Chips and semiconductors: chip design and manufacturing, foundries such as TSMC, Intel and Samsung, memory and HBM, AI accelerators, chipmaking equipment and lithography, and the export controls and subsidies that shape the industry, from the semiconductor trade press, analysts and the tech news, plus any story whose headline is about chips.
Summary
1,538 news articles and 394 research items in this section, back to 10 December 2025. Browse the news ↓Research ↓
78 stories from 30 publishers in the last 24 hours, 595 from 98 this week; 24 stories were covered by more than one newsroom.
Nvidia does not plan to reinstate Jensen Huang, despite the former CEO's suggestion that he would be “honored” to return if the company's board of directors supported the move.
Nvidia’s provision of chips to companies running data centres in Southeast Asia has come under the spotlight amid questions surrounding founder Jensen Huang’s relationship with Alice Huang, the CEO…
Extends Fluxless ThermoCompression (FTC) Technology to Panel-Level Packaging Ships the Industry’s First FTC Panel-Level System – APTURA™ WSX to Two Customers Kulicke & Soffa announced today the…
Most active this week: TechPowerUp (61), Tech in Asia (34), The Hill (34), SemiWiki (23). Built automatically from headlines; the top items are the stories the most newsrooms covered in the last three days.
Analysis
Nvidia's board rules out reinstating Jensen Huang and says it is ending chip sales to China, even as the company reports $115.8bn in quarterly revenue, more than double a year earlier.
The Nvidia spy scandal dominated the 48 hours: board chair Harvey Jones used the company's first earnings call since Huang's resignation to state the board 'does not intend' to bring him back, hours after Huang publicly asked for his job 15. Financially the company appears unscathed, with revenue up 103% year on year and a bullish outlook under interim CEO Colette Kress 1622. The political fallout is widening: Washington has restricted China's remote access to AI chips and imposed an interim rule on exports to Singapore, Huang has agreed to appear before Congress, and an SEC probe is reported 271012. Beyond Nvidia, the sector saw product news at SC26, a Siemens EDA acquisition and continued US capacity investment 891415.
Key developments
Board shuts the door on Huang. Newly appointed board chair Harvey Jones said on the Q3 FY2027 earnings call that the board does not intend to reinstate Huang, only hours after Huang said on X that he wanted to return [1][5][25][27]. The decision came despite public backing for his return from Donald Trump, Elon Musk and Silicon Valley executives, suggesting the board is prioritising governance and national-security concerns over founder loyalty [3][6][13]. Huang's departure followed reports of affairs with Megaspeed executive Alice Huang, who is suspected—though not confirmed—of being a Chinese intelligence asset, and with a senior executive who reported to him [2][13][27]. 135613
Record results despite the turmoil. Nvidia reported revenue of $115.8bn for the quarter ended 25 October, up 20% sequentially and 103% year on year, with gross margins near 75% and GAAP EPS of $2.70, beating expectations and guiding to continued record revenue [16][22][29][30]. The numbers indicate demand for AI accelerators is insulated, at least for now, from leadership chaos and a reported SEC investigation [7]. Interim CEO Colette Kress, previously CFO, also pledged 'disciplined' investment in neoclouds and reaffirmed the open-source strategy and the planned Hugging Face acquisition [18][20]. 716182022
Nvidia exits China. Kress said the company is done selling chips to China, a strategic break from Huang's long effort to keep access to that market [17]. The board simultaneously created a National Security and Export Control Oversight Committee to oversee export compliance and trade risk at board level [11]. The cost of the exit is unquantified in the reports, but the move aligns the company with hardening US policy [17][32]. 111732
Washington tightens access, Singapore caught in the middle. The US has blocked China's remote access to AI chips and the Commerce Department issued an interim rule restricting AI chip exports to Singapore, prompted by scrutiny of Nvidia's supply to Southeast Asian data-centre operators including Megaspeed [2][12]. China condemned the move, while Singapore is seeking talks on 'trusted access' for local firms [2][12]. CEPA argues the affair exposes gaps in US efforts to keep cutting-edge technology from China [32]. 21232
Scandal shadows Trump–Xi talks. The affair has become a factor in US–China diplomacy, with Beijing worried it will fuel pressure for further chip restrictions just as Trump and Xi meet [4]. The Atlantic Council warns that closing export loopholes could invite renewed Chinese rare-earth restrictions, a vulnerability the trade truce never resolved [31]. Huang has agreed to voluntary transcribed interviews with Congress on 1 December, keeping the issue in the political spotlight [10]. 41031
Sector news beyond the scandal. At SC26 Nvidia announced the Rubin HPC GPU for its Vera Rubin platform and, with IBM and Fermilab, a quantum-classical co-design platform plus GPU-accelerated error-correction tools [14][15][34]. Siemens agreed to buy Breker Verification Systems to bolster SoC verification, ACM Research is adding an Oregon R&D and manufacturing site, and Micronics Japan plans to expand test capacity for Samsung [8][9][19]. AMD's CES 2027 invitation hints at a Zen 6 'Olympic Ridge' desktop reveal, though this is speculation based on the invite's motif [24]. 89141519
Themes
Governance reset under political pressure. The board's refusal to reinstate Huang, the creation of a board-level export-control committee and the China exit together mark a deliberate distancing from the founder era [1][11][17]. That this occurred despite presidential and industry lobbying highlights how national-security exposure, not just personal conduct, is driving the response [3][6][13]. 1361113
Export controls spreading to third countries. The Singapore interim rule and the block on remote access show US policy now targeting intermediary routes, not just direct sales to China [2][12]. Allies and partners are scrambling for exemptions while Beijing weighs retaliation via rare earths, raising the risk of broader supply-chain disruption [4][12][31]. 241231
Demand decoupled from drama. Doubling revenue and raised guidance suggest hyperscaler and AI demand is indifferent, so far, to who runs Nvidia [16][22][30]. The open questions are the cost of forgoing China and whether 'disciplined' financing of neoclouds signals a more cautious capital stance [17][18]. 1617182230
What to watch
Huang's 1 December congressional interviews, which could surface new detail on Megaspeed and the alleged intelligence link and shape further legislation [10][2]. 210
Whether Singapore secures a 'trusted access' arrangement under the Commerce interim rule, and how China responds, including any rare-earth measures [12][31][4]. 41231
Confirmation and scope of the reported SEC investigation and whether it touches Nvidia's export compliance or disclosure practices [7][11]. 711
Any quantified guidance on the revenue impact of exiting China and on the pace of Nvidia's neocloud investments [17][18]. 1718
Written by Claude Fable 5.1 (Anthropic), based on 45 recent headlines and excerpts in this section, 2d ago. AI analysis can misread or miss things: the numbers link to the original reporting, which is the authority.
Research 394
Journal articles, academic blogs, think-tank reports and papers, kept apart from the news below.
Pretrained language models (LMs) remain unreliable on exact multi-digit arithmetic without gradient fine-tuning or external tool interpreters. We introduce closed-form, zero-gradient circuit-grafting primitives that compile deterministic computation directly into pretrained LM weights using zero training examples and zero runtime hooks, while bounding per-layer normalization drift and preserving natural-language (NL) competence. Our construction combines three analytical…
Frontier text-to-image (T2I) foundation models predominantly rely on multi-billion parameter Diffusion Transformers (DiTs) pretrained across massive compute clusters, creating an accessibility barrier for academic laboratories. When training DiTs from scratch on single consumer GPUs (e.g., an NVIDIA RTX 4090 with 24\,GB VRAM), practitioners encounter three primary bottlenecks: quadratic attention sequence costs, Just-In-Time (JIT) compilation graph breaks induced by variable…
LocoMuJoCo is an open-source benchmark for imitation learning (IL) and reinforcement learning (RL) on whole-body control. The library couples MuJoCo (CPU) with MJX and the MjWarp backend (GPU) under a single API, exposes 9 humanoid, 3 (musculo-)skeletal, and 4 quadruped embodiments, and ships a curated native motion-capture dataset for every humanoid, to which it can also retarget AMASS and LAFAN1. A modular plug-in architecture makes observations, rewards, control modes…
Weight-only quantization keeps the activations of large language models (LLMs) in a higher-precision format, pushing the on-chip inference cost to the activation arithmetic. Bit-serial, variable-length accelerators offer a promising path to reduce this cost by lowering activation precision while preserving model quality. They process only the mantissa bits that each layer needs. The length has to be chosen per layer and operator, and today it is chosen by offline…
Embedded computing demands high performance and energy efficiency across workloads with widely different computational characteristics. Choosing the most suitable execution model therefore requires a quantitative understanding of how each model maps to different workload classes. In this paper, we compare three execution models for energy-efficient embedded computing: scalar multicore, stream ing multicore, and vector processing. We use three open-source RISC-V accelerators…
Automated analog circuit sizing methods, from black-box optimizers to recent large language model (LLM) agents, typically search the design space with a SPICE simulator in the loop, whereas an experienced designer works out a design by hand analysis. This paper presents CRAFT, a flow in which an LLM agent designs a given topology by calculation, as a designer does, and without an initial sizing. The agent works only through DeX, a service that uses pre-computed lookup tables…
California Attorney General Rob Bonta issued subpoenas to Nvidia regarding reports the company sold advanced AI computer chips to China, his office announced Thursday. “The conduct of Nvidia and its former CEO raises bright red flags, including about the ways in which the company may have misled investors,” Bonta said in a press release. “My...
As in Q2, Nvidia’s Q3 saw a YoYvdoubling of revenues and profits. In the three months ended October 25, Nvidia’s revenue grew by 103% from the same period last year,
A US House committee on Thursday released confidential licence applications from Chinese companies seeking to buy advanced artificial intelligence chips from Nvidia and AMD, information normally protected under US export-control law. Brian Mast, a Florida Republican and chair of the House Foreign Affairs Committee, used his authority under the Export Control Reform Act to make public the details of Nvidia and AMD licence applications to sell AI chips to China. “[The…
It looks like Seasonic's 3,200 W PRIME ENTERPRISE PX-3200 PSU is close to launch. As spotted by Videocardz, the company's official Weibo account posted that the new PSU is coming soon, attaching an image of Supermicro's NVIDIA GB300 AI workstation as well. The PX-3200 offers six 12V-2x6 connectors, ideal for feeding several high-end GPUs directly. It also carries 80 Plus Platinum and Cybernetics Platinum ratings, as far as efficiency goes. There's no official word on pricing…
Boston Dynamics, the global leader in mobile robotics, and GlobalFoundries(opens in new tab) (GF), a manufacturer of essential semiconductors, today announced an expansion of their collaboration to deploy Spot for autonomous enterprise asset management across GF’s global semiconductor fabs.
The semiconductor market continues to grow, driven in part by the expanding use of generative AI. In logic ICs, which perform computing and control functions, transistor structures are becoming increasingly 3D and complex in addition to continued scaling, requiring even more advanced thin-film deposition technologies.
Ecolab today announced it is convening an open AI infrastructure ecosystem to solve the data center industry’s need for faster next-generation data center development.
"We've got this," Colette Kress, interim CEO of Nvidia, said Tuesday night as she closed out the embattled chip company's third quarter earnings call. By Thursday, the company's stock price was hovering at around $150 per share, a nearly 40 percent drop from its all-time high on October 6th, and wiping out a brief post-earnings
The Trump administration approved the shipment of nearly 900,000 AI chips to Defense Department-designated Chinese military companies, according to newly released documents from the House Foreign Affairs Committee. The documents revealed the U.S. signed off on the sale of 879,868 AI semiconductor chips from Nvidia and AMD. The specific chips sold were Nvidia’s sensitive H200...
Amazon SageMaker AI now monitors your generative AI inference endpoints by default and tells you when a change could improve performance or lower cost. It tracks latency, GPU utilization, and serving configuration, then sends a finding through AWS Health with a one-click path to benchmark alternatives in SageMaker Studio before you apply any change.
Enterprise adoption of artificial intelligence is creating new requirements for enterprise AI infrastructure. While much of the early AI investment centered on access to powerful GPUs, organizations moving AI applications and agents into production are discovering that accelerators are only one part of the equation. Agentic AI systems need environments where they can execute code,
Windows 7 was a beloved operating system for many, and if you're still running a machine with that OS, this release might be for you. Three Windows 7 enthusiasts have managed to get current-generation GPUs and DirectX 12 running on Microsoft's 17-year-old OS. Bob Pony shared the project on X, called WDDM2Win7, an unofficial backport of the Windows Display Driver Model (WDDM) 2. The screenshots show Intel UHD Graphics 770, an NVIDIA GeForce RTX 5060, and an AMD Radeon RX 9060…
Insider Brief It turns out teamwork can indeed make the dream work, even when the team isn’t allowed to talk. It helps to have a few entangled qubits, though. In a study, a Google-led team of researchers used a quantum chip to beat the best classical strategies in games that barred players from communicating, demonstrating
NAND flash appears to be finally getting cheaper, at least on the spot market. According to the Seoul Economic Daily, spot prices for NAND flash have dropped roughly 20% since the second half of October. TrendForce says the drop in price took just two weeks, as Chinese traders dumped inventory. A 512 Gb wafer now goes for about $14.30, which is 32% below last quarter, and 1 Tb products are down 19% to $24.90. Apparently, the reasoning is weak consumer demand, since buyers…
Khaled Alashmouny is the founder and CEO of AIDAChip. He spent 13 years at Apple leading analog and mixed-signal design teams and has about two decades of experience building silicon. He founded AIDAChip to solve the problem he saw on every complex program: talented teams losing time, and sometimes silicon, to misalignment across…
NVIDIA and Microsoft are extending their full-stack collaboration to help organizations build, customize and deploy intelligent applications — from next-generation infrastructure on Azure to industrial systems and Windows PCs. At Microsoft Ignite, the companies are expanding access to Azure virtual machines powered by NVIDIA Vera Rubin NVL72 systems, as well as adding customer availability to
As artificial intelligence workloads shift to inference, memory bandwidth decides whether hardware generates tokens quickly or sits idle. D-Matrix Corp. is answering with Raptor, a second-generation chip stacking 3D DRAM beneath its compute logic. AI infrastructure has become a systems game, and winners combine memory, interconnect and software into cohesive designs, according to Satyam Srivastava
US President Donald Trump bought as much as US$265,000 in Nvidia stock and sold as much as US$15,000 worth of the stock in September, weeks before reports surfaced of CEO Jensen Huang’s ties to an alleged Chinese spy, according to Trump’s latest financial disclosure released on Thursday. US lawmakers have been asking questions about Trump’s knowledge of Huang’s personal life, and the president has said he “was amazed” by the allegations against Huang, who is among the…