Issue link: https://iconnect007.uberflip.com/i/1542458
10 SMT007 MAGAZINE I JANUARY 2026 in 2026, in order of increasing scope and challenge: 1. Viable AI use cases will grow, and deploy- ment will expand, albeit not distributed evenly. AI technology is being used in real- world applications, affecting diverse industries and sectors. Productivity gains will kick in. 2. LLMs will advance, yet SLMs will prolifer- ate. 2025 was the year of Large Language Models (LLMs.) In 2026, SLMs will grow. A vast number of parameters that serve as the model's knowledge bank characterize LLMs. They require high computing power and par- allel computing (GPU). SLMs reduce the expense of inference by requiring a lower number of parameters, concentrating on a specific target task to add value. SLMs also facilitate deployment at the edge. They can run on CPU, inference-centric chips (such as LPU), AI-specialized ASICs, or customized chips. 3. Multimodality models will become mainstream. With the release of Google's Gemini 3 on Oct. 22, 2025, we see the model for advancements in reasoning, deep thinking, and processing and integrating multiple forms of data simultane- ously, including text, images, audio, and video. These allow for a more contextual and human- like comprehension. Advancements of these capabilities will continue to gain momentum. 4. Generative AI tools will be ubiquitous. In 2025, we enjoyed the convergence of search and chat. With fierce competition and robust investment, the market is expected to see a proliferation of multimodal AI tools, which will enhance end-user interaction and data pro- cessing by enabling the simultaneous han- dling of multiple input types (e.g., text, images, charts, graphs, and documents such as PDFs). This will lead to improved responses and has broad applications across various industries, including manufacturing, engineering, and document analysis (e.g., interpreting complex technical manuals or financial reports). 5. Connectivity and data quality will be crucial to building a "private" model at the edge. To be cost-effective and time-efficient, the "dis- tillation process" and other emerging tech- niques are taking good, large models to make small models smarter at domain-specific tasks. To leverage AI at the edge, robust 5G connec- tivity and data management are crucial. 6. There will be a demand for tools to reduce hallucinations. Blending text generation with information retrieval using Retrieval-Aug- mented Generation (RAG), such as the Chat- GPT Retrieval Plugin and other emerging techniques, will enhance the accuracy and rel- evance of AI-generated content. 7. AI Agent(s) will be deployed toward an agen- tic AI system. Achieving "Think, Plan, Reason, Adapt, and Act" autonomously while retaining memory by understanding the task, breaking it into steps, choosing the right tools, and exe- cuting and learning from feedback will be the key to success. This means, "Take real-world actions, not just answers." There have been several developments of AI agents and agen- tic AI workflow tools for simple or complex workflows, respectively (Table 1). 8. Electronics manufacturing will adopt and flourish in running AI workloads outside data centers. These will be closer to manufac- turing sites by integrating network, edge com- S M T PE R S PECT I V E S & PRO S PECT S