The 2025 Gartner Hype Cycle for Generative AI (GenAI) provides critical insights for IT leaders navigating a rapidly evolving and often overhyped landscape of GenAI innovations. The report projects that by 2028, more than 95% of enterprises will have integrated generative AI APIs, models, or deployed GenAI applications in their production environments. This underscores the urgent need for organizations to move beyond early proofs of concept and invest strategically in technologies that create tangible value and align with organizational objectives.
Four critical technology areas
Gartner identifies four critical technology areas shaping the GenAI Hype Cycle and warranting strategic investment. The first is GenAI models, where large language models (LLMs) remain the cornerstone and the most mature technology. These foundation models are highly customizable for a wide range of use cases. However, other model types, such as open-source LLMs, domain-specialized GenAI models, and large reasoning models, are rapidly emerging as viable alternatives. Multimodal generative AI is cited as a sample technology in this category, promising more powerful and faster AI results.
AI engineering for scaling
The second key area is AI engineering, which becomes critical as organizations prepare to scale up their GenAI programs. It encompasses the growing ecosystem of tools and techniques designed to build, govern, and customize GenAI applications. AI engineering ensures that GenAI applications serve the organization's strategy, providing frameworks for application orchestration, hallucination reduction, misinformation mitigation, and regulatory compliance. AI TRiSM (Trust, Risk and Security Management) is presented as a sample technology, focused on the safe and effective use of AI.
AI agents and applications
The third area covers AI agents, applications, and use cases. GenAI virtual assistants, such as ChatGPT, are well-known examples leveraging LLMs for advanced capabilities. The long-term vision is to use AI agents to automate complex, multi-step processes at scale, in order to increase productivity, reduce operational costs, and improve customer experience. IA agentique, which perceives, decides, and acts autonomously or semi-autonomously to achieve goals, represents a fundamental shift from passive chatbots toward more interactive, value-creating AI systems. Embodied AI is cited as a sample technology in this space.
Infrastructure and enabling techniques
Finally, infrastructure and enabling techniques form the fourth critical area. The evolution of GenAI relies on a combination of new techniques and established AI practices. Self-supervised learning, for example, reduces the need for massive labeled training datasets and finds applications in fields such as autonomous driving and medical diagnosis, with growing interest across all industries. Specialized infrastructure, including AI chips and tooling, is gaining traction for its role in improving efficiency and reducing model training and inference costs. AI supercomputing is highlighted as a sample technology in this category.
Strategic direction
In essence, the 2025 GenAI Hype Cycle serves as a guide for IT leaders to make informed investment decisions, move beyond the hype, and successfully integrate generative AI into enterprise strategies to drive innovation and create business value.