On the Automation World podcast, Mitsubishi Electric Automation robotics marketing manager Patrick Varley discusses how smaller manufacturers are adopting pre-programmed robotic machine-tending cells under robots-as-a-service contracts with zero upfront CapEx. The pitch targets shops losing production hours to operator turnover on repetitive loading work. Because the guest markets robotics for Mitsubishi Electric, treat it as a vendor view on financing structures rather than an independent shop benchmark.
Neura Robotics, E.ON and enviaM will train robots at TUM RoboGym in Munich on human demonstrations, robot data and simulation, then test inspection, substation monitoring and site security with Mitnetz Strom under real conditions in 2027. Mitnetz runs about 71,000 km of distribution lines, and the utilities keep their data and trained skills. This is a partnership announcement, so the live grid tests are still a year out.
At BIXPO, KEPCO demonstrated a virtual grid that models plants, substations and lines to preview where new renewables or a data center would overload the system, with apps planned for Jeju and Honam in 2027. It sat beside dynamic line rating pilots and LS Electric's 800 V DC to the rack work with LG Uplus. Validation is still planned for this year, but spare capacity on existing wires is being treated as inventory.
Auto Ontology is an Apache-2.0 framework that drafts an ontology from database metadata, grounds agent SQL in certified terms, and exposes read-only MCP tools for Databricks, Postgres, Snowflake and DuckDB. NVIDIA's speed claims are its own and no accuracy scores are published, and the README notes passwords can sit unencrypted without Vault. The common failure for plant data agents is a wrong metric answered fast, and this puts the definitions first.
For data and OT analytics teams: keeping agents on certified metric definitions before they query plant or energy data.
