As we close out 2025, it’s worth taking a hard look at what actually happened with AI in manufacturing versus what we all hoped would happen. Spoiler alert: reality was a bit more humble than the hype suggested.
The biggest story coming out of this year isn’t about breakthrough AI implementations revolutionizing factory floors overnight. Instead, manufacturers discovered that while AI significantly improved their awareness and decision support across forecasting, logistics, and supplier risk scoring, it didn’t magically eliminate uncertainty or deliver the automatic resilience many were banking on. This reality check is probably the most valuable lesson of 2025 for our industry.
The Infrastructure Building Blocks Keep Evolving
While AI was learning to walk before it could run, the foundational technologies supporting AI in manufacturing continued their steady march forward. The Ethernet Alliance just dropped their 2026 roadmap, and it’s laser-focused on next-generation high-performance, AI-driven networking. This isn’t just about faster data pipes – it’s about building the nervous system that will eventually make those AI promises more achievable.
What caught my attention is how automotive applications are driving much of this innovation. The partnership between IAR and SiFive expanding RISC-V adoption in automotive applications might seem tangential to traditional manufacturing, but it’s not. The automotive industry has become a testing ground for industrial automation technologies, and what works in automotive often migrates to broader manufacturing applications.
What This Means for Your Operations
Here’s my take: the manufacturers who treated 2025 as a learning year rather than expecting immediate transformation are probably in the best position moving forward. AI in manufacturing is proving to be more about augmenting human decision-making than replacing it entirely. The companies that figured out how to use AI for better visibility into their processes, rather than expecting it to run the show, likely saw real ROI.
The Ethernet roadmap developments suggest we’re building toward much more capable industrial networks, which will eventually support more sophisticated AI applications. But the key word is “eventually.” The infrastructure improvements happening now are setting the stage for the next wave of capabilities.
Looking ahead, I suspect 2026 will be about refining AI applications based on 2025’s lessons learned, while the networking and processing infrastructure continues maturing in the background. The question for your operations: are you building AI literacy and infrastructure simultaneously, or are you still waiting for the magic bullet?