This is particularly relevant for agentic AI. With machine-centric workflows, infrastructure needs to support systems that act, call tools, retrieve data, and make several intermediate decisions before producing an outcome. These workflows themselves create new latencies and can come with their own costs. They also place demand on observability and control, because the path between request and response isn’t a deterministic code call.
Data sovereignty adds another layer, one that’s of increased importance in territories like the Middle East and Europe. Requirements change by jurisdiction, sector, and individual policy. For enterprises in regulated industries or the public-sector anywhere in the world, the location of data and compute can be a determining factor in architecture.
Joseph also pointed out the differences that are present inside the same country, or even the same county. A remote facility, for example, may not support the same performance as a dense metropolitan location. “After all, you can’t go faster than the speed of light,” he said. From site to site, the differences in latency and network speed mean that the work always has to begin with local business expectations and the tolerated risk profile, and the proviso that every deployment (sometimes a few miles from another) won’t necessarily deliver the same results.
Glover isn’t arguing for a move to the edge, nor dismissing centralised AI compute and resources. Akamai’s position is that enterprise AI in traditional IoT environments requires a more carefully-considered architecture, with training, fine-tuning, post-training and inference placed accordingly, always to what the workload demands.
This meticulousness is what he terms a “continuum of compute”. At any reasonable scale, applications, users, factories, and devices are distributed, and the architecture constructed has to reflect that distribution.
For the attendees at the Edge Computing Expo, the practical takeaway is that proximity is becoming more important than power as AI deployments shift from training models to delivering real ROI in production at scale. Glover wants enterprise leaders to think more deeply about infrastructure for real-time AI in production.
Planning considerations are certainly less glamorous questions than announcements of the power of the latest, locally-hosted model, but they’re much closer to the operational reality for large enterprise companies. The edge is where performance, cost, and resilience can have a much greater effect on the business.
Check out Joseph Glover’s solo keynote presentation at the San Jose McEnery Convention Center on Monday 18th May at the Edge Computing Expo, part of TechEx North America 2026, May 18-19. Joseph will also be a panel member discussing the issues around, “Scaling Edge Deployments With Lessons from Real-World IoT Rollouts”, earlier in the day.
(Image source: Pixabay, under licence.)
Want to learn more about IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information.
IoT News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.