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.
Sanmina’s Revenue Surges 102% as Accelerated Compute Drives Cloud‑AI Sales
Sanmina Corp. posted a 102% year‑over‑year revenue increase to $4.01 billion, driven by a surge in accelerated‑compute shipments for cloud and AI infrastructure. The earnings call highlighted component shortages that could temper growth, but the company sees the trend as a durable shift in enterprise buying patterns.
Amazon Signals Plan to Sell Trainium AI Chips to Outside Customers, Targeting $50B Market
Amazon announced that its custom Trainium AI chips, backed by $225 billion in revenue commitments, could be offered to non‑AWS customers, potentially expanding the chip line to a $50 billion run rate. The shift marks a strategic pivot from cloud‑only acceleration to...
Nvidia Hits New All‑Time High as Market Value Briefly Tops $5 Trillion
Nvidia's shares surged to a fresh all‑time high, lifting its market capitalization past the $5 trillion mark for the first time since November. The rally reflects renewed AI‑related buying, buoyed by strong demand for its chips and supportive moves from memory‑chip...
Wafer Dicing Service | Silicon Wafer Dicing & Die Singulation
Wafer dicing separates processed wafers into individual dies, a step that directly influences yield and downstream packaging reliability. AnySilicon offers a matchmaking platform that connects fabless designers, MEMS developers, and research labs with qualified dicing service providers across a broad...
Google Tensor SoC Thread
Google’s Tensor G3 custom SoC powers the Pixel 8 and Pixel 8 Pro, featuring a Cortex‑X3 prime core, Cortex‑A715 and A510 efficiency cores on Samsung’s 4LPP node, and a Mali G715 MP7 GPU. Leaked data suggests three packaging variants—iPoP, fan‑out panel‑level (Fo‑PLP), and fan‑out wafer‑level...
Power Corner: Renesas’s Grid-to-Core Strategy for 800 V Data Center Power Architecture
Renesas is promoting a grid‑to‑core power architecture that moves data‑center power delivery from traditional side‑car racks to a unified 800 V DC bus using solid‑state transformers. The approach combines GaN‑based Vienna rectifiers, an LLC DCX isolated DC‑DC platform, and matrix transformers...
As Much As Needed, As Little As Possible: Jensen Huang on the Operating Principle Behind Nvidia’s Run
Nvidia CEO Jensen Huang repeatedly emphasized an eight‑word operating principle: “as much as needed, as little as possible.” The mantra guides Nvidia to own only the core AI stack—CUDA, NVLink, compilers—and to refuse non‑essential ventures like its own cloud or...
Prices of Nvidia’s B300 Server at 7 Million Yuan in China on US Curbs: Sources
Strong demand for AI compute in China has pushed Nvidia's B300 server price to about ¥7 million (≈$980,000), nearly double the 2025 level. The surge follows tighter U.S. export restrictions and a crackdown on grey‑market smuggling, creating a scarcity premium. Chinese...
Qualcomm, Samsung, MediaTek Expect Hit in Handset Revenues as Component Shortages Bite OEMs
Chipmakers Qualcomm, Samsung and MediaTek warned that memory‑module shortages, driven by AI data‑center demand, will depress smartphone revenues in 2026. Qualcomm expects Q2 handset revenue to fall from $6 billion to about $4.9 billion, with a bottom in Q3 before a modest...
MediaTek Sees Nearly 10% Annual Revenue Growth Despite Likely Q2 Setback
Taiwan‑based fabless chipmaker MediaTek expects second‑quarter 2026 revenue to be flat or fall up to 6%, ranging between NT$140.2 billion and NT$149.2 billion (approximately $4.43‑$4.77 billion). Despite the near‑term dip, the company projects an almost 10% rise in full‑year revenue, driven by a...
Samsung Electronics Q1 Profit Jumps Fivefold on AI‑driven Memory Chip Demand
Samsung Electronics reported a 750% jump in first‑quarter operating profit to $44 bn, driven by an 86% surge in semiconductor sales as AI‑related memory chips fuel demand. The earnings beat underscores the rapid scaling of AI‑centric manufacturing and its impact on...
Qualcomm Beats Q2 Forecast, Announces Major AI Chip Deal with Leading Hyperscaler
Qualcomm reported Q2 2026 revenue of $10.6 billion and non‑GAAP EPS of $2.65, topping its own guidance. The company also disclosed its first custom AI‑chip shipment to a leading hyperscaler, slated for December, marking a strategic entry into data‑center silicon. The...
Nvidia's Madison Huang Accelerates AI‑Robotics Ties with Samsung, SK Hynix, LG and Doosan
Nvidia senior director Madison Huang met Samsung Electronics, SK hynix, LG Electronics and Doosan Robotics in Seoul, signaling a coordinated push to create a physical AI platform that could deliver industrial humanoid robots by 2028. The talks cover robotics hardware,...
AOI Awarded $20.9m Texas Semiconductor Innovation Fund Grant
Applied Optoelectronics Inc (AOI) has secured a $20.9 million grant from the Texas Semiconductor Innovation Fund to fund a 210,000 ft² manufacturing expansion in Sugar Land, Texas. The new facility will become one of the nation’s largest production sites for AI‑focused data‑center...
Google's Tensor G6 Rumors Nearly Had Me Hyped, but It Might Let Me Down
Leaked specifications suggest Google’s upcoming Tensor G6 SoC will adopt a 1+4+2 core configuration, featuring a 4.11 GHz C1‑Ultra prime core and a mix of C1‑Pro cores. The rumor also claims the chip will use the five‑year‑old PowerVR C‑Series CXTP‑48‑1536 GPU,...
CPPC V4 Support Being Worked On NVIDIA For The Linux ACPI Driver
NVIDIA engineers are adding Collaborative Processor Performance Control version 4 (CPPC v4) support to the Linux ACPI driver, extending the _CPC package from 23 to 25 entries. The new optional fields—OSPM Nominal Performance and Resource Priority—allow the operating system to signal nominal...
Paragraf & Archer Materials Target Quantum Computing With Graphene
Paragraf, a UK graphene‑electronics specialist, has teamed with Australia’s Archer Materials to create graphene‑based structures for qubit detection. The partnership combines Paragraf’s wafer‑scale graphene deposition process with Archer’s quantum‑device expertise, aiming to move quickly from research to functional prototypes. By...
Nexperia China Says Operations Stable After Parent Risk Warning
Nexperia China announced that its manufacturing and financial operations in China remain stable despite a risk warning issued to its parent, Wingtech Technology, after the latter’s latest financial disclosure. The company emphasized that the warning stems from external factors affecting...
Tower Semiconductor and Axiro Push High-Efficiency SiGe for Next-Gen Radar
Tower Semiconductor and Axiro Semiconductor have introduced a new family of silicon‑germanium (SiGe) beamforming ICs for Ku‑ and X‑band radar and satellite communications. The devices, now moving to volume production in Tower’s U.S. fabs, promise higher gain, linearity, output power...
ISM 2.0 Must Support Ecosystem Development, Says American Synopsys
Synopsys senior vice‑president Prith Banerjee urged the Indian government to treat the upcoming ISM 2.0 programme as an ecosystem initiative rather than a collection of isolated incentives. The centre plans to allocate roughly ₹1,000 crore (about $120 million) for FY27, building on the...
Creating Agentic EDA Methodologies
The article examines the emerging push for agentic AI methodologies in electronic design automation (EDA), emphasizing the need for AI to operate across diverse data formats and abstraction levels. It highlights the scarcity of architectural‑level tools and the lack of...
Foundry Capacity Is Limiting Who Competes At Leading Edge Nodes
Leading‑edge semiconductor capacity at 2nm and below is being monopolized by hyperscale customers such as Apple, Nvidia and Broadcom, leaving smaller chip developers with limited wafer access. As a result, firms are turning to advanced packaging and chiplet architectures to...
Unlocking High-Speed Serial Link Signal Integrity With AMI Model
High‑speed interfaces such as PCIe Gen5, USB4, and DDR5 demand rigorous signal‑integrity analysis, but traditional SPICE simulations are too slow for millions of bits. The Algorithmic Modeling Interface (AMI) offers a standardized .ami/.dll approach that embeds FFE and DFE equalization, delivering...
Facilitating Complex SoC Design Through Automation And Integration
The article outlines how Arteris tackles soaring system‑on‑chip (SoC) complexity with a unified automation suite. Magillem Registers creates a single source of truth for hardware‑software interfaces, while Magillem Packaging standardizes IP intent across vendors. FlexGen then auto‑generates network‑on‑chip (NoC) architectures...
Solving Clock Signal Integrity And Jitter Issues
A new blog highlights the growing problem of clock signal integrity and jitter in deep‑submicron chips, where power‑network noise can cause metastability and reduced Fmax. Traditional simulations are slow and often miss subtle timing violations across large clock networks. Synopsys...
From Standards To Systems: The Chiplet Era On Arm
Arm is transitioning from monolithic system‑on‑chips to multi‑die chiplet architectures, backed by the Arm Chiplet Specification Architecture (ACSA) and the OCP Foundation Chiplet System Architecture. While traditional SoCs hit reticle and power limits, chiplets offer modular scaling, better yields, and...
AI Demand Drives Long-Term Memory Chip Contracts, Stabilizing Industry
SK Hynix and Samsung Electronics are seeing a shift in demand patterns. AI-driven shortages are pushing customers toward long-term contracts, signaling a move away from the traditional boom-and-bust cycles in memory chips. The change could be structural. AI demand may bring more...
Intel 18A-P Node Brings Efficiency Gains Without Density Scaling Shift
Intel will unveil the 18A‑P node at the VLSI 2026 Symposium, an optimized spin on its 18A process that targets power and performance gains without altering transistor density. The new variant promises roughly 9% higher performance at the same power envelope...
Intel and SoftBank Explore HB3DM Memory with Higher Bandwidth Than HBM4
Intel and SoftBank’s joint venture Saimemory announced the HB3DM memory architecture, a nine‑layer stack that uses hybrid bonding and roughly 13,700 TSVs per layer to push bandwidth density to about 0.25 Tb/s per mm², delivering an estimated 5.3 TB/s per 171 mm² module....
Mosaic SoC Secures Funding for Spatial Wearable Push
Swiss semiconductor startup Mosaic SoC announced a $3.8 million pre‑seed round led by Founderful and the Kick Foundation. The funding will accelerate development of a proprietary multi‑core chip designed to handle visual and positional sensor data without relying on power‑hungry GPUs....
Samsung Sees No Respite as Memory Shortage Set to Worsen
Samsung Electronics reported a record quarterly profit in its memory division, with operating profit soaring to $36.2 billion – a 49‑fold increase year‑over‑year. The company warned that AI‑driven demand will widen the memory supply gap through 2027, even as it signs...
Harvard Team Achieves Milliwatt UV Light Generation On a Photonic Chip
Harvard researchers have built a chip‑scale ultraviolet light source on thin‑film lithium niobate that delivers 4.2 mW of on‑chip power at 390 nm, roughly 120 times more than prior demonstrations on the same platform. The device uses a frequency‑up‑conversion process that merges two...
Google to Sell Its TPUs to some Customers, Who Also Fancy Big-G GPUs
Alphabet announced that Google Cloud will begin selling its custom Tensor Processing Units (TPUs) to a select group of customers for on‑premise deployment. The move follows strong demand from AI labs, capital‑markets firms, and high‑performance computing users, and complements growing...
TSMC's AI‑driven Pricing Power Fuels Further Upside
TSMC keeps hitting all-time highs, but the AI boom and pricing power suggest there’s more room to run. Geopolitical risk is real, but their tech moat is massive. Semiconductors
AMD: The Overlooked AI Winner You Should've Bought More
In hindsight, AMD has been an obvious play over the past few months. The biggest tech companies clearly signalled in their earnings reports that AI spending was ramping up fast, and we knew they were buying directly from AMD. I even posted...
Iran Crisis a Wake-Up Call for India to Push Indigenous Semiconductor Design, Says L&T's Sanjay Gupta
The Iran conflict has highlighted supply‑chain fragility, prompting L&T Semiconductor’s Sanjay Gupta to call for a faster push toward indigenous chip design in India. Gupta says the short‑term impact on L&T is minimal, but the crisis underscores the need for...
AI Hardware Roadshow Highlights 2025 Milestones, 2026 Vision
At @Synopsys Converge this year, @Sallywf and I were invited to present our AI Hardware Show on the road (aka, the Roadshow) as an amuse bouche ahead of the CEO keynote. We discussed key milestones from 2025, and what we're...
DRAM Oligopoly Masks Price Hikes, Not Volume Growth
The myth of rising memory-chip sales. Oligopoly in action: Samsung, SK Hynix, and Micron aren’t shipping more DRAM they’re just raising prices. https://t.co/QwV7EMKEmy https://t.co/zUVgU49n6t
Polyn Achieves Successful Tapeout of Automotive Chip
Polyn Technology announced the tape‑out of its first automotive chip, VibroSense, which incorporates an analog neuromorphic core for real‑time tire‑road friction monitoring. Developed with GlobalFoundries, the chip processes high‑frequency vibration data to estimate the peak friction coefficient before braking events....
Xizhi Tech Raises HKD2.5 Bn in Hong Kong IPO, Shares Surge 380% on Debut
Xizhi Technology, the world’s first listed AI silicon‑photonic chip maker, closed its Hong Kong IPO with HKD2.53 bn ($323 m) of proceeds. The stock opened at HKD880, 380% above the HKD183.2 offer, giving the company a market value north of HKD80 bn ($10.2 bn)....
TSMC Ramping Up Spending Plans to Meet AI Chip Demand Surge as Rivals Narrow Technology Gap
TSMC announced a near $56 billion capex plan for 2026 to expand capacity for AI‑driven chips, including three new 3‑nm fabs in Taiwan, the United States and Japan that will start production between 2027 and 2028. For the first time the...
Mobileye Posts Weak Earnings but Cites Strong ADAS Hardware Growth Backed by Intel Chips
Mobileye reported earnings that fell short of market expectations, but the company highlighted robust growth in its advanced driver‑assistance systems (ADAS) business. Leveraging a horizontal, Intel‑backed chip‑supplier model, Mobileye says its hardware is now embedded in a growing share of...
Amazon the Chip Company? Tech Giant Says It May Sell AI Chips as a Product, Not Just a Cloud Service
Amazon announced it could begin selling its Trainium AI chips as physical racks to external customers within the next two years, marking a shift from a cloud‑only offering to a product business. The company disclosed $225 billion in revenue commitments for...
Memory Chip Shortage
Samsung reported a first‑quarter 2026 profit surge powered by its memory‑chip division, but warned that supply lags far behind soaring demand and could deteriorate in 2027. Vistance Networks estimates the shortage will shave roughly $30 million off its earnings versus last...
Cohu Inc (COHU) Q1 2026 Earnings Call Transcript
Cohu Inc. reported a 57% year‑over‑year surge in orders, propelled by AI and high‑performance computing (HPC) demand, and highlighted a $750 million pipeline targeting test handlers and HBM inspection. Non‑GAAP revenue reached $125.1 million with a 46.5% gross margin, while recurring revenue...
Chip Gear Supplier KLA Posts Beat-And-Raise Fiscal Q3, But Stock Falls Late
KLA Corp. posted fiscal Q3 earnings of $9.40 per share on $3.42 billion revenue, topping FactSet expectations. Year‑over‑year results rose 12% in earnings and 11% in sales, and the company lifted its FY2026 outlook to $9.87 EPS and $3.58 billion sales. It...
QCOM Beats Modest Forecast; Data‑Center Chips Drive Future
$QCOM expectations were modest this week quarter and it delivered. I stand by my assessment that what will move the needle for the company will be its data center entry with a portfolio of XPU, CPU, Custom, and Networking chips. The...
Qualcomm Announces Entry Into Custom ASIC Market
$QCOM dropped this new nugget on the call. They are officially in the custom ASIC game. Rumors had swirled they were engaged with Google (speculative, but it's likely Google on TPUs). I expect more at FA day...
QCOM's Custom XPU/CPU Business Finally Gains Market Recognition
$QCOM has a custom XPU/CPU business. And the market is finally catching up to this. Some of us tried to warn you. 🙃
Semiconductors Now Exceed Consumer Discretionary in Market Cap
Semis represented more than 10% of global market cap as of Monday... that industry is larger than the Consumer Discretionary sector UBS via @augurinfinity https://t.co/53A9gOEPr5 https://t.co/mUSyBhDNrN