IBM unveiled a 7Å transistor architecture that uses staggered nanosheet transistors stacked on a precisely beveled angle, almost like tiles on a roof. That allows more transistors to be crammed into a given area, boosting performance by 50% or power efficiency by up to 70%. Perhaps even more important, IBM claims a 40% improvement in SRAM scaling, which is orders of magnitude faster and lower latency than HBM. Mass production is slated for ~2031.
Fig. 1: Huiming Bu, IBM VP of Semiconductor Global R&D, shows a staggered GAA transistor model for stacking logic. Source: IBM
Also on the extreme scaling front, the U. of Tokyo created 1nm semiconducting molybdenum disulfide nanotubes for potential use in ultra-scaled GAA transistors.
Dealmaking continues
Onsemiplans to acquireSynaptics for ~$7B for its edge AI compute, wireless connectivity, human-machine interface, touch, display, and biometric technologies. In the release, onsemi CEO Hassane El-Khoury said physical AI will require power, sensing, connected compute, and control to work together, and that Synaptics will help position the company across those four areas. The deal is expected to close mid-2027.
Qualcommplans to acquireModular, which makes a software stack that allows AI to run across different hardware architectures.
Advantest and OpenLight plan to collaborate on photonics test solutions for high-volume manufacturing.
OpenAI and Broadcom designed a chip tuned specifically for LLM inference.
Micron inked a deal with Anthropic spanning memory and storage AI architecture design and supply, use of Anthropic’s models at the company, and additional strategic investment.
Equipment and advanced packaging
Applied Materials introduced an epitaxy system for DRAM fabs, and CMP, deposition, and e-beam inspection systems for advanced packaging.
JCET is reportedly planning a ~$1.15B Shanghai advanced packaging and testing site.
AlixLabs launched the first commercially available equipment for its ALE Pitch Splitting etching technology to increase pattern density.
Groq raised $650M, pivoting to providing an AI inference cloud after Nvidia licensed its LPU design and hired many of the company’s employees last year.
Nearfield Instruments raised $380M to expand production of its 3D metrology, inspection, and process control equipment.
Upscale AI raised $190M for its AI network fabric that connects accelerators, memory, and storage.
The U.S. Commerce Department signed a definitive agreement with I-Pulse for a $250M CHIPS R&D award to develop SiC power semiconductors for next-gen switches.
Agility Robotics will go public through a merger with Churchill Capital in a deal valuing the humanoid robotics company at $2.5B.
Memory and storage
Samsung uncorked UFS 5.0 for mobile devices with a data transfer speed of up to 10.8GB/s and up to 1TB capacity.
In its earnings call, Micron said AI demand and structural capacity limits will keep memory supply tight beyond 2027, prompting 16 customers to sign multiyear take-or-pay agreements with price floors. Also, the company said its 12-high HBM4 ramp is progressing twice as fast as HBM3E, and has already generated more than $1B in shipments.
IDC forecasts continued a supply/demand imbalance for memory through 2027.
SK hynix plans to list ADRs on the Nasdaq, potentially bringing in $29B.
Increased density at advanced nodes, multi-die assemblies, and the rollout of AI everywhere are making it much more challenging to ensure that memory will function properly over its expected lifetime. Here’s how to ensure that memory is fully tested and repaired.
The growing concerns about data centers water usage has put hyperscalers on the hot seat to say how they are addressing the issue, including Amazon claiming a 52% improvement since 2021, Meta using desalinated seawater in India, and Google’s air cooling and recycled water programs.
Classiq and TEA TEK Group are teaming up to build a 128-qubit quantum hub in Naples, Italy.
TNO and ASML are collaborating to scale European photonic chip manufacturing, aiming to move 6-inch InP photonic chips closer to high-volume production.
X-FAB received a ~$145M grant from the German government to expand its MEMS foundry in Erfurt, Germany.
QuantumDiamonds will receive a ~$87M grant from the German government to produce metrology and inspection systems based on quantum sensors.
SiEngine TechnologylicensedArteris FlexNoC interconnect IP for its next-gen automotive SoC platform.
ChipAgentsjoined the Amazon Web Services Partner Network to scale its agentic AI chip design platform.
New Technologies
Data center technology
Qualcomm revealed new data center CPUs, inference accelerators, and a 3D-stacked near-memory computing architecture. Qualcomm will also supplyMeta with data center CPUs.
Nvidia proposes using higher-temperature liquid to cool data centers, an approach the company says improves efficiency and reduces water consumption.
The growing adoption of AI everywhere is shifting the focus from just building the fastest chip at the latest process node to architecting a system around the rapid movement of huge amounts of data.
BrainChip released a physical development reference platform for RF signal classification using its neuromorphic processor.
The rapid development of AI has resulted in new capabilities being provided to verification teams, beyond their ability to rationally insert them into accepted methodologies.
Manufacturing, materials
Bruker introduced new X-ray diffraction systems optimized for batteries, amorphous materials, and catalysts.
Mobile devices
Sony Semiconductor Solutions debuted a 1/2-type ~64-effective megapixel CMOS image sensor for mobile applications.
Research
A Penn State team built a self-powered monolithic 3D chip that stacks graphene chemical sensors, MoS2/WSe2 2D logic and a silicon photovoltaic energy-harvesting layer, which could open the door to low-power autonomous sensing and edge-computing systems.
Binghamton University engineers are developing a single-stage point-of-load converter for AI data centers that steps 48V power down near GPUs, with a lab prototype showing 10% to 12% higher efficiency and twice the slew rate for faster power delivery.
Rice University researchers automated X-ray defect analysis for diamond and other wide-bandgap semiconductors, speeding up material quality checks for power, RF, and quantum devices.
The White House issued an executive order mandating that “high value assets” and “high impact systems” within the federal government migrate to PQC keys by the end of 2030 and PQC digital signatures by the end of 2031.
France’s national cybersecurity agency, ANSSI, will stop certifying security products without quantum-resistant encryption technology beginning next year and will also require all critical infrastructure operators and government agencies to use only post-quantum security technology by 2030.
Japan’s defense forces reportedly used counterfeit USB drives compromised with a China-linked virus for over a year to access top-secret information.
STMicro introduced a secure mobile chip designed to outfit smartphones and personal electronics with quantum-ready security solutions.
Vulnerability
European research hub Paradigm Shift discovered a new iPhone BootROM hardware vulnerability, affecting devices powered by Apple A12, S4/S5, and A13 chips.
IBM and Red Hat are expanding their $5B Project Lightwell open source cybersecurity collaboration to include Palo Alto Networks’ virtual patching technology, providing network-layer protection to help shore up vulnerabilities.
IBM joined OpenAI’s Daybreak Cyber Partner Program, a partnership focused on integrating advanced frontier AI technology into organizations’ cybersecurity operations.
SK keyfoundry developed a Bi-SCR-based on-chip EMC protection technology for automotive semiconductors, adding higher electromagnetic robustness to its 0.13-micron BCD process for power ICs used in harsh vehicle environments.
BAE Systems demonstrated its Endura radiation-hardened SoC, a GlobalFoundries 45nm SOI-based processor.
Autonomous
Lyft now requires its fully driverless AVs on its platform to use redundant multi-modal perception systems that include cameras, radar, and lidar.
Tesla came under fire from federal regulators after one of its Model 3 vehicles sped off a Texas road and crashed into a home, killing a woman. The driver reportedly told police that the vehicle was operating its FSD system at the time of the crash, leading to federal investigations into the incident and the victim’s family filing a lawsuit against the carmaker.
Waymorecalled over 3,800 vehicles following a string of incidents in which the company’s AVs drove past road closure signs and into construction zones.
EVs and batteries
Nissan reportedly halted the development of an electric version of its top-selling Qashqai model in Europe amid global restructuring at the brand.
Signoff Of Synthesis-Optimized Registers: How do you know when you sign off on a complex chip design that everything is going to work? There are more variables, more elements that need to be verified, and more waivers that need to be generated. Synopsys‘ Suresh Barla talks about how to ensure that RTL is fully optimized for PPA targets in large designs that can include hundreds of millions of gates and dozens of design blocks, and that the netlist meets sign-off quality.
Quantum
Quantum Computing Inc. completed its $73M acquisition of NHanced Semiconductors, adding semiconductor and nanophotonics fabrication and advanced packaging capabilities.
IQM proposes using error-correcting directional tile codes to reduce logical qubit error rates without new hardware.
QuEra laid out its roadmap to release a quantum system with 1,000+ logical qubits and capable of 1 billion reliable logical operations in 2028-2029.
IBM Research highlighted an effort to establish benchmarks for quantum optimization.
Qubic raised $2.5M to accelerate commercialization of low-noise cryogenic amplifiers and quantum-enhanced radar.
ORNL and MIT researchers developed an electron-beam method for rearranging atoms inside a crystalline semiconductor at room temperature, creating more than 40,000 quantum defects that could be useful for quantum materials, sensing, dense memory and atomic-scale logic.
SEMICON West in October will include a dedicated Workforce Development Pavilion to connect job seekers with employers and highlight industry efforts to build the semiconductor talent pipeline, along with IC training courses.
U. of Maryland students designed and implemented ASIC projects spanning RF energy harvesting, mixed-signal data converters, neuromorphic computing, and emerging memory.
UNLV launched a semiconductor manufacturing outreach program for high-school students, part of a broader push to move chip-workforce development earlier through K–12 lessons, teacher bootcamps, high-school camps, and regional talent-pipeline programs.
InP and SiPho join CMOS as critical technologies. Lasers, CPO and OCS will be everywhere (indium phosphide, silicon photonics, co-packaged optics, optical circuit switch).
Reducing variation in manufacturing, monitoring behavior over time, and targeting specific workloads can have a big impact on power, performance, and area/cost.
This website uses cookies to improve your experience while you navigate through the website. The cookies that are categorized as necessary are stored on your browser as they are essential for the working of basic functionalities of the website. We also use third-party cookies that help us analyze and understand how you use this website. We do not sell any personal information.
By continuing to use our website, you consent to our Privacy Policy. If you access other websites using the links provided, please be aware they may have their own privacy policies, and we do not accept any responsibility or liability for these policies or for any personal data which may be collected through these sites. Please check these policies before you submit any personal information to these sites.
Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.
Any cookies that may not be particularly necessary for the website to function and is used specifically to collect user personal data via analytics, ads, other embedded contents are termed as non-necessary cookies. It is mandatory to procure user consent prior to running these cookies on your website.
Leave a Reply