發現硬體團隊的趨勢零件、產業訊號和社群搜尋見解。
Anthropic engineer Gurmeet Saran focuses on building trustworthy data infrastructure and observability frameworks to ensure AI model transparency and data privacy. For hardware teams deploying large language models, robust software-level observability and data governance dictate how compute clusters and accelerators scale safely.
Intel officially branded its 'Crescent Island' AI accelerator as the Intel GPU D-Series, featuring the D270 model with 1.3 PFLOPS of dense FP8 performance. The air-cooled, 350W PCIe card bypasses costly HBM by leveraging up to 480GB of LPDDR5X memory for cost-effective enterprise AI inference.
The element14 Community has launched a smart sustainability design challenge inviting engineers to develop electronics-based recycling and waste reduction solutions. Participating hardware teams can leverage component kits and intelligent control systems to prototype real-world environmental applications.
Intelligent Memory is expanding its DDR4 and eMMC product availability to support legacy and long-lifecycle industrial systems. This portfolio update provides hardware engineers with reliable component options as mainstream memory suppliers phase out older process nodes.
Intel officially branded its 'Crescent Island' AI accelerator as the Intel GPU D-Series, featuring the D270 model with 1.3 PFLOPS of dense FP8 performance. The air-cooled, 350W PCIe card bypasses costly HBM by leveraging up to 480GB of LPDDR5X memory for cost-effective enterprise AI inference.
The element14 Community has launched a smart sustainability design challenge inviting engineers to develop electronics-based recycling and waste reduction solutions. Participating hardware teams can leverage component kits and intelligent control systems to prototype real-world environmental applications.
Intelligent Memory is expanding its DDR4 and eMMC product availability to support legacy and long-lifecycle industrial systems. This portfolio update provides hardware engineers with reliable component options as mainstream memory suppliers phase out older process nodes.
Marvell has announced its second-generation Structera X 3406 and 3506 CXL 3.2 memory expanders, bringing six-channel support, PCIe Gen 6 connectivity, and inline LZ4 compression to tackle datacenter AI memory bottlenecks. These controllers provide high-density DDR4 and DDR5 scaling for server architectures targeting memory-heavy workloads like large language model serving.
Engineers can drive LCD graphics on resource-constrained microcontrollers without dedicated display controllers, hardware accelerators, or external DRAM. This approach trims bill-of-materials costs and simplifies hardware design for edge displays.
Applied Materials is investing $5 billion in its new EPIC Center to accelerate collaborative semiconductor R&D and shorten technology commercialization cycles. This multi-node innovation hub aims to clear manufacturing hurdles for future AI accelerators, advanced logic, and high-density memory components.
Fervo Energy has brought its utility-scale Cape Station enhanced geothermal project online in Utah, supplying 33 MW of carbon-free baseload power to the grid. This milestone proves next-generation geothermal scalability, offering a critical clean energy source to power expanding data centers and heavy computational loads.
Ezurio has launched the Veda IF912 and IF913 connected modules, combining Wi-Fi 6/6E, Bluetooth LE 5.4, and an integrated Arm Cortex-M33 MCU into a single 7x11 mm package. This high-integration design allows hardware teams to streamline architectures, shrink footprints, and cut overall BOM costs for space-constrained IoT applications.
Applied Materials is investing $5 billion in its new EPIC Center to accelerate collaborative semiconductor R&D and shorten technology commercialization cycles. This multi-node innovation hub aims to clear manufacturing hurdles for future AI accelerators, advanced logic, and high-density memory components.
Fervo Energy has brought its utility-scale Cape Station enhanced geothermal project online in Utah, supplying 33 MW of carbon-free baseload power to the grid. This milestone proves next-generation geothermal scalability, offering a critical clean energy source to power expanding data centers and heavy computational loads.
Ezurio has launched the Veda IF912 and IF913 connected modules, combining Wi-Fi 6/6E, Bluetooth LE 5.4, and an integrated Arm Cortex-M33 MCU into a single 7x11 mm package. This high-integration design allows hardware teams to streamline architectures, shrink footprints, and cut overall BOM costs for space-constrained IoT applications.
GlobalFoundries announced its 7nm-class FDX Fusion FD-SOI node, marking a strategic return to advanced semiconductor manufacturing for low-power edge AI, automotive, and industrial applications. This development gives hardware teams an alternative planar technology option combining dense digital logic with robust RF, analog, and embedded memory capabilities.
Renesas outlines how humanoid robotics and physical AI require tight integration across brain compute, sensing, actuation, and power management domains. Hardware teams must balance deterministic real-time control with strict latency and thermal limits for next-generation edge systems.
Humanoid robot designs are pivoting toward centralized compute architectures to handle complex AI workloads and sensor arrays. Adopting standardized embedded interfaces will be crucial to reduce integration complexity and power consumption in commercial production.
Humanoid robot designs are pivoting toward centralized compute architectures to handle complex AI workloads and sensor arrays. Adopting standardized embedded interfaces will be crucial to reduce integration complexity and power consumption in commercial production.
