Cedrik Neike of Siemens framed the bar at the AI with Purpose Summit 2025: industry needs multimodal industrial-grade foundation models built around the way machines, workflows and real-world constraints actually behave, not another dashboard. IBM's much-cited figure, repeated at IIoT World Manufacturing Day 2025 and on most conference stages since, is that around 90% of industrial data collected goes unused.
The buyer's bar has hardened to a triple test: industrial-grade reliability under IEC 62443 cyber, brownfield-ready against an existing Siemens, Rockwell, Aveva and SAP fleet, and decision-grade rather than visualisation-grade. Augury sells the result on a Guaranteed Diagnostics contract, Cognite's December 2025 Atlas AI release puts agentic outcomes inside customers like Aker BP and TotalEnergies, and Symbotic posted its first profitable fiscal year (FY25 EPS $1.02, $22 billion backlog) on a warehouse-as-a-service model.
Robots-as-a-service is now the language of every new RFP: the IFR projects the segment from $16 billion in 2025 to roughly $125 billion by 2034, with KUKA, ABB, Locus and Symbotic's GreenBox JV with SoftBank reorganising commercial models around it.