News Posts matching #CoWoS

Return to Keyword Browsing

TSMC's Advanced Packaging Facilities Reportedly at Max. Capacity, Insiders Whisper About Outsourcing

Over the past weekend, Taiwanese semiconductor industry moles alleged that the nation's leading foundry service is considering external advanced packaging production avenues. According to UDN Money, TSMC's "entire CoWoS series of advanced packaging products" are fully booked up—with plenty of demand coming from the usual big AI hardware players. Various flavors of Chip-on-Wafer-on-Substrate (CoWoS) exist, including the currently vaunted CoWoS-L and CoWoS-S processes. A surge in orders—reportedly coming from NVIDIA, Google, Amazon, and MediaTek—has supposedly prompted a revised strategy, beyond planned expansions of first-party facilities (primarily CoWoS-L). In order to meet lofty VIP client requests, TSMC leadership could be bolstering relationships with third-party chip packagers. The UDN Money report cites predictions made by Counterpoint, a research company—they reckon that ASE Technology (in 2026) will benefit greatly from taking on some of TSMC's "spillover."

The premier foundry business is expected to establish an improved "supply-demand" within the next year; possibly influenced by Intel's emerging capabilities—rumors have spread about Apple and Qualcomm weighing up Team Blue's advanced packaging processes. If these whispered "backup strategies" come into effect, key customers could switch loyalties by early 2028. Currently, TSMC's Arizona hub fields a massive disadvantage: a complete lack of on-site advanced packaging. Last we heard—from inside trackers—company leadership was looking into the repurposing of a planned sixth phase fabrication plant (P6 fab). This pivot could pave the way for a "fast track" deployment of US-based CoWoS activities.

TSMC Arizona Supposedly Considering Conversion of "Phase 6 Fab" into Advanced Packaging Facility

Pleasing progress at TSMC's Arizona location has resulted in company leadership expediting a "push into advanced node deployment." Mid-October reports cited an increased domestic demand for 2 nm-class wafer products. Despite the foundry industry leader and NVIDIA's public celebration of debut "made in the USA" NVIDIA "Blackwell" wafers, TSMC's Phoenix-adjacent campus lacks in one key area: advanced packaging. Not long after the conclusion of festivities, computer hardware news outlets highlighted Fab 21-produced (4 nm) Blackwell products being sent overseas—specifically, back to Taiwan—for Chip on Wafer on Substrate (CoWoS) treatments.

Later on in October, TechPowerUp's AleksandarK covered the importance of advanced packaging on North American soil. His special editorial piece focuses on Intel Foundry's potential advantages in this area of expertise; thus poised to take a leading position within the region's semiconductor industry. Yesterday, Taiwan's Liberty Times outlined a major near-future change for one section's of TSMC Arizona's vast spread: "(leadership) plans to convert the land of the planned sixth phase fabrication plant (P6 fab) into an advanced packaging (AP) plant. If the construction goes smoothly, it is expected to be installed as early as the end of 2027 and enter the pre-production stage." Some of the claimed inside track timelines seems to roughly align with previous reports of TSMC and Amkor Technology's $7 billion forthcoming "advanced outsourced semiconductor packaging and testing" facility.

TSMC Showcases Custom C-HBM4E, N3P Logic Dies Target Double Efficiency

At last week's Open Innovation Platform Ecosystem Forum in Amsterdam, TSMC provided more details over its custom HBM for the HBM4 generation, according to reports from HardwareLUXX. The company's custom C-HBM4E logic die is expected to shift to the N3P node, alongside a voltage drop from 0.8 V to 0.75 V. According to TSMC's presentation, the move targets roughly 2× better power efficiency versus today's DRAM processes. Even standard HBM4 base dies will see upgrades. Instead of using a conventional DRAM process as in HBM3E, TSMC plans to manufacture HBM4 base dies on its N12 logic node, reducing operating voltage from 1.1 V to 0.8 V. The company expects this alone to provide around a 1.5× efficiency gain. For C-HBM4E specifically, the base die doesn't just move to the N3P node, it also integrates the memory controllers directly into the stack. Those blocks normally sit on the host SoC, so the PHY becomes a fully custom design as well.

Moreover, TSMC is expanding its advanced packaging lineup with updated InFO and SoW options, but CoWoS remains the company's main growth driver. After moving from 1.5× to 3.3× reticle sizes with support for eight HBM chips, CoWoS now heads toward CoWoS-L, enabling up to 12 HBM3E/HBM4 stacks for 2026 AI parts, followed by a larger A16 generation version planned for 2027. TSMC is also securing major customers for its custom HBM logic dies. Micron has selected the foundry to build the logic base die for its HBM4E parts, with volume production planned for 2027. SK Hynix is also reportedly preparing its first custom HBM4E products for the second half of next year. According to Korea Financial Times, SK Hynix will use TSMC's 12 nm process for mainstream server-grade HBM base dies, while NVIDIA's top-end GPUs and Google's TPUs will step up to a 3 nm node.

Intel and Amkor Team Up to Scale EMIB Packaging Production

Intel is expanding its EMIB packaging capabilities by partnering with long-time OSAT partner Amkor. This is a part of Intel's strategy to quickly increase production volumes to meet unexpected market demand. According to ET News, Intel has begun EMIB assembly at Amkor's facility in Incheon, South Korea, to handle the surge in orders from large AI customers. Outsourcing to Amkor allows Intel to boost output immediately, avoiding delays associated with internal expansions. Some of Intel's EMIB capacity might be reserved for its own upcoming products, making the partnership with an external packaging provider advantageous.

This initiative shows just how much of a capacity crunch for advanced heterogeneous packaging is happening right now, as hyperscalers and AI chip companies place large orders for CoWoS and similar services. While TSMC has been a primary choice for many high-density assemblies, growing interest in Intel's EMIB and Foveros options has led partners like MediaTek, Google, Qualcomm, and Tesla to consider alternatives. Although Intel has packaging capabilities in the United States, shifting production to Amkor reduces lead times and positions EMIB as a significant external revenue source ahead of Intel's next process node launches. This builds on the Intel-Amkor partnership announced in late May of this year.

Intel Sees 5,000 W GPUs Possible with Integrated Voltage Regulators

If the performance requirements for new AI or HPC tasks demand maximum performance density regardless of power or heat constraints, Intel has a project presentation scheduled for the ISSCC conference in February 2026. One of the topics will be a 5,000 W GPU design utilizing integrated voltage regulators (IVRs). While this may seem extreme, Intel plans to leverage advanced packaging technology, specifically the Foveros-B variant, to deliver 5 kW GPUs by 2027. As accelerators scale to support larger AI and HPC workloads, traditional board-level regulators are hitting limits in current density and transient response.

Moving voltage regulation into the package shortens current paths and reduces delivery losses. Intel Foundry and its packaging division are already exploring high-density power delivery and kW-class integrated voltage regulators. Intel's Foveros roadmap targets production-ready integrated power elements by 2027, enabling customers to evaluate IVR-enabled assemblies at scale very soon. While Foveros-B is a 2027 product, customers could potentially evaluate multi-kilowatt designs with IVRs next year. Intel is not a singled-out player with multi-kilowatt designs, as NVIDIA "Rubin" silicon is rumored to have a TDP of up to 2.3 kW for highest-end models, leading to rack-level power consumption exceeding 250 kW.

TSMC Reiterates Node Leadership, Shows Progression from N2 to A14

At the Open Innovation Platform Ecosystem Forum, TSMC outlined its near-term roadmap, highlighting how increasing AI workloads are intensifying the node development and balance between performance and efficiency. The company announced that N2 is now in volume production and expects N2P to ramp up in early 2026. TSMC aims to deliver initial A16 parts, which combine nanosheet transistors with a rear-side Super Power Rail (SPR), by the end of 2026. The progression of nodes moves from N3 to nanosheet-based N2, then to A16 with SPR, and eventually to the more advanced A14 node.

TSMC provided data showing a performance increase of about 1.8x when transitioning from N7 to A14 at constant power, along with an overall efficiency improvement of approximately 4.2x over the same period. For A16, the company predicts clock speeds 8-10% higher than N2P at the same voltage, while reducing energy consumption by 15-20% for similar throughput. For those wanting to remain on FinFET, enhanced FinFET options like N3C and N4C are still available, with N4C already in use by customers. TSMC also highlighted NanoFlex, a cell-level tuning method introduced with N2, which allows design teams to balance speed and efficiency. This approach can achieve 15% frequency gains or energy reductions of up to 30% in suitable designs.

AI Boom Drives Demand for Ultra-Large Packaging as ASICs Expected to Shift from CoWoS to EMIB

TrendForce's latest investigations reveal that the rapid expansion of AI and HPC is increasing the need for heterogeneous integration, positioning advanced packaging as a strategic priority. TSMC's CoWoS platform is currently the leading solution in this area. However, as CSPs accelerate their in-house ASIC development to accommodate more complex functions, their packaging size requirements are growing substantially. As a result, some CSPs are contemplating a switch from TSMC's CoWoS to Intel's EMIB.

TrendForce notes that CoWoS connects compute logic, memory, and I/O dies via an interposer, which mounts them on a substrate. The technology has expanded to include CoWoS-S, CoWoS-R, and CoWoS-L. Demand has been strongly shifting toward CoWoS-L, which incorporates a silicon interposer into the package, as NVIDIA's Blackwell platform approaches mass production in 2025. This trend is expected to continue with NVIDIA's upcoming Rubin architecture, which will boast even larger reticle sizes.

Apple, Broadcom, and Qualcomm May Tap Intel Advanced Packaging

Apple, Broadcom, and Qualcomm might soon become customers of Intel Foundry, but not for the typical reasons. These giants are considering Intel Foundry for its advanced packaging technologies, which could allow Intel to significantly increase its foundry revenue without manufacturing the actual silicon. Job listings from Apple, Broadcom, and Qualcomm highlight Intel's Embedded Multi-die Interconnect Bridge (EMIB) as a key requirement for packaging engineer positions. This suggests they are keen on hiring engineers familiar with Intel's EMIB technology to help design their next-generation products.

Interestingly, Qualcomm's CEO recently stated that "Intel is not an option today," but expressed a desire for Intel to become a viable choice. He noted that Intel's current foundry node, likely the 18A, is not suitable for their mobile chips. The main focus of the 18A node is not on mobile, low-power system-on-chips (SoCs), but rather on mid-range and high-power solutions. As the CEO put it, "We design all of our chips assuming there's a battery on the other side, not plugged into the wall." However, 18A node is not what seemingly attracts customers like Qualcomm. It is the packaging that Intel offers. Last year, Broadcom tested the 18A node but expressed disappointment. However, the situation is now evolving differently, with Intel's advanced packaging emerging as a promising alternative to TSMC's CoWoS and other packaging types.

Socionext Introduces Flexlets Configurable Chiplets

Socionext Inc., a global leader in System-on-Chip (SoC) design and advanced semiconductor solutions, introduces "Flexlets", a new class of configurable chiplets designed to advance heterogenous integration. As traditional monolithic SoC designs face physical and economic limits—reticle size constraints, yield challenges, and thermal bottlenecks- the industry is turning toward chiplet-based design, where designers can integrate their core features and interface functionalities, provided as chiplets, into a packaged device to improve performance, cost, and accelerate time-to-market.

While chiplet technology opens exciting possibilities for modular designs and scalability, many current solutions are derived from fixed-function ASSPs, limiting flexibility and customization. Socionext's Flexlets overcome this by offering a configurable library of chiplet designs at the RTL level. Unlike traditional approaches, Flexlets empower customers to tailor performance to their unique application needs - whether in high-performance computing, advanced networking, or next-generation automotive systems.

NVIDIA's Arizona-Made "Blackwell" GPUs Must Travel Back to Taiwan for Packaging

Just three days ago, NVIDIA and TSMC celebrated the production of the first "Blackwell" GPU at TSMC's Fab 21 complex in Phoenix, Arizona. This marked a significant milestone for both companies, as it signifies the creation of the world's most powerful silicon on U.S. soil. However, the production phase is only part of the story, as the chip benefits not only from an advanced node but also from TSMC's advanced packaging systems. Specifically, the Blackwell family of GPUs utilizes TSMC's Chip on Wafer on Substrate (CoWoS) packaging system.

For CoWoS access, TSMC must send these chips back to the Taiwanese facilities overseas and house the raw die in a CoWoS-S configuration with silicon interposer. This establishes a supply chain dependency on Taiwanese facilities, which remain essential for completing the production of advanced silicon. As a result, the chip isn't entirely manufactured on American soil, indicating that some issues still need to be resolved. TSMC produces NVIDIA's Blackwell chips using a custom 4N 4 nm class node specifically designed for NVIDIA. This die is paired with eight stacks of HBM3E, all housed within a CoWoS-S package.

Alphawave Semi Delivers Cutting-Edge UCIe Chiplet IP on TSMC 3DFabric Platform

Alphawave Semi (LSE: AWE), a global leader in high-speed connectivity and compute silicon for the world's technology infrastructure, has announced the successful tape-out of its cutting edge UCIe 3D IP on the advanced TSMC SoIC (SoIC-X) technology in the 3DFabric platform. This achievement builds on Alphawave Semi's established baseline of leading UCIe IP subsystems and presents a significant evolution in the company's chiplet integration capabilities. By leveraging TSMC's SoIC-X advanced 3D packaging technology, Alphawave Semi continues to push the boundaries of power, performance efficiency and bandwidth for next generation datacenter, AI, and HPC applications.

The IP supports face-to-face (F2F) configurations and provides a 10x improvement in power efficiency over traditional 2.5D die-to-die interfaces. It also delivers up to a 5x increased signal density.

NVIDIA to Tap TSMC's A16 Node for "Feynman" GPUs

NVIDIA is preparing a significant change in its chip strategy for its next-generation GPU design, codenamed "Feynman", as the first customer of TSMC's A16 process. This choice breaks NVIDIA's long-running pattern of adopting slightly older nodes to extract greater efficiency and performance gains for its datacenter accelerators. NVIDIA's biggest AI dies are enormous, at 104 billion transistors currently per each "Blackwell Ultra" die and occupying areas that are breaking reticle limits. Such a scale typically favors one or two generations old nodes to protect yield and thermal characteristics. To keep up with the compute demand, NVIDIA appears to be adapting its architectures to A16 while keeping other families like "Rubin" on refined N3P variants and paired with CoWoS-L packaging.

The most appealing aspect of the A16 node will be its backside power delivery, which separates power and signal routing to reduce losses and simplify floor planning. Implementations range from buried power rails that still use a front-side lowest metal layer to power vias that reach the rear side. TSMC's Super Power Rail concept enables power to be delivered directly to the transistor's source and drain terminals, resulting in lower resistance. Those choices affect efficiency, layout, and thermal behavior. Power demands for high-end AI accelerators are rising fast. Current flagship solutions already draw about 1,400 W in Blackwell Ultra, and next-generation Rubin designs are expected to exceed 2,300 W. Hence, node selection will be critical for the next-generation Feynman design, given its multi-gigawatt power requirement. If A16 reaches volume production on schedule, first silicon could arrive in late 2027 or early 2028.

TSMC Prepares CoWoS to CoPoS Shift with 750 × 620 mm Panels

TSMC is reportedly preparing to transition its advanced packaging strategy from CoWoS to a panelized approach called CoPoS, which stands for Chip-on-Panel-on-Substrate. The company plans pilot lines as soon as 2026 and aims for mass production between late 2028 and the first half of 2029. The change involves swapping around 300 mm silicon interposers for large square and rectangular panels, with early formats measuring around 310 × 310 mm and later options extending to 515 × 510 mm and even 750 × 620 mm. That larger, square format is intended to reduce wasted edge area, allow bigger reticles and masks for AI accelerators, and make it easier to place more dies and HBM on a single package.

Technically, CoPoS combines CoWoS ideas with fan-out panel-level packaging, building RDL layers on glass or sapphire panels rather than on circular silicon. TSMC is lining up CoPoS development and production at its AP7 Chiayi campus, where phases 4 and 5 are expected to host panel work, and it plans related capacity in future Arizona facilities. Equipment and tooling orders are already being placed, and the supplier list includes global names like KLA and Applied Materials, as well as many Taiwanese manufacturers, such as Innolux. The panel route promises higher area utilization, lower cost per packaged accelerator, and fewer warpage issues as package sizes grow, while TSMC continues to support its existing CoWoS lineup during the transition.

Apple M5 Silicon Adopts LMC Packaging, Paving the Way for CoWoS

According to the well-known Apple insider Ming-Chi Kuo, Apple has reworked a part of its semiconductor supply chain by awarding Taiwan's Eternal Materials exclusive roles in supplying key packaging compounds for its 2026 iPhone and Mac processors. Eternal will supply molding underfill, known as MUF, for the A20 iPhone packages and a liquid molding compound, or LMC, for the high-end M5 Mac processors. TSMC has completed the qualification of these materials and validation for CoWoS use is underway, which could lead to initial shipments in 2026 and a larger role for Eternal in 2027 and 2028. This is noteworthy because it places a Taiwanese materials supplier into a space long dominated by Japanese firms and aligns Apple and TSMC around packaging choices that enable higher memory bandwidth and denser compute layouts.

For the iPhone, transitioning from InFO to wafer-level multi-chip module packaging with MUF reduces material consumption. It simplifies assembly by combining underfill and molding into a single wafer-level process, which is expected to improve yields and lower costs. For Macs, adopting CoWoS-compatible LMC for M5 packages provides better structural integrity, enhanced thermal paths, and more consistent manufacturing outcomes, while leaving full CoWoS adoption for future chip generations. Eternal reportedly beat established Japanese suppliers such as Namics and Nagase to win these orders, which could prompt OSAT partners and other manufacturers to trial the new compounds. It also sets a clear path for Apple to scale more complex multi-die designs in the coming years as it prepares the groundwork for CoWoS and later CoPoS-based processors.

Intel Wins Tesla Dojo 3 Packaging Contract in Dual-Supplier Strategy

Intel has secured Tesla as a customer for its specialized module-level packaging services, signaling that billions of investments and years of development are appealing to external Foundry Customers. Tesla's Dojo 3 initiative will see Samsung Electronics Foundry manufacture the custom D3 AI-training chips. At the same time, Intel handles the packaging and testing using its Embedded Multi-Die Interconnect Bridge (EMIB) technology, which connects multiple dies through silicon bridges rather than a full-wafer interposer. This dual-vendor approach diverges from Tesla's previous reliance on TSMC for both fabrication and System-on-Chip packaging. Tesla CEO Elon Musk confirmed that Samsung will focus on mass-producing the next-generation AI6 chips on a 2 nm process, and that the packaging will leverage Intel's EMIB platform.

Tesla's decision rests on the unique demands of its ultra-large Dojo modules, which pack multiple 654 mm² dies into a single array. Traditional packaging methods struggle to handle such sizes efficiently. TSMC's SoW solution faces capacity constraints amid high demand for CoWoS services. Intel's EMIB offers higher modularity, allowing Tesla to tailor interconnect layouts and integrate additional features as needed. Samsung's 2 nm node, paired with Intel's assembly expertise, provides Tesla with optimized yield, cost control, and a faster ramp-up for its AI infrastructure. This collaboration between two historic rivals is unusual, but much needed for both foundries to survive TSMC's competition. By splitting fabrication and packaging, Tesla aims to accelerate production for its full-self-driving systems, robotics ventures, and data-center operations.

TSMC Prepares "CoPoS": Next-Gen 310 × 310 mm Packages

As demand for ever-growing AI compute power continues to rise and manufacturing advanced nodes becomes more difficult, packaging is undergoing its golden era of development. Today's advanced accelerators often rely on TSMC's CoWoS modules, which are built on wafer cuts measuring no more than 120 × 150 mm in size. In response to the need for more space, TSMC has unveiled plans for CoPoS, or "Chips on Panel on Substrate," which could expand substrate dimensions to 310 × 310 mm and beyond. By shifting from round wafers to rectangular panels, CoPoS offers more than five times the usable area. This extra surface makes it possible to integrate additional high-bandwidth memory stacks, multiple I/O chiplets and compute dies in a single package. It also brings panel-level packaging (PLP) to the fore. Unlike wafer-level packaging (WLP), PLP assembles components on large, rectangular panels, delivering higher throughput and lower cost per unit. Systems with PLP will be actually viable for production runs and allow faster iterations over WLP.

TSMC will establish a CoPoS pilot line in 2026 at its Visionchip subsidiary. In 2027, the pilot facility will focus on refining the process, to meet partner requirements by the end of the year. Mass production is projected to begin between the end of 2028 and early 2029 at TSMC's Chiayi AP7 campus. That site, chosen for its modern infrastructure and ample space, is also slated to host production of multi-chip modules and System-on-Wafer technologies. NVIDIA is expected to be the launch partner for CoPoS. The company plans to leverage the larger panel area to accommodate up to 12 HBM4 chips alongside several GPU chiplets, offering significant performance gains for AI workloads. At the same time, AMD and Broadcom will continue using TSMC's CoWoS-L and CoWoS-R variants for their high-end products. Beyond simply increasing size, CoPoS and PLP may work in tandem with other emerging advances, such as glass substrates and silicon photonics. If development proceeds as planned, the first CoPoS-enabled devices could reach the market by late 2029.

Alphawave Semi Tapes Out New UCIe IP on TSMC 2nm Supporting 36G Die-to-Die Data Rates

Alphawave Semi, a global leader in high-speed connectivity and compute silicon for the world's technology infrastructure, announced the successful tape out of one of the industry's first UCIe IP subsystem on TSMC's N2 process, supporting 36G die-to-die data rates. The solution is fully integrated with TSMC's Chip-on-Wafer-on-Substrate (CoWoS ) advanced packaging technology, unlocking breakthrough bandwidth density and scalability for next-generation chiplet architectures.

This milestone builds on the recent release of the Alphawave Semi AI Platform, proving readiness to support the future of disaggregated SoCs and scale-up infrastructure for hyperscale AI and HPC workloads. With this tape-out, Alphawave Semi becomes one of the industry's first to enable UCIe connectivity on 2 nm nanosheet technology, marking a major step forward for the open chiplet ecosystem.

AMD Celebrates Four Decades of FPGA Innovation - From Invention to AI Acceleration

This year marks the 40th anniversary of the first commercially available field-programmable gate array (FPGA), introducing the idea of reprogrammable hardware. By creating "hardware as flexible as software," FPGA reprogrammable logic changed the face of semiconductor design. For the first time, developers could design a chip, and if specs or requirements changed mid-stream, or even after manufacturing, they could redefine its functionality to perform a different task. This flexibility enabled more rapid development of new chip designs, accelerating time to market for new products and providing an alternative to ASICs.

The impact on the market has been phenomenal. FPGAs launched a $10+ billion industry and over the past four decades we have shipped more than 3 billion FPGAs and adaptive SoCs (devices combining FPGA fabric with a system-on-chip and other processing engines) to more than 7,000 customers across diverse market segments. In fact, we've been the programmable logic market share leader for the past 25 consecutive years, and we believe we are well positioned for continued market leadership based on the strength of our product portfolio and roadmap.

Intel Details EMIB-T Advanced Packaging for HBM4 and UCIe

This week at the Electronic Components Technology Conference (ECTC), Intel introduced EMIB-T, an important upgrade to its embedded multi-die interconnect bridge packaging. First showcased at the Intel Foundry Direct Connect 2025 event, EMIB-T incorporates through-silicon vias (TSVs) and high-power metal-insulator-metal capacitors into the existing EMIB structure. According to Dr. Rahul Manepalli, Intel Fellow and vice president of Substrate Packaging Development, these changes allow a more reliable power supply and stronger communication between separate chiplets. Conventional EMIB designs have struggled with voltage drops because of their cantilevered power delivery paths. In contrast, EMIB-T routes power directly through TSVs from the package substrate to each chiplet connection. The integrated capacitors compensate for fast voltage fluctuations and preserve signal integrity.

This improvement will be critical for next-generation memory, such as HBM4 and HBM4e, where data rates of 32 Gb/s per pin or more are expected over a UCIe interface. Intel has confirmed that the first EMIB-T packages will match the current energy efficiency of around 0.25 picojoules per bit while offering higher interconnect density. The company plans to reduce the bump pitch below today's standard of 45 micrometers. Beginning in 2026, Intel intends to produce EMIB-based packages measuring 120 by 120 millimeters, roughly eight times the size of a single reticle. These large substrates could integrate up to twelve stacks of high-bandwidth memory alongside multiple compute chiplets, all connected by more than twenty EMIB bridges. Looking further ahead, Intel expects to push package dimensions to 120 by 180 millimeters by 2028. Such designs could accommodate more than 24 memory stacks, eight compute chiplets, and 38 or more EMIB bridges. These developments closely mirror similar plans announced by TSMC for its CoWoS technology. In addition to EMIB-T, Intel also presented a redesigned heat spreader that reduces voids in the thermal interface material by approximately 25%, as well as a new thermal-compression bonding process that minimizes warping in large package substrates.

AMD Prepares Instinct MI450X IF128 Rack-Scale System with 128 GPUs

According to SemiAnalysis, AMD has planned its first-ever rack-scale GPU cluster for the second half of 2026, when it shows its first rack-scale accelerator, the Instinct MI450X IF128. Built on what's expected to be a 3 nm-class TSMC process and packaged with CoWoS-L, each MI450X IF128 card will include at least 288 GB of HBM4 memory. That memory will sustain up to 18 TB/s of bandwidth, driving around 50 PetaFLOPS of FP4 compute while drawing between 1.6 and 2.0 kW of power. In our recent article, we outlined that AMD split the Instinct MI400 series into HPC-first MI430X and MI450X for AI. Now for AI-focused MI450X, the company created both an "IF64" backplane for simpler single-rack installs and the full-blown "IF128" for maximum density. The IF128 version links 128 GPUs over an Ethernet-based Infinity Fabric network and uses UALink instead of PCIe to connect each GPU to three built-in Pensando 800 GbE NICs. That design delivers about 1.8 TB/s of unidirectional bandwidth per GPU and a total of 2,304 TB/s across the rack.

With 128 GPUs each offering 50 PetaFLOPS of FP4 compute and 288 GB of HBM4 memory, the MI450X IF128 system delivers a combined 6,400 PetaFLOPS and 36.9 TB of high-bandwidth memory, and MI450X IF64 provides about half of that. Since AI deployments require massive density of rack systems, AMD plans to possibly outnumber NVIDIA's upcoming system known as "Vera Rubin" VR200 NVL144 (144 compute dies, 72 GPUs), which tops out at 3,600 PetaFLOPS and 936 TB/s of memory bandwidth—about half of what AMD's IF128 approach promises. AMD will have a possibly more powerful system architecture than NVIDIA until the launch of VR300 "Ultra" NVL576, which has 144 GPUs, each carrying four compute dies, totaling 576 compute chiplets.

TSMC Outlines Roadmap for Wafer-Scale Packaging and Bigger AI Packages

At this year's Technology Symposium, TSMC unveiled an engaging multi-year roadmap for its packaging technologies. TSMC's strategy splits into two main categories: Advanced Packaging and System-on-Wafer. Back in 2016, CoWoS-S debuted with four HBM stacks paired to N16 compute dies on a 1.5× reticle-limited interposer, which was an impressive feat at the time. Fast forward to 2025, and CoWoS-S now routinely supports eight HBM chips alongside N5 and N4 compute tiles within a 3.3× reticle budget. Its successor, CoWoS-R, increases interconnect bandwidth and brings N3-node compatibility without changing that reticle constraint. Looking toward 2027, TSMC will launch CoWoS-L. First up are large N3-node chiplets, followed by N2-node tiles, multiple I/O dies, and up to a dozen HBM3E or HBM4 stacks—all housed within a 5.5× reticle ceiling. It's hard to believe that eight HBM stacks once sounded ambitious—now they're just the starting point for next-gen AI accelerators inspired by AMD's Instinct MI450X and NVIDIA's Vera Rubin.

Integrated Fan-Out, or InFO, adds another dimension with flexible 3D assemblies. The original InFO bridge is already powering AMD's Instinct cards. Later this year, InFO-POP (package-on-package) and InFO-2.5D arrive, promising even denser chip stacking and unlocking new scaling potential on a single package, away from the 2D and 2.5D packaging we were used to, going into the third dimension. On the wafer scale, TSMC's System-on-Wafer lineup—SoW-P and SoW-X—has grown from specialized AI engines into a comprehensive roadmap mirroring logic-node progress. This year marks the first SoIC stacks from N3 to N4, with each tile up to 830 mm² and no hard limit on top-die size. That trajectory points to massive, ultra-dense packages, which is exactly what HPC and AI data centers will demand in the coming years.

Cadence and TSMC Advance AI and 3D-IC Chip Design with Certified Design Solutions for TSMC A16 and N2P Process

Cadence today announced it is furthering its longstanding collaboration with TSMC to accelerate time to silicon for 3D-IC and advanced-node technologies through certified design flows, silicon-proven IP and ongoing technology collaboration. As a leading provider of IP for TSMC N2P, N5 and N3 process nodes, Cadence continues to deliver cutting-edge AI-driven design solutions to the TSMC ecosystem for multiple horizontal applications from chiplets and SoCs to advanced packaging and 3D-ICs. The deep collaboration encompasses certified tools and flows for TSMC's N2P and A16 technologies, paves the way for TSMC's A14 and further unlocks 3D-IC possibilities by extending support for TSMC 3DFabric design and packaging. In addition, Cadence and TSMC are extending tool certification for newly announced TSMC N3C technology based on available N3P design solutions.

Cadence is driving innovation in AI chip design with certified tools and optimized IP for TSMC's advanced N2P and A16 process technologies. Reinforcing its memory IP leadership, Cadence offers TSMC9000 pre-silicon-certified DDR5 12.8G IP for N2P. Cadence digital, custom/analog design and thermal analysis solutions are certified for TSMC N2P and A16 technologies. Combined with continued collaboration on AI-driven digital design solutions for N2P, including leveraging large language models (LLMs), these advancements play an important role in improving digital design flows for future process nodes.

TSMC Unveils Next-Generation A14 Process at North America Technology Symposium

TSMC today unveiled its next cutting-edge logic process technology, A14, at the Company's North America Technology Symposium. Representing a significant advancement from TSMC's industry-leading N2 process, A14 is designed to drive AI transformation forward by delivering faster computing and greater power efficiency. It is also expected to enhance smartphones by improving their on-board AI capabilities, making them even smarter. Planned to enter production in 2028, the current A14 development is progressing smoothly with yield performance ahead of schedule.

Compared with the N2 process, which is about to enter volume production later this year, A14 will offer up to 15% speed improvement at the same power, or up to 30% power reduction at the same speed, along with more than 20% increase in logic density. Leveraging the Company's experience in design-technology co-optimization for nanosheet transistor, TSMC is also evolving its TSMC NanoFlex standard cell architecture to NanoFlex Pro, enabling greater performance, power efficiency and design flexibility.

GUC Announces Tape-Out of the World's First HBM4 IP on TSMC N3P

Global Unichip Corp. (GUC), the Advanced ASIC Leader, announced today that it has successfully taped-out the world's first HBM4 controller and PHY IP. This test chip was implemented using TSMC's cutting-edge N3P process technology and CoWoS -R advanced packaging technology.

The HBM4 IP supports data rates of up to 12 Gbps under all operating conditions. By leveraging a proprietary interposer layout, GUC has optimized signal integrity (SI) and power integrity (PI) to achieve these high speeds for all types of CoWoS technology. Comparing with HBM3, GUC's HBM4 PHY delivers 2.5x bandwidth while improving 1.5x power efficiency and 2x area efficiency. In line with previous GUC HBM, GLink, and UCIe IPs, this HBM4 IP integrates proteanTecs' interconnect monitoring solution to provide high visibility for testing and characterizing the PHY while improving in-field performance and reliability for end products.

GUC Launches First 32 Gbps per Lane UCIe Silicon Using TSMC 3nm and CoWoS Technology

Global Unichip Corp. (GUC), the Advanced ASIC Leader, today announced the successful launch of industry's first Universal Chiplet Interconnect Express (UCIe) PHY silicon, achieving a data rate of 32 Gbps per lane, the highest speed defined in the UCIe specification. The 32G UCIe IP, supporting UCIe 2.0, delivers an impressive bandwidth density of 10 Tbps per 1 mm of die edge (5 Tbps/mm full-duplex). This milestone was achieved using TSMC's advanced N3P process and CoWoS packaging technologies, targeting AI, high-performance computing (HPC), xPU, and networking applications.

In this test chip, several dies with North-South and East-West IP orientations are interconnected through CoWoS interposer. The silicon measurements show robust 32 Gbps operation with wide horizontal and vertical eye openings. GUC is working aggressively on the full-corner qualification, and the complete silicon report is expected to be available in the coming quarter.
Return to Keyword Browsing
Nov 17th, 2026 22:36 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts