KOSPI +1.40%: Foreign Capital Rotates Into HBM Supply Chain
Foreign investors poured into South Korea’s high‑bandwidth memory (HBM) and related AI‑infrastructure supply chain, lifting the KOSPI 1.4% as the KRW weakened. The inflow coincided with a post‑ceasefire risk‑premium compression and a 2.1% rally in the S&P 500, signaling targeted conviction rather than a broad risk‑on move. Market participants now hinge the trade on the March US CPI: a print at or below the 1% month‑over‑month consensus would validate the Korean positioning, while an overshoot could trigger a rapid unwind. The outcome will ripple through US semiconductor leaders such as Nvidia, Micron and Applied Materials.
SPINS Project Aims for Millions of Stable Semiconductor Qubits
The EU‑backed SPINS project secured a €50 million (~$54 million) investment to create a pan‑European research and production hub for semiconductor spin qubits. Coordinated by imec and involving 25 organisations, the consortium will develop three material platforms—Si/SiGe, Ge/GeSi and SOI—to deliver stable,...
Snapdragon 8 Gen 6 Pro: The 2nm Powerhouse Inside the Samsung Galaxy S27 Ultra
Leaks suggest Samsung’s upcoming Galaxy S27 Ultra will be powered by Qualcomm’s Snapdragon 8 Elite Gen 6 Pro, a 2 nm chipset featuring a 2‑plus‑3‑plus‑3 core layout and the Adreno 850 GPU with 18 MB of dedicated graphics memory. The device is also expected to be the first smartphone to...
Jim Cramer Shows Positive Sentiment Toward Broadcom’s Google and Anthropic Deals
Jim Cramer praised Broadcom Inc. after the chipmaker announced two AI‑focused contracts, one with Google and another with Anthropic. The endorsement helped the stock surge more than 6% in after‑hours trading. Cramer highlighted Broadcom’s $1.5 trillion market cap and its broad...
Google's Compute Domination
Google’s Tensor Processing Unit (TPU) fleet expanded 11.5‑fold over seven quarters and now consumes more electricity than Microsoft’s entire AI compute stack. The growth rate is accelerating, with Q4 2025 adding more compute in a single quarter than xAI has built...
F&S M.2 AI Accelerator Uses NXP Ara-240 for Edge Inference Workloads
F&S Elektronik Systeme launched an M.2 AI accelerator powered by NXP Ara‑240, delivering up to 40 TOPS for edge inference. The module uses a standard M.2 Key‑M 2280 form factor, supports PCIe Gen3/Gen4 x4, and includes up to 16 GB LPDDR4 memory while drawing...
Intel Nova Lake-S Could Come with Optional 2L-ILM: Leak Suggests a Flatter IHS Contact for Enthusiast Boards
A leak from VideoCardz suggests Intel’s upcoming Nova Lake‑S platform may include an optional two‑lever independent loading mechanism (2L‑ILM) for high‑end enthusiast motherboards. The 2L‑ILM would provide a flatter contact surface between the CPU’s integrated heat spreader (IHS) and the cooler,...
TSMC Is Upgrading Japan’s Second Plant to the 3-Nanometer Process. Kumamoto Is Transitioning From a Backup Site to a True...
Taiwan Semiconductor Manufacturing Co. (TSMC) has received approval to launch 3‑nanometer production at its second Japanese fab in Kumamoto, with equipment installation slated for 2026 and volume output expected in 2028. The plant will initially run at a capacity of...
SiFive Raises $400 Million; Nvidia Bets on RISC-V for Data Centers
SiFive announced a $400 million Series G financing round that values the RISC‑V IP company at $3.65 billion. Nvidia participated as an investor, reinforcing a joint plan to deliver high‑performance RISC‑V CPUs with NVLink Fusion connectivity for data‑center AI workloads. The capital will...
Samsung’s Profit Surges On AI-Related Gains, Japan Moves to the Industrial Deployment of AI
Samsung projected a first‑quarter operating profit of about $38.7 billion, driven by soaring demand for AI‑optimized DRAM and HBM memory. South Korean chip makers, led by SK Hynix’s 60% HBM share, saw stock gains as memory prices are forecast to jump 58‑63%...
CRDO As An AI Play
Credo Technology Group (CRDO) is shifting from a niche SerDes and active cable supplier to a full‑stack AI connectivity fabric architect. The company aims to close the reliability gap in massive GPU clusters by delivering vertically integrated interconnect solutions that...
YieldHUB Expands Its Impact with New Technology and a New Website
YieldHUB has launched a redesigned website and introduced YieldHUB Live, a real‑time manufacturing intelligence layer for semiconductor test floors. The new portal groups solutions by product lifecycle, device architecture and user role, while the live platform delivers continuous visibility, anomaly...
NXP Expands Arteris NoC Deployment to Scale Edge AI Architectures
NXP announced an expanded deployment of Arteris’s NoC and cache‑coherent IP suite—including FlexNoC®, Ncore®, CodaCache® and the Magillem® integration platform—across its upcoming edge‑AI silicon. The move targets the growing need for deterministic latency, high bandwidth, and safety‑critical isolation in heterogeneous...
SiFive Raises $400M To Double Down On High Performance RISC-V For Data Centers
SiFive announced a $400 million Series G financing round to accelerate its high‑performance RISC‑V offerings for data‑center workloads. The round was oversubscribed, with lead investors including NVIDIA and Apollo Global Management. Proceeds will fund new CPU core designs, accelerators, and system IP,...
RISC-V Optimized Strnlen Implementation For Linux 7.1 Yields Big Speed-Up
A hand‑optimized RISC‑V implementation of the kernel’s strnlen() function is slated for Linux 7.1. Developed by Feng Jiang of KylinOS, the assembly version includes a generic path and a Zbb‑enabled variant, delivering up to a 427.5% speed increase in benchmarks. The...
Architecting Intelligence: The Rise of RISC-V CPUs in Agentic AI Infrastructure
SiFive announced a $400 million Series G round that lifts its valuation to $3.65 billion, earmarked for next‑generation RISC‑V CPU IP aimed at agentic AI data‑center workloads. The funding will accelerate hardware co‑design that tightly integrates scalar, vector and matrix compute units to...
HighPoint Announces Rocket 1604L Compact PCIe Gen5 X16 Retimer AIC for AI and Industrial Edge
HighPoint Technologies unveiled the Rocket 1604L, a 4‑M.2 PCIe Gen5 x16 add‑in card powered by Astera Labs' Gen5 retimer. The compact 167 mm module delivers full 32 GT/s bandwidth with near‑zero latency, targeting AI edge and high‑velocity compute workloads. It integrates a Smart Firmware Layer...
AMD Making It Easier To Embed Lemonade AI Capabilities Into Other Apps
AMD’s Lemonade open‑source AI server has released version 10.2, focusing on embeddable binaries for Linux and Windows. The new artifacts contain only the Lemond daemon, CLI and essential resources, removing web UI and Electron components. Lemonade 10.2 continues to support...
AMD Between the Memory Crisis and Price Stability: Is the 8GB Radeon RX 9060 XT a Damage Control Measure or...
AMD officially launched the Radeon RX 9060 XT on May 20 2025 with two memory options – an 8 GB model priced at $299 and a 16 GB version at $349 – targeting the 1440p segment. At the same time, a global DRAM price surge of roughly...
Framework Warns of Further Increases in RAM and SSD Costs: The Memory Crisis Is Affecting Repairable PCs as Well
Framework updated its April 6 price list, openly warning that DRAM, LPDDR5x and NAND costs are only temporarily stabilised and will likely rise through the rest of 2026. The company left DDR5 RAM prices unchanged but raised the 4‑TB WD Black...
AMD Ryzen 9 9950X3D2 Listed in Stores Ahead of Launch: Early Store Listings Show a Release Window, but No Confirmed...
AMD’s new Ryzen 9 9950X3D2 dual‑edition CPU is appearing in retailer catalogs ahead of its official launch, with multiple stores listing an April 22 pre‑order window. The processor retains the 16‑core/32‑thread Zen 5 core count but expands L3 cache to 192 MB and raises TDP...
AMD Announces Ryzen 9 9950X3D2 Pricing and Availability
AMD unveiled the Ryzen 9 9950X3D2 Dual Edition, a desktop processor featuring dual 3D V‑Cache. The chip launches on April 22 with a suggested retail price of $899. Targeted at workstation users, it aims to deliver higher cache density for compute‑intensive workloads. AMD...
Foundryecosystem Report: Nvidia GPU Delays; Tools; IC Prices
The latest Foundryecosystem Report highlights several critical shifts in the semiconductor sector. Nvidia’s new Rubin GPU ramp is trimmed to 1.5 million units after HBM4 qualification setbacks at SK Hynix and low yields at Micron, delaying mass production to September. Applied Materials...
Intel, Musk, and the Tweet That Launched a 1000 Ships on a Becalmed Sea
Intel announced a partnership with Elon Musk’s Terafab project, joining SpaceX, xAI and Tesla to develop a 1‑terawatt‑per‑year AI compute fab. The deal follows Intel’s $11.1 billion federal rescue, converting unspent grants into a 9.9% U.S. government equity stake. Musk’s ecosystem...
From SoC to System-in-Package: Transforming Automotive Compute with Multi-Die Integration
Automotive electronics are moving from monolithic system‑on‑chips to multi‑die system‑in‑package solutions to meet soaring compute, safety, and longevity demands. By stacking or side‑by‑side heterogeneous dies, manufacturers can combine CPUs, GPUs, AI accelerators and high‑bandwidth memory within a single package. This...
Equal1’s Silicon Qubits Gain Autonomous Calibration with Q-CTRL
Equal1 has partnered with quantum‑control specialist Q‑CTRL to embed its Boulder Opal Scale Up software into the company’s Bell‑series silicon qubit systems. The integration adds autonomous calibration, eliminating the need for manual, expert‑driven tuning of quantum hardware. This software‑driven autonomy...
Why Google’s TurboQuant Algorithm Is Disrupting the AI Memory Chip Market
Google’s TurboQuant algorithm compresses large language models by up to six times and accelerates processing as much as eightfold, delivering the same accuracy with far less memory. The technology combines PolarQuant and a Quantized Johnson‑Lindenstrauss routine to cut inference costs...
Asia Daily: April 8, 2026
Chinese researchers at Xidian University have developed a method to fabricate short‑wave infrared (SWIR) chips using standard silicon‑germanium CMOS processes, slashing production costs by up to 99% to roughly $10 per unit. The low‑cost chips retain military‑grade performance, opening civilian...
Intel Releases OpenVINO 2026.1 With Backend For Llama.cpp, New Hardware Support
Intel unveiled OpenVINO 2026.1, its latest quarterly update that expands generative AI capabilities across Intel’s hardware portfolio. The release adds official support for Wildcat Lake SoCs and the new Intel Arc Pro B70 32 GB GPU, while introducing Qwen3 VL on both CPU and...
AMD Medusa Point Leaked: 10 Cores, 32 MB L3 Cache, and First Geekbench Results for a Zen 6 APU
AMD’s engineering sample of the rumored Medusa Point APU surfaced on Geekbench, revealing a 10‑core, 20‑thread Zen 6 design with 32 MB of L3 cache. The sample runs at a 2.40 GHz base clock and peaked near 2.01 GHz, scoring 1,210 single‑core and 7,323...
Broadcom Secures Google’s Commitment Through 2031, and Anthropic Invests in the Next Generation of TPUs
Broadcom announced a multi‑year agreement with Google to develop and supply future generations of custom Tensor Processing Units and the networking components for AI racks through 2031. In parallel, Anthropic secured access to roughly 3.5 GW of Google‑based TPU capacity starting...
Onvo L90 to Adopt Nio's In-House Developed Shenji Smart Driving Chip, Report Says
Nio’s sub‑brand Onvo will launch the 2026 L90 electric SUV on April 21, featuring the company’s in‑house Shenji smart‑driving chip for the first time. The chip, fabricated on a 5 nm process, delivers up to 1,000 TOPS and is said to match the...
Rigetti Announces General Availability of 108-Qubit System
Rigetti Computing has made its 108‑qubit Cepheus‑1‑108Q system generally available through its Quantum Cloud Services platform and Amazon Braket. The modular chiplet‑based architecture stitches together twelve 9‑qubit chiplets, delivering a median two‑qubit gate fidelity of 99.1% with ~60 ns gate times...
Aehr Test Systems Q3 FY2026: The Burn-In Proxy for Silicon Photonics HVM
Aehr Test Systems reported Q3 FY2026 revenue of $10.3 million, a 44% year‑over‑year decline, while bookings surged to $37.2 million, yielding a book‑to‑bill ratio above 3.5x. The company disclosed that at least two silicon photonics customers are now ordering wafer‑level burn‑in (WLBI)...
UALink Consortium Publishes 4 Specifications Defining In-Network Compute, Chiplets, Manageability and 200G Performance
The UALink Consortium ratified four new specifications, including UALink Common Specification 2.0 with In‑Network Compute, a 200 Gbps data‑link and physical‑layer spec, a Manageability spec, and a Chiplet spec. These updates enable tighter compute‑communication integration, higher bandwidth, centralized control and chiplet‑level...
Google Controls the Most AI Computing Power, Driven by Its Custom TPUs
Google now controls about 25% of all AI compute sold since 2022, with roughly 75% of that capacity delivered by its custom‑designed Tensor Processing Units (TPUs). This makes Google the single largest owner of AI‑focused hardware, while other hyperscalers continue...
Intel Joins Elon Musk’s Fab Project
Intel announced on April 7 that it will serve as the manufacturing and packaging partner for Elon Musk’s Terafab project, a $20‑$25 billion semiconductor fab planned for Austin, Texas. The partnership positions Intel to help deliver 1 terawatt of compute power annually...
Ubuntu 26.04 Provides More Performance For AMD Ryzen AI Max "Strix Halo"
Ubuntu 26.04 brings significant performance gains to AMD’s Ryzen AI Max+ 395 “Strix Halo” APUs compared with the prior Ubuntu 25.04 release. The new distribution runs the Linux 7.0 kernel and GCC 15.2, delivering double‑digit CPU speedups and noticeable Radeon 8060S GPU improvements on the same Framework...
Broadcom Signs Expanded Deals with Google and Anthropic, Giving the AI Startup Access to 3.5 Gigawatts of Compute Capacity
Broadcom will manufacture future generations of Google’s AI chips and has expanded its agreement with Anthropic, granting the startup access to roughly 3.5 gigawatts of Google Tensor Processing Unit compute, up from 1 GW. Anthropic disclosed annualized revenue now exceeds $30 billion,...
Intel QAT Driver With Linux 7.1 Adding Zstd Offload Support
Intel’s QuickAssist (QAT) driver for the Linux 7.1 kernel now supports Zstandard (Zstd) offloading across Gen 4, Gen 5, and Gen 6 accelerators. Gen 4/5 offload compression only for buffers 8 KB‑512 KB, while Gen 6 offloads both compression and decompression with no size ceiling but falls...
[OFC 2026] Part 3 of 5: Packaging: Glass Substrates to 3D Hybrid Bonding
The OFC 2026 session highlighted that AI‑driven data centers are now limited by packaging rather than raw silicon speed. Co‑Packaged Optics (CPO) aims to overcome this bottleneck by integrating optical engines directly into switch packages, but the choice of substrate...
Intel Partners with Tesla and SpaceX on Terafab
Intel announced a partnership with Tesla, SpaceX, and xAI to create the Terafab joint venture, a massive vertically integrated semiconductor facility in Austin, Texas. The fab aims to deliver about 1 terawatt of AI compute per year, supporting billions of...
Air Liquide Launches Advanced Materials Plant in Taiwan – Just Now, the Chemistry Behind AI Chips Is Becoming a Strategic...
On March 25, 2026 Air Liquide inaugurated its first large‑scale advanced materials plant in Taichung, Taiwan, dedicated to deposition and etch chemicals essential for sub‑2 nm semiconductor nodes. The facility focuses on atomic‑layer‑deposition precursors needed for AI and high‑performance‑computing chips. The launch...
SK Hynix Links Its Record Order From ASML to Its Plans for a U.S. Stock Market Listing, Sending a Pretty...
SK hynix announced an 11.95 trillion‑won (≈ $7.97 bn) order for ASML EUV lithography tools, the largest publicly disclosed single ASML customer deal. The machines will equip the new Yongin semiconductor cluster and the M15X site in Cheongju, supporting HBM4 and advanced 1c‑nm DRAM...
Nvidia’s RTX 50 SUPER Refresh Continues to Face Delays, and Even the RTX 60 Could Be Pushed Back Further
Rumors in early 2026 indicate NVIDIA has postponed the mid‑cycle RTX 50 SUPER refresh originally expected at CES 2026, and the mass‑production timeline for the next‑generation RTX 60 series may shift from late 2027 to 2028. The delays stem from tight GDDR7 memory supplies, which...
Intel "Nova Lake" To Use Xe3 Graphics and Xe3P Display/Media Engine
Intel’s upcoming Nova Lake CPUs, part of the Core Ultra 400 series, will integrate Xe3 graphics and the Xe3P display/media engine rather than the newer Xe4 IP. The platform targets a late‑2026 launch with early‑2027 market availability and will natively support...
YieldWerx Delivers a Master Class in Co-Packaged Photonics Implementation
YieldWerx announced a forthcoming webinar that will teach semiconductor engineers how to implement co‑packaged photonics (CPO) across the full product lifecycle. The session, led by CEO/CTO Aftkhar Aslam, will detail the 12 cross‑domain challenges—from optical data complexity to test‑flow discontinuities—and...
[OFC 2026] Part 2 of 5: CPO and the AI Interconnect Challenge
AI compute clusters are doubling annually, pushing Meta's rack designs from 72 GPUs to over 256 nodes and exceeding 1 MW of power per rack. At these scales, copper backplanes face insurmountable limits in power delivery, bandwidth density, and routing complexity....
[OFC 2026] Part 2 of 5: CPO and the AI Interconnect Challenge
AI training racks are scaling rapidly, with Meta's upcoming ORW rack doubling node count and pushing single‑rack power beyond one megawatt. Copper backplanes now face insurmountable limits in bandwidth, power density, and routing complexity. Meta used OFC 2026 to set...
Intel Launches Core Ultra 7 251HX Arrow Lake Processors with 18 CPU Cores and 3 Xe GPU Cores
Intel quietly added the Core Ultra 7 251HX to its Arrow Lake HX family, slotting between the Core Ultra 5 245HX and the Core Ultra 7 255HX. The processor features 18 cores – six Performance and twelve Efficient – with a 30 MB Smart Cache and a...