Today's Semiconductors Pulse
SMIC clears final hurdle for $5.97B takeover of SMNC
China’s Semiconductor Manufacturing International Corp (SMIC) secured regulatory approval to acquire the remaining 49 % of its Beijing‑based unit SMNC, valuing the stake at 40.6 billion yuan (≈$5.97 bn). The transaction will be funded by issuing 547.2 million A‑shares at ¥74.20 ($10.91) each, with a 12‑month lock‑up for sellers, making SMNC a wholly‑owned subsidiary.
From Simulation Checkpoints To Continuous Physics
Semiconductor teams have relied on iterative, checkpoint‑based simulation, but growing design complexity is exposing its limits. Advanced packaging now demands physics insight that adapts instantly to geometry, material, and load changes. Continuous physics reasoning injects solver‑grounded analysis directly into the design loop, delivering early thermal, stress and warpage feedback. Determinism remains essential, ensuring repeatable, trustworthy results for downstream validation.
AMD Ryzen AI Max+ PRO 495 Spotted With 192GB Memory and New iGPU
AMD’s upcoming Ryzen AI Max+ PRO 495 APU surfaced in a PassMark database, indicating early benchmark testing. The chip packs 16 CPU cores, 32 threads and a Radeon 8065S integrated GPU, an upgrade from the prior 8060S. It also supports a...
Advantech Adds Intel Core Series 3 to Edge AI Systems
Advantech announced it will embed Intel’s Core Series 3 processors into its industrial embedded boards and edge AI systems, beginning with a lineup launching in April 2026. The hybrid six‑core silicon combines performance and efficient cores, an Xe3 GPU and Intel NPU 5.0,...
AMD’s Ryzen AI Halo Box Is Set to Launch in June – Strix Halo Apparently Gets Its Own Developer Enclosure
AMD is preparing a dedicated developer system called the Ryzen AI Halo Box, slated for a June 2026 release. The box will be built around the Ryzen AI Max+ 395 APU, which packs 16 Zen 5 cores, a Radeon 8060S iGPU with 40 RDNA 3.5 compute units, an XDNA 2...
Logic Scaling Boosts Efficiency, Yet Costs Rise Sharply
Scaling logic continues to deliver better performance per watt, but it’s becoming harder, more expensive, and increasingly customized. https://t.co/MzILzSSFrN #2nm #semiconductor
JEDEC Releases New Memory Interface Logic and Expanded MRDIMM Roadmap
JEDEC’s JC-40 and JC-45 committees announced the release of the DDR5MDB02 multiplexed rank data buffer standard and signaled an imminent DDR5MRCD02 clock‑driver specification. The groups are also finalizing the MRDIMM Gen2 module standard and prototyping raw‑card designs that reach 12,800 MT/s....
AMD Leaks First 16‑Core Ryzen 9 PRO 9965X3D with 3D V‑Cache
AMD has leaked specifications for the Ryzen 9 PRO 9965X3D, its first 16‑core processor in the PRO portfolio and the inaugural PRO chip to carry 3D V‑Cache. Built on the Zen 5 architecture, the chip targets professional, content‑creation and AI workloads while promising...
Nvidia Adds Taiwan's JPC as Second Certified Supplier for Vera Rubin AI Power Cables
Nvidia confirmed that Taiwan‑based JPC Connectivity has become its second certified supplier of power cables for the Vera Rubin AI supercomputer platform. The move reflects growing demand for high‑power components as AI workloads surge.
BrainChip Strikes IP Licence Deal with ASICLAND to Integrate Akida AI Technology
BrainChip Holdings (ASX:BRN) has signed a non‑exclusive, worldwide IP licence agreement with ASICLAND to embed its Akida neuromorphic AI technology into ASICLAND’s system‑on‑chip designs. The deal provides upfront evaluation and production licence fees per customer plus volume‑based royalties on net...
From AI Servers to High-Speed PCBs: Why CCL Prices Are Entering a New Upcycle
Copper‑clad laminate (CCL) prices are entering a new upcycle as high‑frequency, AI‑driven servers, 5G infrastructure, and electric vehicles demand faster, thinner, and greener boards. Environmental mandates for lead‑free and halogen‑free formulations add complexity, while tighter dimensional tolerances push material costs...
Jefferies: AI Capex to Consume 92% of Hyperscalers' Cash Flow, Powering DRAM Surge
Jefferies projects that the four biggest U.S. hyperscalers will allocate 92% of their operating cash flow to capital spending on AI infrastructure in 2026, up from 41% in 2023. The surge, part of a $800 billion AI‑related capex wave, is channeling...
Tata Electronics Aims to Be $30 Billion Business with Fab Play: CEO & MD Randhir Thakur
Tata Electronics, the Tata Group’s semiconductor arm, has surged from a ₹400 crore base to a ₹1.3 lakh crore (≈$15 bn) revenue run‑rate in four years and now aims to double that to $30 bn by 2031. The company is building India’s first fab in...
Adiuvo Explorer Board Aims to Bring Artix UltraScale+ FPGA to $99 Platform
Adiuvo is engineering the Explorer Board, a $99 FPGA development platform built around Xilinx’s Artix UltraScale+ AU7P. The board packs 37K LUTs, 216 DSP slices, 3.6 Mb of block RAM and integrated GTH transceivers, and adds external HyperRAM for larger memory...
Doubling Throughput with Speculative Decoding on DGX Spark
Pretty sure I can double the t/s Isn’t possible to do speculative decoding with a DGX spark on the same network?
PsiQuantum Appoints Intel CEO Lip‑Bu Tan to Board, Sharpening Quantum Hardware Roadmap
PsiQuantum announced that Intel CEO Lip‑Bu Tan has joined its board of directors, bringing decades of semiconductor leadership to the quantum‑computing startup. The move coincides with a reshuffle that placed former AMD president Victor Peng as interim CEO and co‑founder...
US Cloud Giants Commit $700 Billion to AI Infrastructure, Sparking ASIC Demand Uncertainty
Amazon Web Services, Microsoft Azure, Alphabet’s Google Cloud and Meta’s data centers together pledged roughly $700 billion in AI‑focused capital spending for 2026. While Alphabet disclosed a $190 billion AI budget, analysts warn that the surge could outpace ASIC supply, leaving timing...
Boardcon PICOT536 SoM and EMT536 SBC Feature Allwinner T536 Edge AI Processor
Boardcon has introduced the PICOT536 system‑on‑module and the EMT536 development board, both built around Allwinner’s T536 edge‑AI processor. The T536 combines a quad‑core Cortex‑A55 CPU, RISC‑V co‑processors and a 2 TOPS NPU, supporting up to 8 GB LPDDR4/LPDDR4X and 64 GB eMMC storage....
Self-Healing Synaptic Transistor Recovers Memory After Damage
Researchers have created a fully self‑healing, stretchable synaptic transistor that regains most of its function after being cut in half. The device restores about 80% of its operating current and over 90% of its memory within 24 hours without external triggers,...
South Korea’s April Exports Surge 48% to Record $85.9 B, Driven by AI‑Fueled Chip Boom
South Korea’s exports rose 48% year‑over‑year in April, reaching $85.9 billion, the second‑highest monthly total on record. The surge was powered by a 173.5% jump in semiconductor shipments tied to artificial‑intelligence demand, while auto exports slipped amid Middle‑East logistics disruptions. The...
New Wafer Inspection and Metrology Platform From TRI
TRI unveiled the AI‑powered TR7950Q SII wafer metrology and inspection platform, targeting back‑end process and advanced‑packaging applications. The modular system handles 6‑inch to 12‑inch wafers on a high‑stability granite base and offers automated visual inspection for particles, scratches, and contamination....
CadenceLIVE Unveils Chip Design Challenges and SaaS Tool Power
.@Cadence’s CadenceLIVE Silicon Valley 2026 took place on April 15–16, 2026 in Santa Clara, a week after #GTC2026. Cadence has just put the discussion between @nvidia CEO, Jensen Huang & @Cadence CEO, Anirudh Devgan on YouTube. It’s a great listen...
Nvidia's China AI Accelerator Share Hits 0% Amid Export Controls
JUST IN: Jensen Huang says Nvidia's market share of AI accelerators in China has dropped to 0%, calling U.S. export controls strategically counterproductive.
IShares Semiconductor ETF SOXX Posts 50% YTD Gain, Beats Market Amid Reddit Caution
BlackRock's iShares Semiconductor ETF (SOXX) surged about 50% year‑to‑date, closing near $450 and outpacing the broader market. The rally reflects AI‑fuelled demand for chips, even as Reddit’s retail community flags the move as potentially overbought.
Selling Older GPUs to China Boosts US Security
The argument for selling deprecated GPUs to China: Selling them less advanced GPUs than we have in America decreases the odds that they develop more advanced GPUs than America that go down a different, more power intensive tech tree. Pro national-security.
TSMC Futures Leap 40 Points After Trump Halts Iran Action, Taiwan Market Poised for Rebound
TSMC futures jumped 40 points in pre‑market trading after former U.S. President Donald Trump announced a halt to hostile actions against Iran. The move lifted expectations that Taiwan's main index will open higher on Monday, despite continued foreign‑investor selling. Analysts...
Nvidia’s Next Big Market
The episode explores NVIDIA’s evolution from a gaming GPU maker to a dominant force in AI and robotics, highlighting Jensen Huang’s strategy of creating "zero‑billion‑dollar" markets—products with no existing customers that later become essential. Guest author Stephen Witt explains how...
Inference Revives AI Chip Startups' Market Prospects
Inference is giving AI chip startups a second chance to make their mark https://t.co/Jj2cR4IgwS https://t.co/7sYm6hPqyw
AMD Champions Open, Modular AI for Flexible Space Missions
AMD promotes open platforms for space AI, arguing modular systems reduce vendor lock-in risks while improving flexibility across complex multi-supplier mission environments. https://t.co/t81WiURscM
Inference Is Giving AI Chip Startups a Second Chance to Make Their Mark
AI adoption is moving from model training to inference, creating a fragmented workload landscape. Startups are exploiting this heterogeneity by pairing specialized chips—such as Groq's SRAM‑heavy LPUs—with GPUs for pre‑fill tasks, while others use wafer‑scale or optical accelerators for decode....
Inference Is Giving AI Chip Startups a Second Chance to Make Their Mark
AI adoption is moving from model training to inference, creating a heterogeneous workload landscape. Startups see an opening to specialize in either pre‑fill or decode phases, where GPUs, LPUs, and wafer‑scale chips excel in different ways. Nvidia’s $20 billion Groq acquihire...
TSMC Unveils 1.6nm A16 Process, Promising 10% Speed Gain and 20% Power Cut for Q4 2026
TSMC said its upcoming A16 1.6nm node will enter mass production in the fourth quarter of 2026, offering an 8‑10% speed improvement or a 15‑20% power reduction over its current 2nm N2P technology. The announcement positions the company at the...
Anthropic in Early Talks to Buy DRAM-Less AI Inference Chips From UK Startup — Fractile's SRAM Architecture Reduces Need for...
Anthropic is in early talks with UK chip startup Fractile to add a fourth AI inference silicon supplier alongside Nvidia, Google and Amazon. Fractile’s SRAM‑based design eliminates the need for off‑chip DRAM, promising dramatically lower latency and cost. The chips...
Vicor Posts $113 Million Q1 Revenue, CFO Schmidt Flags Strong Growth and Capacity Plans
Vicor Corporation announced first‑quarter 2026 revenue of $113 million, a 20.2% year‑over‑year increase. CFO James F. Schmidt emphasized expanding gross margins, a robust backlog and a strategic push to boost manufacturing capacity. The results signal continued momentum for the power‑electronics specialist...
PlayStation 6 Launch Pushed to 2028‑2029 as AI‑driven RAM Shortage Bites
Sony Interactive Entertainment is reportedly postponing the PlayStation 6 launch to 2028 or 2029 because of a global shortage of high‑bandwidth RAM driven by AI data‑center demand. Leaks suggest a major hardware leap—Zen 6 CPU, RDNA 5 GPU, 30 GB DDR7 memory—and a...
China Weekly Wrap: Cambricon Earnings Edition
Cambricon’s Q1 earnings sparked a 160% revenue jump and 185% net‑profit surge, lifting the STAR50 index about 8.1% and driving on‑shore Chinese benchmarks higher. On‑shore indices such as the CSI 300 and SSE Mega‑cap posted gains of 0.8%‑2.4%, while dividend‑focused stocks...
Scientists Built a Memory Chip that Breaks the Rules of Miniaturization
Scientists at the Institute of Science Tokyo have created a 25‑nanometer ferroelectric tunnel junction memory cell using hafnium oxide, a material that retains polarization at atomic thicknesses. By heating the electrodes to form a semicircular, near‑single‑crystal structure, they eliminated leakage...
GeForce RTX 5070 Laptop GPU with 12 GB: Leak Benchmarks Show Why More VRAM Does Not Automatically Mean More FPS
NVIDIA quietly added a 12 GB VRAM option for the GeForce RTX 5070 Laptop GPU via driver 596.36. Leaked benchmarks show the 12 GB variant matches the 8 GB model in traditional gaming tests, with only a ~2% variance, but it pulls ahead in...
KOSPI Hits Record Highs, Up 30% in April as AI Chip Boom Drives $4 Trillion Market Cap
South Korea’s KOSPI surged to a new all‑time high of 6,750 points, delivering a 30.6% rise in April. The rally, powered by semiconductor giants Samsung and SK Hynix, lifted Korea’s market capitalization to about $4 trillion, but analysts are flagging valuation pressures...
Nvidia’s Push Into Physical AI Sparks Rally in Asian Partners
Nvidia is extending its AI strategy beyond chips into "physical AI," encompassing robotics and autonomous systems. The move sparked sharp share gains for Asian partners—including LG Electronics (+15%), Nanya Technology (+10%), Huizhou Desay and Pateo Connect—after announcements of joint projects....
Qualcomm Unveils Agentic CPU and Plans First Data‑Center Chip Shipments by December 2026
Qualcomm CEO Cristiano Amon announced a dedicated agentic CPU for data‑center AI workloads and confirmed that the company will ship its first custom data‑center ASICs to a major hyperscaler by December 2026. The news sent the stock up more than...
Cerebras Systems Targets $4 Bn IPO at $40 Bn Valuation, Eyes Nasdaq Debut
Cerebras Systems announced plans to raise as much as $4 bn in an initial public offering that would value the AI‑chip maker at roughly $40 bn. The filing, backed by Morgan Stanley, Citigroup, Barclays and UBS, follows a $10 bn multi‑year compute deal...
Nvidia's AI Boost Cuts Quantum Error‑Correction Time by 2.5×, Accuracy Up 3×
Nvidia announced AI‑driven models that accelerate quantum error‑correction decoding by 2.5× and improve accuracy threefold. The move positions the chipmaker to become a key software layer for emerging quantum hardware, a hurdle that has long limited real‑world quantum advantage.
Nvidia’s Push Into Physical AI Sparks Rally in Asian Partners
Nvidia is rapidly expanding its AI ecosystem across Asia, prompting a wave of stock rallies among regional partners. In the past week, South Korea’s LG Electronics, Taiwan’s Nanya Technology, and Chinese firms Desay SV Automotive and Pateo Connect announced collaborations...
Rebellions Raises $400M to Launch RebelRack and RebelPOD AI Inference Systems
Rebellions announced a $400 million pre‑IPO financing round and the immediate availability of two new AI inference hardware offerings, RebelRack and RebelPOD. The funding, led by Mirae Asset Financial Group and the Korea National Growth Fund, brings total capital to $850 million...
Qualcomm Shares Jump $15B on Q2 Beat and 5G/AI Growth Outlook
Qualcomm's shares surged 12% after the chipmaker posted fiscal Q2 results that beat forecasts, highlighted a 38% jump in automotive revenue and outlined a roadmap for custom silicon and AI‑enabled 5G chips. The rally added roughly $15 billion to the company's...
An AI Analysis of Qualcomm: The Edge Is Where AI Will Live
Qualcomm is positioning itself as the primary supplier of edge‑AI silicon as artificial intelligence moves from data‑center servers to billions of endpoint devices. The company’s Snapdragon line, built over two decades for power‑efficient inference, targets smartphones, PCs, cars, industrial sensors...
Utah Germanium Refining Boosts Semiconductor Material Demand
5N Plus sees strong demand for semiconductor materials as germanium refining expands in Utah. https://www.metalnomist.com/2026/05/5n-plus-semiconductor-materials-demand.html
All Viable Accelerators Could Win Hyperscaler Adoption.
I maintain my thesis that all viable compute (in this context accelerators) has a chance of being adopted by hyperscalers and frontier labs. Hence why @cerebras is getting more attention and now @tenstorrent. BUT, @AMD has more upside, and maybe...
Nvidia Reclaims $5 Trillion Market Cap, Sets Sights on $6‑$10 Trillion
Nvidia's shares jumped more than 21% in April, pushing the chipmaker back above a $5 trillion market valuation. The rebound is tied to strong demand for AI inference chips and projections that the company could reach $6‑$10 trillion in market cap within...
Taiwan Semi Is Selling ARM Stock. Should You?
Taiwan Semiconductor Manufacturing Co. (TSMC) has fully exited its stake in Arm Holdings, selling roughly 1.1 million shares for about $231 million. Despite the divestment, Arm’s shares continued to rise, reflecting strong investor confidence in its AI‑driven growth. The chip‑design firm posted...