TechEx: Power, Data Centers and Security Shape AI Rollouts

At TechEx North America in Santa Clara, day-one panels across Edge Computing, IoT, the Data Center Congress and Cyber Security and Cloud outlined practical barriers to moving AI from pilots into production.

The Edge Computing sessions, chaired by Ed Doran of the Edge AI Foundation, addressed running AI close to machines. Representatives from Akamai, Spectro Cloud, Scylos, TÜV Rheinland, the OPC Foundation and Schneider Electric discussed latency, distributed inference, agentic network operations and immutable edge infrastructure. Panelists noted that local decision-making can reduce latency and dependence on central cloud services, and they raised questions about observability, operator control and changing risk profiles when compute shifts to the edge.

IoT Tech Expo sessions on Industrial IoT and digital twins focused on the gap between demonstrations and durable deployments. Speakers from Rockwell Automation, Ford, Siemens, LG CNS and Boston Dynamics examined what they called “pilot purgatory,” where proofs of concept stall when they encounter older equipment, legacy software or unclear ownership of outputs. Presenters argued that digital twins should function as operational models that support maintenance and decisions rather than only as visual replicas, and they said AI must integrate into daily workflows to avoid unused dashboards.

Data Center Congress roundtables covered construction, procurement, cooling and the network spine that dense AI compute requires. Santa Clara officials described the city’s data center trajectory. Multiple speakers pointed to pressures on power, water, land and permitting and emphasized that data center capacity and related infrastructure take years to plan and build, while AI economics change quickly.

Cybersecurity sessions examined security culture, compliance, ransomware, shadow AI and data exfiltration. Panelists highlighted risks from employees using AI tools inside workflows without approval or logging, which they said collapses data governance and cyber governance into a single problem. Discussions also covered legacy systems, open-source dependencies and the relationship between CISOs and the C-suite.

Across tracks, presenters emphasized links among technical layers. Edge and IoT panels raised the same concerns as cybersecurity sessions about modern intelligence meeting older plant systems, and data center speakers pointed to power and cooling as nonsoftware bottlenecks that affect timelines and costs.

Panels concentrated on deployment discipline, scaling strategies for multi-site operations, and designing systems that operators and machines can use in daily work. Day-one programming presented specific infrastructure and security issues that industry participants identified as factors affecting whether AI projects move from pilot to steady-state production.