Samsung Foundry aims to strengthen its AI chip offering for the robotics and automotive sectors by leveraging Cadence's design platforms. This collaboration highlights the growing importance of optimized hardware solutions for artificial intelligence in critical domains, where performance, efficiency, and data sovereignty requirements drive dedicated architectures and on-premise or edge deployments.
SOND, a startup led by Bose's former head of sleep products, has announced $7 million in funding. The company focuses on developing AI-powered smart earbuds designed to improve sleep quality, marking a new step in integrating AI into wearable devices for health and wellness.
Nvidia has initiated restricted access distribution of its Vera CPU, an 88-core processor, for initial Linux benchmarks. Early tests indicate Vera can compete with or outperform AMD's Epyc and Intel's Xeon counterparts in selected tests. This development marks a significant step for Nvidia in the server processor arena, with implications for high-performance workloads and on-premise infrastructure.
Itera has developed a prototype fluid circuit board, based on a patented architecture utilizing glass and liquid metal. This hardware solution, aimed at PCB engineers, promises to revolutionize development. Its ability to be physically reconfigured in under a minute could accelerate hardware iterations by a thousandfold compared to traditional printed circuit boards, offering a super-fast approach to prototyping.
Samsung Electronics has commenced construction of a new semiconductor testing facility in Vietnam, backed by an initial investment of $1.5 billion. Located north of Hanoi, this facility marks the company's first site in the country exclusively dedicated to this activity, with operations slated to begin by November 2027.
Daxin Materials, led by Chairman and CEO Cheng-yih Lin, is expanding its offerings into the low dielectric constant (Low Dk) and low dissipation factor (Low Df) materials market. This strategic move aims to support the booming demand for high-performance AI servers and Low Earth Orbit (LEO) satellite applications, sectors that require advanced components to ensure signal integrity and energy efficiency.
ICP DAS has announced a significant expansion in semiconductor controller orders and a strategic acceleration towards Tensor Processing Units (TPUs) dedicated to the medical sector. The company also plans to expand its manufacturing facilities to support this growth and new market direction.
SmartSens and Unisoc have announced a collaboration to develop optical interconnects based on Micro LED technology. The goal is to enhance the efficiency and performance of AI clusters, a critical aspect for on-premise deployments of Large Language Models. This innovation aims to overcome the limitations of current connectivity solutions, offering higher bandwidth and lower latency, which are decisive factors for scaling local AI infrastructures and managing data sovereignty.
Transcend is set to showcase its new "AI-ready" enterprise SSD and DDR5 memory solutions at COMPUTEX 2026, designed to meet the demanding requirements of AI infrastructure. These components are crucial for supporting intensive Large Language Model (LLM) workloads and AI applications, offering fundamental performance and reliability for on-premise deployments and data sovereignty.
Broadcom and Taiwanese chipmakers are at the forefront of a growing adoption of Application-Specific Integrated Circuits (ASICs). This trend marks a significant evolution in hardware dedicated to artificial intelligence, particularly for Large Language Models. The emphasis on customized solutions promises optimal efficiency and performance, with direct implications for on-premise deployment strategies and Total Cost of Ownership (TCO).
A comparative analysis of NVIDIA RTX 5090 and RTX 6000 PRO (MaxQ and WS/SE) performance for AI diffusion workloads highlights the trade-offs between power consumption and speed. Tests conducted with various power configurations reveal how energy efficiency can be a decisive factor in on-premise deployments, where TCO and thermal management are priorities for CTOs and infrastructure architects.
Huawei, with its Tau Law series 2 initiative, focuses on crucial enabling technologies for artificial intelligence, such as advanced packaging, AI interconnects, and EDA. These developments are fundamental for creating high-performance, efficient, and scalable AI infrastructures, directly impacting on-premise deployment strategies and data sovereignty for companies managing LLM workloads.
Wah Hong Industrial Corp., a Taiwanese optoelectronics materials maker, is focusing its strategy on the growing demand for high-end Printed Circuit Boards (PCBs) and chip packaging. This move highlights the importance of advanced materials for modern computing infrastructure, including systems dedicated to on-premise Large Language Models (LLMs), where performance, reliability, and thermal management are crucial for Total Cost of Ownership (TCO) and data sovereignty.
After two decades, Nvidia is phasing out its iconic Control Panel, consolidating all driver updates and management features into the new Nvidia App. This unification aims to streamline user experience and hardware management, with significant implications for enterprise environments and on-premise deployments relying on Nvidia GPUs for intensive workloads like LLM Inference.
AMD has introduced a controversial update to its Vivado software licensing, significantly impacting FPGA developers working on Linux. The new tier model dictates that future free versions of Vivado will be exclusively available for Windows, leaving Linux users without a free option for FPGA development with this tool. This move raises questions about AMD's support strategy for the Open Source ecosystem.
Tryx has unveiled its new Holo 360 AIO liquid cooler, featuring a unique holographic display. This aesthetic innovation, based on beam splitters, emerges in a context where thermal management is crucial for system performance, including those dedicated to LLM inference and training in on-premise environments. Available in white and black, the product highlights the continuous evolution of hardware innovation.
GlobalPlatform has announced the launch of Pavona, a new open-source silicon ecosystem. The initiative, supported by founding members such as Meta, Qualcomm, and the University of Oxford, aims to provide a certifiable foundation for hardware development. This approach promotes control and sovereignty over components, crucial aspects for on-premise LLM deployments and infrastructures requiring high security and compliance standards.
NVIDIA is set to launch its Vera CPU, an ARM-based data center processor designed for agentic AI workloads. Featuring proprietary Olympus cores, Vera promises competitive performance against Intel and AMD x86_64 CPUs, a significant milestone for ARM architecture in the data center sector. Early benchmarks on Linux suggest a substantial leap forward for on-premise AI infrastructures.
SK Hynix has introduced iHBM, a new thermal architecture that integrates cooling elements directly into the HBM memory interface. This innovation aims to reduce thermal resistance by 30%, enhancing efficiency for HBM5 accelerators and high-density AI data centers. The solution addresses critical thermal challenges for on-premise Large Language Model deployments.
Intel has announced the development of pmtctl, a new Linux kernel tool designed to interface with its Platform Monitoring Technology. This tool will enable more granular management of platform telemetry data, offering IT professionals greater control and visibility over underlying hardware, a crucial aspect for on-premise deployments and efficient resource management.