Social Implementation of AI and Accelerating Infrastructure Investment

◆2025 marked the year AI took its first step toward collaborating with humans. Technologically, advancements in AI model sophistication and multimodal capabilities enabled the processing of more complex tasks and expanded the types of information AI could handle, broadening the range of tasks AI could perform. Furthermore, the emergence of lightweight yet high-performance AI models increased options for corporate adoption. Against this backdrop of improved AI model performance, proof-of-concept experiments focused on social implementation intensified even in practical domains like AI agents and physical AI.

◆From 2026 onward, the social implementation of these AI agents and physical AI are expected to advance further. While challenges remain in areas such as permission management, security, and institutional frameworks, implementation is likely to progress first in routine tasks with relatively minor impacts and in closed environments like factories. The year 2026 is anticipated to mark the turning point when the entire economy transitions toward an AI-based industrial structure.

◆Investment in AI infrastructure such as data centers by various AI companies has been increasing year by year. While some voices suggest this represents overinvestment, this trend is expected to continue into 2026. While technological efficiencies are improving, the performance gains of AI models are outpacing these efficiencies, leading to strained computational resources. Therefore, it is difficult to immediately label this as overinvestment. On the revenue side, however, corporate AI adoption is still in the trial phase, and monetization is likely to take time, necessitating close monitoring of developments.

◆Monetization depends not only on the performance of AI models but also on how widely and deeply they are adopted and established within companies. For the time being, it is crucial to carefully assess the expansion of implementation areas and usage track records, evaluating them from a long-term perspective rather than as short-term trends. The utilization of specialized AI models is also advancing, and their combined use with optimization tailored to specific applications is expected to grow. Beyond technological progress and infrastructure development, the key to AI market growth will likely be whether the level of adoption within companies translates into sustained demand.