Today's Semiconductors Pulse
AMD CEO says AI boom revives focus on CPUs
AMD CEO Lisa Su said the rapid expansion of AI workloads is refocusing attention on central processing units. AMD also announced it will invest more than $10 billion in Taiwan’s ecosystem to accelerate AI infrastructure deployment.
Applied Materials Boosts Its Outlook as AI Chipmakers Scramble to Add More Production Capacity
Applied Materials lifted its outlook for the current quarter, forecasting earnings of $3.36 per share and revenue of $8.95 billion, both above analyst estimates. The company posted Q2 revenue of $7.9 billion, an 11% year‑over‑year increase, and net income of $2.806 billion. Strong demand for AI‑focused chips is driving record sales in its semiconductor systems, DRAM, and services divisions. CEO Gary Dickerson said the firm expects overall equipment sales to grow more than 30% this year as AI compute builds out globally.
Mature Nodes, Missing Links And Niche Markets
China is accelerating its mature‑node semiconductor fab capacity, growing four times faster than global demand between 2015 and 2023. The country now accounts for roughly half of new mature‑node capacity projected over the next three to five years, squeezing Japanese...
AI Chip Maker Cerebras Systems Hits $95B Valuation After IPO, Giving Its Founder a 2nd Big Win
Cerebras Systems went public, pricing shares at $185 and closing at $311, which lifts its market cap to roughly $95 billion. The IPO sold 30 million shares, raising $5.55 billion, making it the biggest U.S. tech listing since Uber in 2019. The company’s...
Cerebras Risked It All on Dinner Plate-Sized AI Accelerators a Decade Ago. Today It’s Worth $66 Billion
Cerebras Systems completed a blockbuster IPO on Thursday, raising $5.55 billion and debuting with a market valuation exceeding $66 billion. The company’s wafer‑scale engines—giant chips the size of a dinner plate—have evolved from the first‑gen WSE to the current WSE‑3, delivering up...
Cerebras Stock Nearly Doubles on Day One as AI Chipmaker Hits $100 Billion — What It Means for AI Infrastructure
Cerebras Systems went public on Nasdaq at $350 per share, nearly double its $185 IPO price, instantly achieving a $100 billion market value. The company sold 30 million shares, raising $5.55 billion, the largest U.S. tech IPO since Uber in 2019. Its wafer‑scale...
US Clears H200 Chip Sales To 10 China Firms
U.S. Commerce Department has granted export licenses to roughly ten Chinese companies—including Alibaba, Tencent, ByteDance and JD.com—to purchase Nvidia's H200 AI chip, the company's second‑most powerful accelerator. Each approved buyer may acquire up to 75,000 units, but no shipments have...
Cerebras Opens at $385. IPO Price Was $185 Raising $5.5B at the IPO Price
Cerebras Systems debuted on Nasdaq at $385 per share, more than double its $185 IPO price, delivering a 108% first‑day gain. The offering raised $5.55 billion, valuing the AI‑chip maker at over $100 billion on a fully diluted basis. Cerebras’ wafer‑scale engine,...
Cerebras CTO Explains Building the World's Largest Chip
was watching a bunch of Cerebras YouTube videos and CTO Sean Lie does a solid explainer here on how it builds its wafer-scale chip aka “world’s largest chip” https://t.co/yg5Phe1t92
Cerebras Files for $48 B IPO, Upsizing to $4.8 B Offering on Wafer‑Scale AI Chip
Cerebras Systems filed to go public with a $150‑$160 per‑share price range, valuing the wafer‑scale AI chip maker at about $48 billion and targeting a $4.8 billion raise. The company expanded the offering to 30 million shares after securing Amazon Web Services and...
Nvidia Hits $5.5 Trillion Valuation With Zero Revenue From Its Former Biggest Growth Market: Could Huang Trip to China Accelerate...
Nvidia’s market capitalization surged to $5.5 trillion, eclipsing Apple and Google as the world’s most valuable public company. The jump reflects exploding demand for its AI‑focused GPUs, yet the firm reported zero revenue from China, its former biggest growth market. Analysts...
Power Corner: IDEAL’s SuperQ Tapping Si’s Potential Beyond Superjunction
iDEAL Semiconductor unveiled SuperQ, a silicon MOSFET architecture that replaces traditional superjunction P‑N pillars with an ultra‑thin high‑K dielectric in high‑aspect‑ratio trenches. This charge‑compensation redesign shrinks the blocking region to as little as 5% of the die, freeing silicon for...
Rethinking Superjunction’s Charge-Compensation Architecture with SuperQ
iDEAL Semiconductor’s SuperQ technology replaces the traditional p‑type pillars in superjunction MOSFETs with a nanometer‑thin high‑k dielectric film, reclaiming up to 95% of the silicon cross‑section for conduction. This asymmetric charge‑compensation architecture reduces specific on‑resistance while preserving high blocking voltages...
ADP: DRAM Shortage Is Affecting Ransomware Recovery
ADP’s chief strategy officer Stacy Hayes warns that the global DRAM shortage is crippling ransomware recovery more than backup. AI workloads are gobbling up memory, driving up hardware costs and lengthening lead times for the infrastructure MSPs rely on. As...
Chinese Companies Are Ramping up Homegrown AI Chips, Even if Nvidia Is Coming Back
Chinese tech giants Tencent and Alibaba say domestic AI chip supply will surge this year as they expand capital spending and integrate home‑grown GPUs into their data centers. The push follows a year‑long U.S. export ban that sidelined Nvidia, prompting...
T-Global Partners with SiPearl to Launch Taiwan–France Advanced R&D Initiative
T-Global Technology has teamed with French fabless CPU designer SiPearl in a two‑year joint R&D effort to create high‑thermal‑conductivity materials and two‑phase liquid‑cooling modules for high‑performance computing chips. The project, approved under Taiwan’s Ministry of Economic Affairs “A+ Driving Industrial...
Diodes Launches PCIe 7.0 Clock Generator for AI Servers
Diodes Incorporated introduced the PI6CG33A06, a six‑output ultra‑low jitter clock generator built for PCI Express 7.0 platforms in AI servers and high‑performance data‑center equipment. The device delivers RMS jitter below 30 femtoseconds, comfortably beating the PCIe 7.0 maximum of 67 fs, and supplies stable 25 MHz...
Tower Signs Customer Contracts for $1.3bn Silicon Photonics Revenue for 2027
Tower Semiconductor announced $1.3 bn of silicon photonics contracts for 2027, backed by $290 m in prepaid capacity reservations. The deals cover more than 50 active customers and set the stage for a larger wafer commitment in 2028. Tower is ramping multi‑fab...
Chiplets Need A New Workflow
Chiplet architectures are reshaping semiconductor development from single‑die design to a system‑level, multi‑die challenge. Engineers must adopt coordinated workflows that integrate design, packaging, verification, test, and reliability early to avoid costly failures before tape‑out. Multi‑physics analysis—thermal, mechanical, power, and signal...
Flash Getting Stacked High-Bandwidth Version
Sandisk, in partnership with SK Hynix, has unveiled a high‑bandwidth flash (HBF) memory stack that mirrors the footprint of HBM but delivers far greater capacity—up to 3 TB per 16‑die stack—and 1.6 TB/s read bandwidth. The NAND‑based architecture targets AI inference workloads, keeping...
AMD Adds 3D V-Cache to Ryzen Pro 9000 CPUs, Boosting Workstation Performance
AMD unveiled its Ryzen Pro 9000 series, equipping select workstation CPUs with 3D V-Cache and raising power limits to 170 W. The top model, the Ryzen 9 Pro 9965X3D, offers 16 Zen 5 cores, 128 MB of stacked L3 cache and a 5.5 GHz boost, targeting real‑time...
Rigaku Working with Imec to Accelerate Next-Gen Semiconductor Metrology Development
Rigaku Corp. has launched a three‑year collaboration with European research hub imec to fast‑track X‑ray‑based metrology for next‑generation semiconductors. The partnership will develop 3D device metrology, ultra‑sensitive ultrathin‑film detection, and non‑destructive defect inspection to support emerging architectures such as Gate‑All‑Around...
MRAM Gets Its Own SIG
Magnetoresistive RAM (MRAM) has reached a maturity level that prompted the creation of the MRAM Alliance Special Interest Group (SIG) within the Storage Networking Industry Association. The SIG brings together foundries, chip makers, memory vendors, equipment suppliers and end‑users to...
SMIC Wins Shanghai Stock Exchange Approval for $5.9 B Wafer Foundry Takeover
Semiconductor Manufacturing International Corp (SMIC) obtained clearance from the Shanghai Stock Exchange’s M&A Review Committee to purchase the remaining 49% stake in SMIC North for RMB 40.6 billion ($5.9 billion). The deal makes SMIC North a wholly owned subsidiary and marks the biggest...
EU Launches €50 Million SUPREME Quantum Manufacturing Project in Helsinki
The European Union has inaugurated the SUPREME quantum manufacturing project in Helsinki, allocating €50 million ($56 million) to expand high‑quality superconducting cat‑qubit chip production. The effort aims to shore up Europe’s quantum hardware supply chain and give emerging firms like Alice & Bob...
UK AI Chip Startup Fractile Raises $220M to Tackle the Growing Inference Bottleneck
UK AI‑chip startup Fractile announced a $220 million Series B round to accelerate its next‑generation inference hardware. The funding, led by Accel, Factorial Funds and Founders Fund, backs Fractile’s effort to cut the time and cost of producing massive token outputs for...
Flyback ICs Extend Operating Power to 440 W
Power Integrations has launched the TOPSwitchGaN family, extending flyback converter capability to 440 W. The new ICs combine PowiGaN switches with TOPSwitch architecture, delivering up to 92% efficiency and eliminating the need for heatsinks in many applications. Operating at up to...
IC-Link by Imec Joins TSMC 3DFabric® Alliance to Accelerate Advanced Packaging and 3D IC Innovation
IC‑Link by imec has become a member of TSMC’s 3DFabric Alliance within the Open Innovation Platform, giving it direct access to TSMC’s 2.5D/3D silicon stacking and advanced packaging technologies such as SoIC, CoWoS and InFO. The partnership expands IC‑Link’s ASIC...
D2D China: Are the GPU Sanctions Working?
The U.S. ban on GPU shipments to China has halved Nvidia's market share there, while Chinese firms accelerate domestic semiconductor development. Microsoft’s recent certification of Chinese GPUs and a surge in locally sourced wafer‑fab equipment signal a growing home‑grown ecosystem....
US Prosecutors Uncover $2.5 B Nvidia GPU Smuggling Ring to China and Russia
Federal investigators disclosed that a trio of suspects used encrypted WeChat messages to coordinate the illegal shipment of Nvidia A100, H100 and H200 GPUs to China and Russia. The scheme, valued at millions per order, highlights gaps in U.S. export...
NEO Semiconductor Demonstrates 3D X-DRAM Proof-of-Concept, Secures Strategic Investment to Advance AI Memory
NEO Semiconductor announced that its 3D X‑DRAM proof‑of‑concept chips meet key DRAM performance targets, including sub‑10 ns read/write latency and over 10¹⁴‑cycle endurance, using mature 3D NAND manufacturing lines. The results validate a new scaling path for high‑density, low‑cost AI memory....
Samsung Labor Talks Stall, Threat of 50,000‑worker Strike Could Halt AI Memory Chip Output
Samsung Electronics' wage negotiations with its labor union have broken down, prompting threats of a strike involving more than 50,000 workers from May 21 for up to 18 days. The dispute could cripple production of advanced AI memory chips, a...
Arm Preparing The Linux Kernel For 128-Bit Page Table Entries "FEAT_D128"
Arm announced an optional feature, FEAT_D128, for Armv9.3 and later that enables 128‑bit page table entries in the Linux kernel. The new translation system, VMSAv9‑128, expands both physical and virtual address spaces and frees bits for future MMU controls. Linux...
SpacemiT K3-Powered DC-ROMA RISC-V Motherboard III Is Made for the Framework Laptop 13
SpacemiT has launched the DC‑ROMA RISC‑V Mainboard III, an octa‑core K3 64‑bit SoC board that slots into the Framework Laptop 13. The module delivers up to 60 TOPS of AI performance, supports 16 GB or 32 GB LPDDR5 RAM and an optional 1 TB NVMe SSD....
Samsung Negotiates for an Agreement: Impending 18-Day Strike Could Affect AI Memory and HBM Supply Chains
Samsung Electronics is in mediated talks with its largest union over wages and profit‑linked bonuses, with an 18‑day strike slated to begin May 21, 2026 if no deal is reached. The dispute targets the chip division that posted roughly $100 billion in...
Tesla AI6.5 Reportedly Before Foundry Switch: Weibo Rumor Brings Intel Into Play Instead of TSMC
According to a Weibo leak cited by Wccftech, Tesla may consider moving production of its upcoming AI6.5 accelerator from TSMC to Intel under alleged U.S. political pressure. The claim is unverified; Tesla’s current AI5 chip is already fabricated by TSMC...
ASML Appoints Marco Pieters as CTO and Adds Him to Board of Management
ASML announced that Marco Pieters, a 25‑year veteran, has officially taken the role of Chief Technology Officer and joined the Board of Management after the company’s April 22, 2026 AGM. The move replaces retiring CTO Martin van den Brink and...
Maskless Laser Printing Enables 3D Carbon Circuits on Glass, Paving Way for Next‑Gen Chip Packaging
A global research team led by Prof. Seunghwoi Han (Chonnam National University) and Prof. Young‑Jin Kim (KAIST) demonstrated a mask‑free, femtosecond‑laser process that directly writes crystalline carbon circuits on glass substrates, promising cheaper, higher‑density interconnects for future semiconductor packages.
China’s AI Suppliers Can’t Keep Up as Component Shortages Bite
Chinese AI hardware vendors are confronting a severe component shortage that threatens to curb the rapid growth they enjoyed earlier this year. Demand for GPUs, ASICs and edge‑AI processors remains robust, but supply chain bottlenecks in semiconductors and advanced packaging...
Beijing Put Huawei’s Secret Chip Lab on National TV Two Days Before Trump Arrives. The Message Wasn’t for Chinese Viewers.
Beijing aired footage of Huawei’s secret chip laboratory on prime‑time CCTV just two days before President Trump’s state visit, turning the broadcast into a diplomatic signal. The Lianqiu Lake campus, a $1.4 billion complex that will host 35,000 researchers, showcases Huawei’s...
TSMC Arizona Fab Posts $514 M Profit Amid Water and Labour Hurdles
Taiwan Semiconductor Manufacturing Co. reported a NT$16.14 billion ($514 million) profit from its Arizona fab in 2025, the first full year of mass production. The gain comes as the company wrestles with water shortages, power constraints and a tight labor market, while...
Micron Shares Double Since March as Memory Chip Supercycle Gains Momentum
Micron Technology's stock has more than doubled since the end of March, climbing to about $742 and up 6.7% on the day, as AI‑driven demand and a global DRAM shortage spark a sector‑wide supercycle. The rally lifts the large‑cap semiconductor...
Micron Samples 256GB DDR5 RDIMM Built on 1-Gamma DRAM for AI Servers
Micron Technology announced sampling of a 256 GB DDR5 registered DIMM built on its 1‑gamma DRAM platform, capable of 9,200 MT/s—about 40% faster than current volume‑produced modules. The module uses 3D‑stacked dies connected by through‑silicon vias, delivering higher capacity and power efficiency,...
Supply Crunch Forces Japanese Chip Makers to Go Monochrome
Global supply chain disruption so bad they got the Japanese chip brands switching to black and white packaging to save on ink 😭
Ambiq AI Chips Power Tiny Devices, Stock Soars
Ambiq Micro @Ambiq_AI is up 31% today. Going to CES paid off. Here is its CTO/founder Scott Hanson talking about the AI chips it is making for small devices (rings, watches, glasses, medical devices). I got a tour of...
Beyond Tool Interoperability: The Emerging Governed Convergence Problem in Semiconductor Design
The semiconductor sector is moving past tool interoperability toward a "governed convergence" challenge, where specialized tools, domains, and organizations must align to produce manufacturable, high‑performance systems. Fragmentation now spans abstraction layers, physics domains, and decision authority, creating an "entropy wall"...
Tech Revolutions Start Slow, Explode Despite Early Doubt
One remarkable stat is when CUDA first launched in 2007, it was downloaded only 13,000x “Difficult to use” “Parallelism is not better” “High R&D cost” “Distraction from core customers” $NVDA tanks 90% But all technological revolutions look like this “Gradually then suddenly”
Nvidia’s Huang Wants to Sell Chips to China. Trump Has Other Priorities.
Nvidia CEO Jensen Huang will not join President Donald Trump on the administration’s upcoming trip to China, according to media reports. The omission suggests the White House is scaling back efforts to open the Chinese market for U.S. semiconductor firms....
Cerebras Targets $33B IPO as It Challenges Nvidia in AI Chips
Cerebras Systems announced an upsized IPO targeting $4.8 billion in proceeds at a roughly $33 billion valuation, with shares priced between $150 and $160. The chip maker, known for its wafer‑scale processor the size of a dinner plate, has secured a multi‑year...
Peraso's Q1 Revenue Plummets to $1M as Asian Supplier Delays Hit Shipments
Peraso Inc. reported first‑quarter 2026 net revenue of $1 million, a 66% decline from the prior quarter, after an Asia‑based supplier disrupted shipments of its millimeter‑wave and legacy memory chips. The company said it has now secured alternative sources and expects...
Cerebras IPO and the Four Bottlenecks in Its Custom-Everything Architecture
Cerebras Systems went public this week with an IPO that was oversubscribed by roughly 20‑times, underscoring strong investor enthusiasm for its wafer‑scale AI chips. The company’s flagship Wafer Scale Engine 3 (WSE‑3) integrates 900,000 cores, 44 GB of SRAM and 21 PB/s of...