Marco Mezger’s Post

In its latest financial report, SK hynix stated that in the #DRAM sector in 2023, its main products #DDR5 DRAM and #HBM3 experienced revenue growth of over fourfold and fivefold, respectively, compared to the previous year. This is powered by the AI trend that needs fast and high-performance memory. According to research by TrendForce, the mainstream HBM in the market in 2023 was HBM2e. To meet the evolving demands of AI accelerator chips, manufacturers like Sk hynix, Samsung Semiconductor and Micron Technology are planning to launch new products like HBM3e in 2024, expecting HBM3 and HBM3e to become the market norm. Trendforce has captured the timeline of the big 3 memory manufacturers in a roadmap that shows that the higher-spec HBM4 is expected to launch in 2026. With the push for higher computational performance, HBM4 is set to expand from the current 12-layer (12hi) to 16-layer (16hi) stacks, spurring demand for new hybrid bonding techniques. HBM4 12hi products are set for a 2026 launch, with 16hi models following in 2027. Read more about it here: https://bit.ly/42fmMJ7 #semiconductor #memory #DRAM #NAND #flashmemory #innovation #productroadmap

Intel's 5th Gen Xeon processors are powering up AI capabilities in data centers and servers, with a focus on both general computing and specific AI tasks. With innovations like AXM & AVX, AI inference performance sees a 42% boost. Intel's commitment to CPU-based AI deployment shines in applications such as recommendation systems and voice & image recognition. Their energy-efficient Xeon processors are tailored for cloud-native designs, meeting global upgrade requirements for optimized energy performance. #Intel #AI #CloudComputing

Supercharging Computing Capabilities: Harnessing the Power of HPC and 5G Integration. Join the discussion at Embedded World and explore how the integration of HPC and 5G is reshaping the landscape. From enhanced communication networks and real-time data processing to advanced simulations and artificial intelligence applications, this forum is the go-to hub for those passionate about the future of these technologies. https://lnkd.in/eKhVuptX Three outstanding speakers: 🔹 Sravanthi Kota Venkata, Director of AI and GPU at Intel 🔹 Sabrina Liao, Senior Manager at Supermicro 🔹 Martin Ting, CEO of 7StarLake embedded world Exhibition&Conference, 7Starlake, Intel Corporation, Supermicro #EW24 #EmbeddedWorld #Embedded #EmbeddedSystems #EmbeddedSolutions #HPC #5G #Intel #SuperMicro

TSMC Dominates Another AI Battlefield: Securing HBM4 Base Die Orders This statement holds true as industry sources indicate that TSMC's subsidiary, @Global Unichip Corporation (GUC), secured a critical design order for SK hynix's HBM4 base die, a crucial component for next-generation AI servers. TSMC, already dominant in AI chip manufacturing, is poised to control another critical AI market segment: high-bandwidth memory (HBM). The company, along with its subsidiary GUC, secured a significant order for HBM4 base die chips from a major player (likely SK Hynix). This move highlights the increasing demand for HBM, fueled by the rise of AI. Next-generation HBM4, expected by 2026, promises to address bandwidth limitations with increased DRAM layers and the innovative addition of a base die logic IC for enhanced speed. #TSMC #AI #Semiconductors #HBM #HBM4 #ArtificialIntelligence #ChipManufacturing #AdvancedPackaging #Technology #Innovation #MarketTrends #SemiconductorIndustry

SK hynix and TSMC to collaborate on HBM4 innovation #SK hynix has signed an MoU with #TSMC for work together on next-generation #HBM, with plans to mass produce #HBM4 #chips from 2026 💡 HBM is a type of high-speed memory that is used in high-performance computing applications. Its high bandwidth and low latency makes it ideal for applications that require rapid data transfer. #Korean #memory specialist SK hynix currently uses a proprietary #technology to make base dies up to HBM3E, but with this MoU it plans to adopt TSMC’s advanced logic process for HBM4’s base die. This will allow additional functionality can be packed into limited space and "enable breakthroughs in memory performance through trilateral collaboration between product design, #foundry, and memory provider." Thanks again to Dennis Dahlgren and Evertiq - Global for the full article with more background and insights via the link below 🙏💡👇 https://lnkd.in/eiZEaSRF #semiconductorindustry #semiconductormanufacturing #innovation #it #ai #taiwan #semiconductors #cloud #computing #computer #dram #ic

"Lights Outside Tunnel", Doug Yu, Vice President of TSMC R&D and TSMC Distinguished Fellow, TSMC, kicks off the Advanced Packaging Conference at #SEMICONEuropa 2023. https://lnkd.in/ehC3DSec High Performance Compute (HPC) and AI/ML have been realized by advanced nodes IC and advanced system integration technologies. Device /chip scaling and heterogeneous system integration, eg. TSMC 3DFabricTM, which consists of CoWoS®, InFO and SoIC®, become the twin engine to drive semiconductor technology. Recent new wave of generative AI with LLM showed that HPC and AI/ML have tremendous room for future growth. In the meantime, higher performance compute with higher energy efficiency become even more critical requirements than ever to support the demand. TSMC will continue the scaling of both device/IC and advanced system in classical m-electronics based computing system. Furthermore, photonics-based system integration technology will be added which are complementary to the classical system integration to meet the ever-increasing energy efficient performance requirements for future HPC and AI/ML.

📈 Yole Group Press - DRAM: an industry in full flight During the #memory market winter, #DRAM demand bits stayed low, except for #AI servers and automotive #electronics. Generative AI applications like #ChatGPT drove interest in high-speed memory technologies like DDR5 DRAM and HBM, especially in data centers. According to Simone Bertolazzi, from Yole Group: “Samsung Electronics, SK hynix, and Micron Technology diverted more wafer capacity to HBM to meet demand, slowing overall bit production and increasing undersupply for non-HBM products. HBM wafer production is set to double in 2024.” Driven by AI computing demand, HBM is expected to outpace the overall DRAM market significantly. HBM bit shipments grew by 93% in 2023 and are projected to grow by 147% in 2024, with a 45% CAGR from 2023 to 2029, while data center DRAM bits will grow by 25%. HBM market revenue could rise from US$2.7 billion in 2022 to US$14 billion in 2024, accounting for 3% and 19% of overall DRAM revenue, respectively. Read the full press release on: https://zurl.co/1ZXF

Trendforce predicts robust market growth for HBM (high-bandwidth memory), along with up to 10% price increases in 2025. TrendForce says the HBM market is poised for “robust growth,” driven by significant pricing premiums and increased capacity needs for AI chips. HBM unit sales prices are several times higher than those of conventional DRAM. This pricing, combined with product iterations in AI chip technology that increase single-device HBM capacity, are expected to “dramatically raise” HBM’s share in both the capacity and value of the DRAM market from 2023 to 2025. #StoragePRSpecialists (Via BlocksAndFIles.com) https://a3.ax/QJpkz

5 Must Watch Talks on Semiconductor Industry: Opportunities and Future 1. ISSCC 2003 Plenary Presentation by Gordon Moore, Intel 2. Open-Source IC Design: Viewpoints from Academic and Industry 3. Semiconductor Industry: Present & Future by Kevin Zhang, TSMC 4. Computing in the Era of Generative AI by Jonah Alben, Nvidia. 5. Architecting the Future through Heterogeneous Computing, Lisa Su, Current CEO, AMD. Want to find out the links of these, Check first comment. For all semiconductors and beyond, Follow TechoVedas #VLSI #AI

Great to see Yole Group note that 1T0C memory, based on the floating body effect, is a contender that is being actively explored by #DRAM companies. More efficient than gain cells and also capacitor-free, Unisantis Electronics Stacked Dynamic Flash Memory (DFM) offers the potential to break the DRAM cell size barrier in fewer layers than 3D stacked 1T1C DRAM. #semiconductors https://lnkd.in/e_vrbq8n