 on the back, each carrying 8 GB for a total of up to 160 GB. This is a deliberate choice made by Intel to sidestep the HBM supply issues while keeping costs down, though LPDDR5X trades bandwidth for lower pricing compared to HBM standards.

The rest of the PCB shows 13 populated VRM stages out of 18 total positions, and power comes in via a single 16-pin connector on the back of the board. A USB Type-C port on the side appears to be for engineering and testing purposes rather than a production feature. Intel Crescent Island GPUs are still a few months out, but the leaked PCB looks close to a final design according to the source. However, no gaming variant is expected based on current rumors. The company recently introduced the Arc Pro B70 and B65 professional cards based on Battlemage Xe2 silicon, while rumors surrounding a higher-end Arc B770 gaming model have largely gone quiet ahead of Computex.

Samsung Starts Shutting Down Chip Production Lines Ahead of 18-Day Worker Strike

Samsung is gradually winding down its factory production as an 18-day strike looms over the company facilities. According to South Korean news outlets, Samsung is slowing its manufacturing operations to enter what is referred to as emergency management mode, which essentially limits production capacity. According to the Seoul Economic Daily, about 43,286 factory workers have joined the picket line for the strike, which is more than half of the entire semiconductor division (DS unit) labor force. As factories face worker shortages, production lines are gradually stalling. In emergency mode, Samsung will restrict the number of new wafers introduced into the production process, which is a long and demanding procedure that typically runs 24/7. Interestingly, the factory can still operate at minimum production capacity thanks to the extensive automation involved, but operating machines still requires human involvement.

TrendForce estimates that about 3-4% of the world's global DRAM supply could be disrupted, along with about 3% of the world's NAND Flash supply. At a time when DRAM and NAND Flash shortages are at their peak, this will exacerbate the global supply chain situation. The targeted production lines in the strike are for HBM, (LP)DDR5, and some custom logic production. If Samsung's customers perceive uncertainty from the production capacity stall, many could shift to competitors like SK hynix or Micron. With more than half of the workforce absent, there will not be enough factory workers to keep production lines running, and output will drop significantly.

Intel "Razor Lake-AX" Could Bring Back On-Package Memory

Intel plans to reintroduce on-package memory in the future, with the "Razor Lake-AX" potentially being the product to carry this feature. This could be a successor to the "Nova Lake" generation, expected to arrive in late 2026. The last time Intel used on-package memory was with the "Lunar Lake" Core Ultra 200V series of CPUs, which included LPDDR5X-8533 memory in a single package alongside the CPU, GPU, and I/O die. Intel had previously stated that future CPU generations after "Lunar Lake" would use off-die memory, as sourcing and integrating additional memory modules in its SoCs is challenging, despite having technologies like EMIB and Foveros 3D to support integration. However, this decision seems to be reconsidered, as Intel is reportedly planning to integrate memory back on-package with "Razor Lake-AX."

Intel's AX variants of SoCs have long been rumored to offer significant integrated graphics capabilities to handle demanding graphics workloads. This lineup is intended to compete directly with AMD's "Halo" series, which includes "Strix Halo," "Gorgon Halo," and other future revisions like "Medusa Halo." Intel aims to achieve excellent iGPU performance, bundled with substantial integrated memory such as LPDDR5X or the upcoming LPDDR6 memory. Eventually, Intel might reintroduce HBM memory, similar to what it did with "Kaby Lake G," but we are still awaiting more details from rumors and leaks. Amidst a significant memory shortage in the PC gaming world, Intel would likely face challenges in acquiring sufficient DRAM stock. However, by the time "Razor Lake-AX" arrives in the coming years, the situation could be quite different.

SK hynix Eyes Intel EMIB 2.5D Packaging for HBM Memory

SK hynix is collaborating with Intel to utilize its Embedded Multi-die Interconnect Bridge (EMIB) 2.5D packaging technology for HBM memory. As SK hynix aims to diversify its supply chain and customers are increasingly considering Intel Foundry, the South Korean memory giant is exploring research and development efforts with Intel on 2.5D packaging technology. Intel's premier 2.5D packaging technology is its EMIB, which interconnects multiple silicon dies using bridges embedded in a packaging substrate. SK hynix is interested in integrating this technology into its HBM memory, presumably to bring its HBM4 memory modules up to standard for EMIB integration, should its AI chip partners choose Intel Foundry for advanced packaging of their next-generation solutions.

We have already covered that small silicon EMIB bridges—available in variants like EMIB-M with embedded MIM capacitors and EMIB-T with TSVs—provide low-cost, high-density shoreline connections ideal for logic-to-logic and logic-to-HBM interfaces. However, until now, SK hynix has been using TSMC and its CoWoS 2.5D packaging technology. As CoWoS is gradually reaching its limits and customers are seeking alternative packaging methods, EMIB is emerging as a strong candidate to continue the scaling of chiplets in many directions beyond the traditional reticle limit of 830 mm² of silicon area.

Intel Prepares HBM Killer: HB3DM Memory Stacks with Z-Angle Technology

Intel and SoftBank, through their subsidiary Saimemory, have been developing an alternative technology to the popular high-bandwidth memory (HBM) to provide more bandwidth and capacity for memory modules used with powerful AI accelerators. At VLSI 2026 in June, Saimemory is scheduled to present a paper on the newly developed HB3DM memory, which is based on Z-Angle Memory (ZAM) technology. This name refers to the vertical (Z-axis) stacking of dies, similar to traditional HBM. However, Intel aims to achieve impressive results using state-of-the-art manufacturing technology. The first generation of HB3DM will feature a total of nine layers, stacked using a hybrid bonding technique for 3D chip placement. At the base will be a logic layer that manages data movement within the chip, with eight DRAM layers on top for data storage. Each layer will include about 13,700 TSVs for hybrid bonding.

In terms of capacity, HB3DM will offer about 1.125 GB per layer, translating to 10 GB per memory module. Intel can achieve approximately 0.25 Tb/s of memory bandwidth per mm², and for a 10 GB module with a 171 mm² die area, we can expect around 5.3 TB/s per module. These impressive figures could quickly overshadow competing HBM4 memory, as HB3DM offers much higher bandwidth. HBM4 provides speeds of around 2 TB/s per stack, less than half of what HB3DM will deliver. However, HB3DM is limited by capacity, with only 10 GB available, whereas HBM4 can reach up to 48 GB per stack. Intel may increase the number of layers in production as HB3DM progresses, but for now, it is emerging as a bandwidth leader.

SEMI Reports Worldwide Silicon Wafer Shipments Increase 13% Year-on-Year in Q1 2026

The SEMI Silicon Manufacturers Group (SMG) reported today, in its quarterly analysis of the silicon wafer industry, that worldwide silicon wafer shipments increased 13.1% year-on-year to 3,275 million square inches (MSI) from the 2,896 MSI recorded during the same quarter of 2025. Sequentially, shipments declined 4.7% quarter-over-quarter from the 3,437 MSI recorded during the fourth quarter of 2025 in line with typical seasonality.

"Silicon wafer demand related to AI data centers continues to be strong, including advanced logic and memory, and also now extending to power management devices," said Ginji Yada, Chairman of SEMI SMG and Managing Executive Officer, General Manager, Sales and Marketing Division at SUMCO Corporation. "Overall, silicon wafer demand has improved, but the recovery is not uniform. Many device companies have noted improvements in the industrial semiconductor segment, and this is creating a more broad-based recovery as wafer inventory is absorbed. Weaker smartphone and PC shipments in the first quarter of this year may show the impact of tighter supply of memory due to AI high bandwidth memory (HBM) allocation decisions."

TSMC Debuts A13 Technology at 2026 North America Technology Symposium

TSMC (TWSE: 2330, NYSE: TSM) today debuted its latest innovation in its most advanced process technology at the Company's 2026 North America Technology Symposium. TSMC's new A13 process is a direct shrink of its industry-leading A14 node announced in 2025, enabling even more compact and efficient designs to address insatiable customer demand in computational requirements for next-generation artificial intelligence, high performance computing (HPC), and mobile applications.

Representing TSMC's commitment to continuous improvement, A13 provides 6% area savings from A14. Design rules are fully backward compatible with A14, enabling customers to quickly migrate their designs to TSMC's latest nanosheet transistor technology. In addition, A13 delivers increased power efficiency and performance gains through design-technology co-optimization, and is scheduled to enter production in 2029, one year after A14.

Intel Foundry Nears Major Customer Wins as Apple, AMD, Google, and NVIDIA Weigh Deals

Intel Foundry has been attracting significant attention from the industry lately, with some of the world's largest companies expected to announce new foundry commitments this fall, according to UBS Group. The investment bank has reported that Intel Foundry is on the brink of securing many new contracts scheduled for this fall, citing the release of version 1.0 of the 14A node process design kit (PDK) as a key catalyst. There have been numerous rumors that Apple, AMD, NVIDIA, Google, and Broadcom are considering utilizing Intel's silicon manufacturing, including the 18A, 18A-P, and 18A-PT nodes, as well as the upcoming 14A node. Apple is rumored to manufacture some of its M-Series "Apple Silicon" laptop processors in 2027 using the 18A-P node, while Google might leverage Intel's EMIB and Foveros 3D advanced packaging technology for some of its TPU designs.

Companies regularly evaluate leading industry nodes but often choose TSMC's manufacturing due to its reliability, high volume capacity, and advanced packaging, which has produced numerous high-performance and low-power designs at scale. However, Intel Foundry has been investing significant resources and logistics into attracting external clients, leading UBS to anticipate multiple foundry commitments being announced this fall. Late last year, we learned that Apple was waiting for Intel to release the 18A-P PDK version 1.0 or 1.1, scheduled for Q1 and Q2 of 2026, respectively. As we are now in Q2, we are awaiting further confirmation to see if Apple proceeded with the deal, but UBS expects it did.

ASML Targets 60+ EUV Shipments in 2026 as Memory Demand Surges

In the latest Q1 2026 quarterly figures, ASML announced plans to ship over 60 EUV units this year, including both High-NA and Low-NA EUV lithography scanners. This is a significant increase from the 48 units shipped in 2025, but it doesn't break any records compared to the company's expectations for 2027. ASML predicts shipping about 80 EUV systems in 2027, driven by massive demand from memory makers. As memory has become the most sought-after commodity in AI data centers, there is simply not enough supply to meet demand. In Q1 2026, ASML reported that nearly half—45% to be precise—of its revenue came from South Korea.

While ASML hasn't provided specific data per manufacturer, we know that South Korean semiconductor makers primarily focus on memory, storage, and to a lesser extent, logic production. More than half of the quarterly shipments, at 51%, are dedicated to memory production. Companies like SK hynix and Samsung are purchasing many EUV systems to support their growing memory and storage businesses, which include the latest technologies like GDDR6, HBM3, HBM4, and DDR4/DDR5, all consumed by AI data centers worldwide. SK hynix plans to install 20 Low-NA EUV units in the next two years, all designed for HBM memory and advanced storage solutions, making SK hynix one of ASML's largest customers.

AMD Instinct MI355X GPUs Surpass 1M Tokens/Sec in MLPerf 6.0

In its MLPerf Inference 6.0 submission, AMD did not simply revisit familiar benchmarks with a faster GPU. It expanded into first-time workloads, crossed the 1-million-tokens-per-second threshold at multinode scale and showed that partners can reproduce the results across a broader ecosystem. That combination matters because our customers no longer evaluate inference platforms on one metric alone. They want competitive single-node performance, efficient scale-out, faster bring-up on new models, reproducible results across partner systems and confidence that the software stack can keep pace. MLPerf Inference 6.0 let us show all of that in one submission.

Just as important, we showed that these results are not isolated. A broad partner ecosystem submitted across four AMD Instinct GPU types that closely reproduced numbers submitted by AMD and the first three-GPU heterogeneous MLPerf submission demonstrated that AMD hardware and AMD ROCm software can orchestrate meaningful inference throughput even across systems in different geographies.

SEMI Projects Double-Digit Growth in Global 300 mm Fab Equipment Spending for 2026 and 2027

Worldwide 300 mm fab equipment spending is expected to increase 18% to $133 billion in 2026 and 14% to $151 billion in 2027, SEMI reported today in its latest 300 mm Fab Outlook. This strong growth reflects surging AI chip demand for data centers and edge devices, as well as the growing commitment to semiconductor self-sufficiency across key regions through localized industrial ecosystems and supply chain restructuring. Looking further out, the report projects investment will continue to increase 3% to $155 billion in 2028 and another 11% to $172 billion in 2029, respectively.

"AI is resetting the scale of semiconductor manufacturing investment," said Ajit Manocha, President and CEO of SEMI. "With global 300 mm fab equipment spending projected to exceed $150 billion in 2027 for the first time, the industry is making historic, sustained commitments to the advanced capacity and resilient supply chains needed to power the AI era."

NVIDIA's "Rubin Ultra" Reportedly Faces Issues With CoWoS-L Packaging

NVIDIA is reportedly experiencing manufacturing issues with its next-generation "Rubin Ultra" GPU design, one of the company's most ambitious chip development projects, due to the limitations of modern packaging technology. The world's largest company is already shipping customer samples of the standard "Rubin" GPUs, with mass shipments set to begin this summer. However, the current roadmap for the upgraded "Rubin Ultra" design may be encountering technological limitations, as NVIDIA's design goals are too ambitious for TSMC's packaging capabilities. Reportedly, NVIDIA plans to double the regular "Rubin" two-die package with 8 HBM4 modules into a new "Rubin Ultra" package that will include four silicon dies and 16 HBM4E modules in a single package. This configuration is scheduled for 2027, but the sheer volume of silicon may be too much for TSMC's packaging, according to Global Semi Research.

In a typical CoWoS package, TSMC usually combines multiple smaller dies and multiple HBM memory modules into a unified package that supports the entire AI build-out. However, with the ambitious "Rubin Ultra" design, NVIDIA planned to use CoWoS-L, which was expected to handle the design and concept that "Rubin Ultra" was based on. It is rumored, however, that in a 2+2 die package—meaning four dies as in this architecture—TSMC is encountering warping issues. The package—which includes a substrate—is bending in multiple directions, causing the compute dies of "Rubin Ultra" to not make complete contact with the underlying substrate. This instability means that TSMC has to explore alternatives within its packaging portfolio. One of these alternatives is a panelized approach called CoPoS, which stands for Chip-on-Panel-on-Substrate.

SK Group Warns Memory Shortage May Continue into 2030

Earlier this week we reported that the memory shortage was going to persist until late 2028, but news out of Korea now suggests it might stick around all the way until 2030. SK Group Chairman Chey Tae-won is quoted by The Korea Times saying that "The shortage stems from a lack of wafer capacity, and securing additional wafers takes at least four to five years," continuing "We expect the industry-wide supply shortfall to persist at over 20 percent through 2030."

The company is apparently looking at ways to stabilise pricing, but as to when this will happen, is anyone's guess at this point in time. In this specific case, it's down to SK hynix CEO Kwak Noh-jung to try and resolve the immediate situation. SK Group doesn't seem interested in building fabs outside of South Korea either, as Chey also told The Korea Times that "Korea already has the infrastructure in place, allowing for a much faster response. That is why we are concentrating our efforts here." There is at least some good news here, as it doesn't seem like SK hynix will put its entire focus on HBM memory, as the company is aware that might lead to further shortages of DRAM, which will affect the broader tech industry.

Intel Advanced Packaging Complex in Malaysia to Go Live Later This Year

Intel is accelerating its advanced packaging push with a major move in Malaysia. According to reports from The Edge Malaysia, the company's new complex there, part of Project Pelican, is now 99% complete and set for full operations later this year. Malaysian Prime Minister Datuk Seri Anwar Ibrahim confirmed he met with Intel CEO Tan Lip-Bu and executives to review the project progress. The first phase will launch assembly and testing capabilities for advanced packaging, marking a key step in Intel's foundry expansion strategy. The facility is designed to handle die sort, prep, and full production flows across both EMIB (Embedded Multi-die Interconnect Bridge) and Foveros technologies, critical for supporting high-volume chiplet-based designs. As we reported in December last year, Intel values the project at approximately $7 billion and aims to transform Malaysia into a major regional hub for its advanced packaging operations. An additional $200 million investment has already been committed to finish the site.

Intel continues to advance its EMIB packaging approach. Unlike traditional silicon interposers used by NVIDIA in Blackwell, EMIB embeds conductive bridges directly into the substrate. This approach is cheaper, more efficient, and more suited for high-density chips. Intel is now aiming for 120 x 120 mm packages from the current 100 x 100 mm. These larger dies can support up to twelve HBM stacks, compared with eight in current designs. By 2028, Intel plans 120 x 180 mm packages capable of handling twenty-four HBM stacks. The company has also upgraded EMIB-T, its latest version incorporating through-silicon vias (TSV), to support next-gen HBM4 memory, now entering mass production. That is critical as AI chipmakers demand higher bandwidth and tighter integration. However, bigger isn't always easier. Scaling up package size increases warpage risk and yield challenges during manufacturing.

Samsung and AMD Expand Strategic Collaboration on Next-Generation AI Memory Solutions

Samsung Electronics Co., Ltd. today announced it has signed a Memorandum of Understanding (MOU) with AMD to expand their strategic collaboration on next-generation AI memory and computing technologies. The signing ceremony was held at Samsung's most advanced chip manufacturing complex in Pyeongtaek, Korea, attended by Dr. Lisa Su, Chair and CEO of AMD, and Young Hyun Jun, Vice Chairman & CEO of Samsung Electronics.

"Samsung and AMD share a commitment to advancing AI computing, and this agreement reflects the growing scope of our collaboration," said Young Hyun Jun, Vice Chairman & CEO of Samsung Electronics. "From industry-leading HBM4 and next-generation memory architectures to cutting-edge foundry and advanced packaging, Samsung is uniquely positioned to deliver unrivaled turnkey capabilities that support AMD's evolving AI roadmap."
Return to Keyword Browsing
Nov 17th, 2026 13:49 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts