Manufacturing Wire

Researchers from Ruhr University Bochum and the Max Planck Institute for Security and Privacy have published a systematic review of 187 publications spanning two decades of hardware reverse engineering (HRE) research. The paper, titled 'SoK: From Silicon to Netlist and Beyond,' examines HRE as a foundational discipline for establishing trust in semiconductor-based computing systems. The analysis maps the evolution of techniques used to deconstruct and analyze chip designs from physical silicon up through logical netlists.

Georgia Tech researchers published a technical paper identifying CPU bottlenecks as a primary performance limiter in multi-GPU systems running large language model inference workloads. The study characterizes how CPU-side processing constrains throughput even when GPU resources remain underutilized. This represents a systematic architectural finding with direct implications for how AI inference infrastructure is designed and provisioned.

Plant Engineering published an analysis by a Caterpillar Electric Power sales manager examining the transition from reactive to predictive power reliability management in industrial facilities. The piece addresses how unplanned outages, aging electrical infrastructure, and constrained maintenance resources trap operations teams in a perpetual firefighting cycle. The argument centers on service agreements and forecasting methodologies as a structural remedy to that cycle.

A SimScale survey finds that engineering teams actively using AI tools are demonstrating measurable productivity gains, with adoption accelerating across manufacturing and industrial sectors. The Germany-based simulation software provider positions proper AI integration as a competitive differentiator for engineering organizations. The data suggests the gap between early adopters and laggards is widening.

Siemens has released on-premises drivetrain monitoring software that runs on industrial PCs rather than cloud infrastructure. The system ingests vibration and analog sensor data locally, communicating via MQTT, gRPC, and OPC UA protocols. This positions it as a plant-floor-native condition monitoring solution without requiring external network connectivity for core analysis functions.

Bold Laser has introduced a UV precision cleaning system built around a 349 nm Nd:YLF laser operating as a Class 1 enclosure. The platform integrates machine vision and CAD/CAM controls to perform thin-film removal and surface preparation with high repeatability. The system targets applications requiring precise, localized cleaning without abrasive or chemical contact.

Kubotek Kosmos has released version 8.0 of its Model-Based Definition (MBD) utility software, adding saved cutting plane views, CUI (Controlled Unclassified Information) report markings, and expanded CAD format support including PTC Creo 12 and Siemens NX 2312. The update extends compatibility with current-generation CAD platforms used widely across discrete manufacturing and defense supply chains.

Wonderful PCB has expanded its reverse engineering service portfolio to include PCB cloning, firmware extraction, and design recovery targeting legacy, industrial, automotive, and IoT product segments. The service is positioned to address situations where original design documentation is unavailable or manufacturers need to recreate obsolete board-level assemblies. This expansion reflects growing demand for design recovery capabilities as aging industrial equipment faces end-of-life component and documentation challenges.

Perfect Batch is an industrial AI analytics platform targeting batch manufacturing processes, designed to identify and replicate 'golden batch' profiles from historical production data. The system differentiates itself from legacy SCADA and DCS approaches by replacing static alert thresholds and manual parameter settings with dynamic AI-driven optimization. The tool is positioned for sectors where batch consistency directly drives yield and quality outcomes, such as specialty chemicals, pharmaceuticals, and food processing.

Axtra3D has expanded its additive manufacturing platform with elastomer resins covering Shore 48A to 90A hardness range, alongside integrated wash, cure, and sensor-based process monitoring tools. The additions target production-grade flexible part applications where consistent mechanical properties are critical. The workflow tooling suggests a push toward tighter process control and repeatability in elastomeric additive workflows.

Semiconductor Engineering's weekly roundup covers several developments with direct manufacturing implications: GlobalFoundries has filed 11 patent lawsuits, quantum computing milestones are projected for 2029, new fab construction is underway across multiple regions, and helium atom beam lithography is emerging as a next-generation patterning technology. Additional items include a 2D materials roadmap, AI tooling cost reductions of roughly 90% for generative AI workloads, and Intel's latest security report.

Electrified Thermal Solutions has opened a new headquarters and production facility in the Boston area, specifically to manufacture its electrically conductive firebricks, branded as E-Bricks. The facility expands production capacity for these components, which serve as the core storage medium in the company's Joule Hive Thermal Battery system. The move signals a scaling of domestic manufacturing for industrial thermal energy storage hardware.

Food manufacturers are exploring AI, automation, and robotics to address regulatory compliance and food safety requirements, according to experts at a virtual Food Manufacturing Summit organized around Deloitte research. The discussion centered on how these technologies can be operationalized within food production environments to meet evolving U.S. regulatory standards. The summit highlighted practical applications rather than theoretical potential, framing technology adoption as a compliance and risk mitigation strategy.

Simon Fraser University and Queen's University have announced a partnership to develop a secure, domestically built supercomputing infrastructure intended to support Canada's artificial intelligence initiatives. The collaboration aims to reduce Canadian dependence on foreign computing resources for AI workloads. No specific hardware vendors or deployment timelines were disclosed in the announcement.

Ketjen Corp. and Saudi Aramco Technologies Company have signed a Joint Development Agreement to co-develop, test, and deploy next-generation Fluid Catalytic Cracking catalysts and additives. The collaboration targets performance improvements across Aramco's refinery network and affiliated operations. FCC units are central to converting heavy crude fractions into higher-value products including gasoline, diesel, and petrochemical feedstocks.

A new line of industrial bag filters has been introduced capable of handling gas volumes up to 880,000 actual cubic feet per minute (ACFM). The filter elements combine a fabric bag with a support cage mounted from the clean gas side, and the product line is designed to accommodate a broad range of dust types, particulate loadings, and flow rates. The offering targets large-scale industrial gas filtration applications across process industries.

Boston Micro Fabrication (BMF) has released the microArch S150 series, a compact high-resolution 3D printer offering 25 µm optical resolution with automated setup and HEPA13 filtration. The system is designed for deployment across lab, office, and industrial environments. The printer targets applications requiring microscale precision parts manufacturing.

Brooks Tractor, a Wisconsin-based dealer, has been designated as a Trimble Technology Outlet, authorizing it to sell and support Trimble's grade control, site positioning, and machine control correction services for John Deere earthmoving equipment. The partnership expands Trimble's dealer network in the upper Midwest, giving regional contractors and earthmoving operators access to precision machine guidance technology through an established local channel. Trimble's grade control systems use GPS and GNSS-based positioning to automate blade and bucket control to within centimeter-level accuracy.

NAVER Cloud is expanding its AI infrastructure in South Korea through a collaboration with AMD, deploying 6th-generation EPYC Venice CPUs alongside MI455X GPUs for AI training, inference workloads, and cloud service development. The partnership signals continued investment in high-performance compute infrastructure outside of dominant US-based hyperscalers. This builds on existing cooperation between the two companies in the Korean cloud market.

AI inference workloads are fundamentally restructuring data center network architecture, blurring the boundaries between memory and storage as latency and bandwidth demands intensify. Traditional tiered network designs are proving inadequate as inference pipelines require near-simultaneous access to large model weights and real-time data. The shift signals a hardware and topology overhaul across hyperscale and edge data center infrastructure.