Follow the latest coverage, related explainers and connected technology stories.
Open-weight AI models give organizations greater flexibility for local deployment and reduce reliance on service providers, but they do not automatically mean that training data, code, or usage rights are available. Semiconductor Engineering’s analysis confirms that their impact extends to memory design, interconnects, security, and chip-design-tool workflows.
Cadence explains that physical AI depends not only on inference accuracy, but also on its ability to sense, make decisions, and execute actions within strict constraints on latency, energy, heat, reliability, and data movement. The article highlights the role of edge computing, memory, interconnects, and input/output interfaces in ensuring predictable system behavior.
An analysis published in Semiconductor Engineering argues that the future of semiconductor manufacturing will depend not only on shrinking components, but also on developing new materials through simulation, artificial intelligence, and early collaboration among chip manufacturers, equipment suppliers, and materials suppliers. The analysis presents examples including dry manufacturing processes, photonic interconnects, and gallium nitride substrates.
Operation at the 1-kilowatt level does not represent a single test condition, as risks vary according to the workload, heat location, operating duration, and power-delivery path. Semiconductor Engineering’s analysis argues that verification must cover the chip, package, test system, and field data—not merely the detection of electrical failures.
A sponsored opinion article argues that securing artificial intelligence systems that interact with the physical world cannot be limited to protecting networks or testing models; it must cover data, sensors, hardware, software, and system behavior during operation. This approach is becoming increasingly important with agentic systems that make decisions and adapt in real time to changing environments.
Humanoid robots need a combination of centralized and distributed computing to process sensor data and control movement, but their direct proximity to people increases the risks of hacking, from disrupting systems to causing physical harm. An analysis by Semiconductor Engineering examines the challenges of designing the computing architecture, protecting artificial intelligence models, and securing communications as well as voice and visual interaction.
imec presented two modules to improve the manufacturability of CFET transistors: a backside-contact structure that increased drive current and reduced access resistance to the bottom devices, and a dipole-middle gate structure for tuning threshold voltage within a suitable thermal budget. The results represent research progress, but do not yet demonstrate CFET readiness for large-scale commercial production.
Brian Bailey argues that introducing artificial intelligence into chip design verification imposes a new verification layer, because the system may produce answers that appear plausible or circumvent the objective instead of adhering to the specifications. He emphasizes that the promised productivity gains may be offset by extensive manual work to review everything the AI produces and its provenance.
A compilation of results from 80 semiconductor companies showed that AI-related demand growth extended from data center processors to memory, advanced packaging, testing and manufacturing equipment. Eleven companies recorded growth exceeding 100%, while results for phones and consumer electronics were more mixed.