As manufacturers accelerate their use of AI, the limiting factors are no longer theoretical. Energy availability, infrastructure resilience and deployment reality are now shaping what can be achieved. This perspective helps separate genuine progress from hype.
From agentic AI and defence‑led validation to energy constraints and edge‑first architectures, emerging technologies are advancing quickly. The real question for manufacturers is not what is possible, but what is deployable under real‑world constraints.
Pablos Holman has built a career identifying technologies that are technically possible but operationally misunderstood. With experience spanning aerospace, energy, automation and applied science, his work focuses on translating complex ideas into systems that must perform reliably, economically and at scale.
His keynote and subsequent discussions at 3DEXPERIENCE World earlier this year offered a pragmatic counterpoint to much of the surrounding AI conversation, grounding emerging technology in the realities of deployment, infrastructure and operational risk.
Holman’s recent commentary offers a counterbalance to technology hype. Rather than predicting sweeping transformation, he emphasises that the success or failure of emerging technologies is increasingly determined not by capability alone, but by how well they operate within real organisational and infrastructure constraints.
A defining feature of the current technology cycle is which constraints matter most. Energy availability, infrastructure resilience and governance requirements are no longer abstract considerations. They are active planning variables influencing where and how AI‑enabled systems are deployed.
Agentic AI systems represent one area of real progress. Designed to perform defined tasks and improve over time, these systems can support workflow preparation, quality monitoring and exception handling. Their value depends entirely on clearly defined objectives, boundaries and escalation paths.
Defence continues to act as a proving ground for advanced technologies. Tools validated under defence constraints tend to flow into civilian manufacturing within a few years, offering a roadmap for adoption rather than a call for early experimentation.
Energy has emerged as a silent but decisive constraint. AI‑heavy workloads depend on reliable power, and manufacturers are increasingly factoring energy stability into digital and automation strategies. Incremental deployment and edge‑first architectures are becoming practical responses in an uncertain operating environment.
Holman also cautions against assuming governance challenges have been solved by scale. In manufacturing, where IP and process knowledge are core assets, sensitive workloads should remain close to where value is created, with selective cloud integration applied deliberately.
The emerging technology landscape remains noisy, but the signal is clear. Manufacturers that sequence adoption carefully, respect operational limits and anchor decisions in real KPIs will be best placed to extract lasting value as conditions continue to tighten through 2026. Those that chase novelty without discipline are likely to be disappointed.