Advantech is pushing further into the physical AI space tools designed to deploy and manage AI models across diverse edge hardware at scale. At the company's recent Edge AI Conference in Taipei, executives from Advantech, Nvidia, Qualcomm and Bosch Rexroth made the case that edge AI is poised for major growth, projecting the U.S. market alone to reach $197 billion by 2034.

For process control engineers, the more relevant takeaways may be the cautionary ones: panelists pointed to persistent gaps between OT and IT architectures, the difficulty of achieving physical-world model accuracy comparable to simulation tools used in chip design, and a sobering reminder that 40% of AI projects fail when applied to pre-existing processes rather than designed around them. For continuous process operations evaluating where edge AI and physical AI fit into existing control architectures, those integration challenges are worth weighing alongside the upside.

Keith Larson was in Taipei for the conference. You can read his report on our sister publication, Automation World, here.