Nvidia H200 Chips Still Not Sold to China...Chinese Chipmaker Horizon Robotic Launches 5-Nanometer chip...China’s Low Cost EVs to Be Fitted...
U.S. officials confirm Nvidia’s flagship H200 AI chips have not been shipped to Chinese firms, despite a Trump‑era export licence that remains stalled by regulatory disputes. Meanwhile, Chinese chipmaker Horizon Robotics unveiled a 5‑nanometer automotive processor, the Starry 6P, delivering 650 TOPS and 273 GB/s to challenge Nvidia and Qualcomm. China’s low‑cost EV makers are preparing to equip $8,800 models with lidar sensors, a technology once limited to premium cars, while DJI entered the heavy‑lift drone market with 200‑kg payload units priced around $14‑20k. These moves highlight a broader push by China to close technology gaps as U.S.–China tensions affect trade and travel, evident in a near‑50% cut in China‑Japan flights.
Intel OpenVINO 2026.1 Integrates llama.cpp with Wildcat Lake and Arc Pro B70—And Suddenly Makes Intel’s AI Strategy More Tangible
Intel’s OpenVINO 2026.1 toolkit adds a preview backend for llama.cpp and expands model support to include Qwen3 VL, GPT‑OSS 120B, and a WhisperPipeline for Node.js. The release explicitly targets Intel Core Series 3 (codenamed Wildcat Lake) CPUs and the Arc Pro B70 GPU with 32 GB of memory,...
RoboSense Releases Image-Grade SPAD-SoC Platform as LiDAR Competition Heats Up
Chinese LiDAR leader RoboSense unveiled its Eocene digital architecture and two flagship chips, Phoenix and Peacock, at a Tech Day event ahead of the Beijing Auto Show. The Phoenix chip is the world’s first automotive‑grade monolithic SPAD‑SoC, delivering a 2,160‑beam...
The End of the Neutral Stack: Choose Your Ecosystem
The automotive industry is abandoning the long‑standing neutral stack model as OEMs begin locking in compute and software platforms early in the vehicle development cycle. Simultaneously, data‑sovereignty regulations are forcing a split between China and non‑China vehicle architectures. Suppliers that...
JEDEC Previews LPDDR6 Roadmap Expanding LPDDR Into Data Centers and Processing-in-Memory
JEDEC previewed a next‑generation LPDDR6 roadmap that extends the mobile‑focused memory standard into data‑center and accelerated‑computing markets. The update adds a narrower per‑die interface, flexible metadata handling, and targets 512 GB densities per module. JEDEC is also developing a SOCAMM2 module...
Transitioning Voltage Regulator Design From Unidirectional To Bidirectional
The article by Nazzareno Rossetti outlines the shift from traditional unidirectional voltage regulators to bidirectional converters, a change driven by modern energy‑management systems that must move power between electric vehicles, photovoltaics, home batteries, and the grid. It highlights how synchronous...
How to Overcome the Advanced Node Physical Verification Bottleneck
Synopsys unveiled its IC Validator platform as a new answer to the mounting physical‑verification bottlenecks that accompany advanced‑node designs. The tool’s HyperSync distributed‑processing engine delivers 2‑4× faster full‑chip antenna and PERC ESD checks while slashing CPU usage by up to 70 %. Elastic...
Intel LLM-Scaler vllm-0.14.0-b8.2 Released With Official Arc Pro B70 Support
Intel released LLM‑Scaler vllm‑0.14.0‑b8.2, officially adding support for the Arc Pro B70 GPU. The update refreshes the Docker platform image to intel/llm-scaler-platform:26.18.8.2 and continues the Project Battlematrix push for multi‑GPU AI inference on Intel Arc hardware. The B70, a 32 GB VRAM card...
Raspberry Pi RP2350 Board Offers NB-IoT Cellular Connectivity, GNSS, and Wi-Fi Indoor Location
Challenger+ has released a Feather‑compatible RP2350 board that adds certified NB‑IoT and GNSS connectivity via the ST87M01 modem. The dual‑core RP2350 MCU runs at 150 MHz and includes 8 MB PSRAM, 8 MB flash, and a USB‑C power interface. Cellular performance reaches up...
The Last 1.5mm of AI Power: Three Numbers From Vicor’s Q1 2026 Earnings Call
Vicor Corp. disclosed its second‑generation Vertical Power Delivery (VPD) module on the Q1 2026 earnings call, highlighting a 3 A/mm² current density, up to 40× current multiplication, and a 1.5 mm package thickness. The company also announced a capacity boost at its...
Google Is in Talks with Marvell About Two New AI Chips—Putting Pressure on Broadcom
Google is reportedly in talks with Marvell to develop two custom AI chips—a Memory Processing Unit and a new TPU optimized for inference. The discussions come after Google signed a long‑term AI‑chip supply agreement with Broadcom that runs through 2031....
Onto Innovation Acquires Stake in Rigaku for $710M
Onto Innovation agreed to purchase a 27% stake in Japan’s Rigaku Holdings for $710 million, gaining a board seat while keeping Rigaku’s management independent. The investment builds on an existing collaboration that integrates Onto’s AI‑driven Ai Diffract software with Rigaku’s CD‑SAXS...
Why Indie Semiconductor (INDI) Just Made A Move Beyond Automotive
Indie Semiconductor announced on March 23 its first 399 nm ultraviolet‑visible distributed feedback (DFB) laser diode, a single‑frequency device tailored for quantum‑computing systems that use cooled ytterbium atoms. The UV laser adds to the company’s existing LXM‑U and narrow‑linewidth visible DFB lines,...
Is Intel About to Take Flight?
Intel announced a strategic partnership with Elon Musk to supply custom AI inference chips for Tesla, SpaceX and other ventures, leveraging its under‑utilized Hillsboro fab that already houses ASML EUV equipment. The deal offers Musk a queue‑free production line while...
China AI Circuit Board Firm Raises $2bn in Year’s Top HK Listing
Victory Giant Technology Huizhou, a Chinese maker of high‑end printed circuit boards for AI servers, raised HK$17.3 billion (about $2.2 billion) in Hong Kong, the largest IPO in the market this year. The stock jumped 59.6% to HK$335 ($43) on debut, reflecting...
Meta's Heterogeneous Fleet
Meta’s data‑center strategy now embraces a highly heterogeneous chip fleet, deploying five to six distinct hardware SKUs each year. While this diversity historically caused underutilized servers and software friction, the company has refined its processes to turn the complexity into...
Linux 7.1 Lands Workaround For Arm C1-Pro Erratum
Linux kernel version 7.1 now incorporates a mitigation for a hardware erratum in Arm’s C1‑Pro CPU affecting its Scalable Matrix Extension (SME). The bug, identified as erratum 4193714 and tracked as CVE‑2026‑0995, could allow TLB invalidation to complete before SME...
Initial Linux Driver Patches For Smart Data Accelerator Interface "SDXI"
Initial patches for the Smart Data Accelerator Interface (SDXI) have been posted to the Linux kernel mailing list, marking the first vendor‑neutral driver implementation for memory‑to‑memory data movement offload. The driver targets PCIe‑hosted SDXI 1.0 hardware and complies with the latest...
On Taiwan’s Breathtaking Order Boom
Taiwan’s Ministry of Economic Affairs reported a near‑70% year‑over‑year jump in AI‑hardware export orders for March, the strongest surge since a 72% rise in January 2010. Economists had only expected a 45% increase, underscoring the unexpected strength of global demand. Orders...
According to a Leak, Microsoft’s Next Surface Lineup Will Feature a New Division: Snapdragon X2 for Consumers, Panther Lake for...
A leak reported by VideoCardz suggests Microsoft will split its upcoming Surface refresh between Qualcomm Snapdragon X2 processors for consumer devices and Intel’s Panther Lake (Core Ultra Series 3) for business models. The rumor aligns with recent announcements: Windows 11 26H1 is optimized for Snapdragon X2,...
Intel Is Reportedly Reconsidering Its Desktop Sockets, but Raptor Lake, of All Things, Might Still Be in the Running
Intel is reportedly planning to lengthen desktop socket lifespans, aiming for AMD‑style multi‑generation support. An insider leak suggests a Raptor Lake refresh could extend the aging LGA1700 platform alongside the upcoming Arrow Lake refresh on LGA1851 and Nova Lake on...
AMD Bringing Back Ryzen 7 5800X3D to Save the World From DDR5
AMD announced a limited‑edition re‑release of the Ryzen 7 5800X3D, pricing it at $449 to mark the 10th anniversary of the AM4 socket. The move gives gamers a high‑performance, 3D‑V‑Cache CPU that can run on existing DDR4 platforms, sidestepping today’s steep DDR5...
Inside Nvidia’s Silicon Photonics Roadmap
Nvidia unveiled its next‑generation scale‑up system, NVL1152 “Kyber,” which will pack 1,152 GPUs—16 times the capacity of the current NVL72 rack. The machine, slated for the 2028 “Feynman” GPU generation, will combine copper interconnects with silicon photonics and co‑packaged optics...
Equal1’s Silicon Processors Power Kvantify’s Quantum Simulation Workloads
Equal1 and Kvantify have formed a partnership to bring silicon‑based quantum processors to life‑science workloads. Equal1’s Bell‑1 server, built on standard silicon, is being shipped as the company’s first‑generation quantum machine, and Kvantify has been named its preferred partner for...
Accelsius Announces General Availability of NeuCool IR150 and Launches NeuCool HyperStart Program
Accelsius announced the general availability of the NeuCool IR150, the first fully integrated 42U rack that combines a two‑phase coolant distribution unit with up to 150 kW of cooling capacity. The plug‑and‑play system is designed for hyperscale, neocloud and enterprise data...
WEBINAR: Intrinsic Techniques in RF Power Amplifier Design
Keysight’s second RF Power Amplifier Design master‑class webinar dives into intrinsic modeling techniques for GaN transistors, complementing traditional load‑pull methods. The session showcases a Class J PA built around an intrinsic‑node model, demonstrating how waveform engineering and Fourier‑based impedance mapping boost...
Analog Bits Demos Real-Time On-Chip Power Sensing and Delivery on N2P at the TSMC 2026 Technology Symposium
Analog Bits demonstrated a suite of real‑time on‑chip power‑sensing and delivery IP at the TSMC 2026 Technology Symposium, running on the company’s N2P test chip. The portfolio includes an on‑die LDO with 30% area reduction, droop detector, glitch catcher, ultra‑low‑power...
Taiwan Semi’s (TSM) Impressive Quarter Sets The Tone For 2026
Taiwan Semiconductor Manufacturing Co. (TSM) reported a stellar first‑quarter 2026, posting NT$1.13 trillion (≈$36.3 billion) in revenue, a 35.1% year‑over‑year increase. March alone saw NT$415.19 billion (≈$13.3 billion), up 45.2% YoY and 30.7% sequentially, underscoring robust demand for AI‑related chips. Wall Street analysts...
According to a Leak, Intel LGA1954 Will Feature 2L-ILM on High-End Motherboards: More than Just a New Lever on the...
Intel’s upcoming LGA1954 desktop socket may include an optional two‑lever Independent Loading Mechanism (2L‑ILM) on premium motherboards, according to a leak reported by VideoCardz and analyzed by Tom’s Hardware. The 2L‑ILM would replace the traditional single‑lever design, aiming for more...
Disaggregating LLM Inference: Inside the SambaNova Intel Heterogeneous Compute Blueprint
Intel and SambaNova announced a heterogeneous inference blueprint that assigns distinct phases of large‑language‑model serving to the most suitable hardware. GPUs accelerate the parallel prefill stage, while SambaNova’s Reconfigurable Dataflow Units (RDUs) handle the memory‑bound decode phase. Intel Xeon 6 CPUs...
Tesla Fab Chip Partners Are Critical for the Next Few Years
Tesla is turning to a set of chip partners to bridge shortages of CPUs, GPUs, and memory over the next two to four years. Intel will supply the central processing units, while Nvidia provides graphics processors for AI workloads. Advanced‑node...
DongshanPI K510 SoM Appears with Dual RISC-V Cores and AI Acceleration
DongshanPI unveiled a new System‑on‑Module built around the Kendryte K510, featuring dual 64‑bit RISC‑V cores clocked to 800 MHz and a dedicated RISC‑V DSP. The module adds a second‑generation KPU AI accelerator capable of 1.22 TOPS (INT8) or 1.22 TFLOPS (BF16), targeting computer‑vision and...
AMD Zen 6 “Medusa Point” Appears on Geekbench – 10 Cores, 32 MB L3 Cache, and the First Credible Hint...
A Geekbench entry for an engineering sample named “AMD Plum‑MDS1” reveals a 10‑core, 20‑thread Zen 6 mobile processor with 32 MB of L3 cache, the first concrete data point for AMD’s rumored Medusa Point platform. The sample runs at a modest 2.40 GHz base...
From Efficiency to Exhaustion: Rethinking the Limits of Taiwan AI Supply Chains
Taiwan’s semiconductor sector, anchored by Hsinchu Science Park, now drives over $800 billion in GDP and accounts for 18.5% of national output. While TSMC’s overall attrition fell to 3.5% in 2024, first‑year turnover remains high at 8.9%, reflecting long hours and...
Linux 7.1 Adds New AMD SMCA Bank Types, Presumably For Upcoming EPYC Venice
Linux 7.1’s kernel update expands the AMD Machine Check Exception (mce_amd) driver with ten new Scalable Machine Check Architecture (SMCA) bank types. The additions cover data‑acceleration front‑end and back‑end units, eDDR5 interconnects, and several secure‑processor components. While the patches do...
The 7 Taxes on AI
TSMC posted a record Q1 with $35.9 billion revenue, 40.6% YoY growth, and a 62.3% gross margin, while ASML reported €8.77 billion (~$9.6 billion) revenue, beating forecasts. Both companies lifted full‑year 2026 guidance, yet their stocks were rattled by unrelated macro headlines—a US...
Intel Bartlett Lake on Z790 via BIOS Mod: A Community Discovery Reveals the Line Between Socket Compatibility and Firmware Locks
A community modder successfully ran Intel’s Bartlett Lake‑S Core 9 273PQE, a 12‑core embedded processor, on an Asus Z790 consumer motherboard by applying a custom BIOS. The hack bypasses Intel’s firmware restrictions, proving that the socket and electrical design are compatible, but the...
Intel Officially Launches the Core Series 3: Wildcat Lake Brings 18A and AI Features to More Affordable Laptops
Intel launched the Core Series 3 (codenamed Wildcat Lake) on April 16, 2026, targeting affordable laptops, education, small‑business PCs and edge devices. The line uses the Core Ultra Series 3 architecture on Intel’s 18A process, offering up to 40 TOPS AI performance, Thunderbolt 4, Wi‑Fi 7 and...
AMD FP-DSS Security Bug For Zen 1 CPUs Made Public, Linux Kernel Patched
A floating‑point divider state sampling (FP‑DSS) vulnerability has been disclosed for AMD Zen 1 and Zen 1+ processors, including early Ryzen and EPYC chips. The flaw is a transient‑execution issue that could let a locally‑privileged attacker extract data through the floating‑point divisor...
You're Not Talking to Somebody Who Woke Up a Loser
Jensen Huang, CEO of Nvidia, engaged in a roughly 40‑minute debate with investor Dwarkesh Patel about whether the United States should continue exporting AI‑focused chips to China. Patel raised concerns that Chinese firms could use Nvidia hardware to train powerful...
Valve Developer Lands RADV/ACO Changes For AMD's GFX11.7 / RDNA 4m
Valve’s Linux graphics team, led by engineer Rhys Perry, merged ACO compiler and RADV Vulkan driver updates to support AMD’s upcoming GFX11.7 GPU, branded as RDNA 4m. The changes extend Mesa’s RadeonSI Gallium3D and RADV drivers to include new shader capabilities...
Silicon Storage Technology Assigned Four Patents
Silicon Storage Technology (SST) has been assigned four U.S. patents covering advanced non‑volatile memory (NVM) innovations. The patents include an adaptive bias decoder, a coarse‑and‑fine programming method, a planar split‑gate/FinFET integration process, and an output block with on‑chip ADC conversion....
Jensen Vs. Dwarkesh on China Chips
NVIDIA CEO Jensen Huang and analyst Dwarkesh Patel sparred over China’s access to advanced chips. Huang argued that retaining a U.S.‑centric technology stack and collaborative research would keep China dependent, while Patel warned that supplying the world’s best GPUs makes...
The Intel Arc Pro B65 Is Launching, and Intel Is Deliberately Positioning the Workstation Graphics Card Closer to the Gaming...
Intel introduced the Arc Pro B65 workstation GPU, a 32 GB GDDR6 card that blurs the line between professional and gaming graphics. The card ships with 20 Xe cores, 160 XMX engines, 608 GB/s bandwidth, a 200‑W TDP and PCIe 5.0 ×16 support. A WHQL driver...
Intel Core Series 3 “Wildcat Lake” Processor Family Launched for Entry-Level Laptops and Edge AI Systems
Intel officially launched the Core Series 3 “Wildcat Lake” processor family, its first hybrid AI‑ready Core line aimed at entry‑level laptops and edge AI devices. The SoC combines two performance cores with four efficiency cores, up to 2‑core Xe3 graphics, up...
Samsung Foundry VP Joins Intel Foundry
Samsung Foundry executive vice president Shawn (Seung Hoon) Han will join Intel Foundry Services as senior vice president and general manager in May, succeeding Kevin O’Buckley. Han will report to Intel’s EVP Naga Chandrasekaran and brings three decades of Samsung...
Wellcome Leap Announces $2M Prize in $50M Quantum for Bio Challenge Program
Wellcome Leap announced that Algorithmiq earned the $2 million prize in its $50 million Quantum for Bio (Q4Bio) Challenge, marking the first end‑to‑end quantum‑classical workflow that simulates a photosensitizer drug for photodynamic cancer therapy. The program, launched in 2023, devoted $40 million to...
The Latest News In IC Packaging & Test
The semiconductor packaging and test ecosystem saw a wave of strategic investments and acquisitions in early 2026. Morocco’s RifSol Corp. announced a $1.7 billion 200 mm fab aimed at automotive‑grade legacy nodes, while India’s Kaynes Semicon opened an OSAT campus in Gujarat...
Our First Ryzen AI 5 430 Benchmarks Are in and They're a Mixed Bag
AMD’s new Ryzen AI 5 430, the latest mid‑range Zen 5 mobile processor, replaces the 2025 Ryzen AI 5 330 in budget‑to‑mid‑range laptops such as the Lenovo IdeaPad 5 2‑in‑1 15. Benchmarks show the 430 delivers up to 30% faster single‑threaded performance and a 1.5‑2× jump in integrated graphics speed...
WEBINAR: Beyond Moore’s Law and The Future of Semiconductor Manufacturing Intelligence
The SemiWiki webinar on April 23 2026 gathered leading AI, EDA and fab executives to discuss how semiconductor manufacturing is moving beyond Moore’s Law. Speakers highlighted AI‑driven digital twins, predictive metrology, agentic AI and generative design as tools to tame the complexity...