# gartner-hype-cycle-genai-2025-critical-innovations-2025-07-29

## Veille

Gartner Hype Cycle GenAI 2025 - Critical innovations - LLMs - AI engineering - Agentic AI - Infrastructure

## Titre Article

The 2025 Hype Cycle for GenAI Highlights Critical Innovations

## Date

2025-07-29

## URL

https://www.gartner.com/en/articles/hype-cycle-for-genai

## Keywords

Generative AI, GenAI, Hype Cycle, AI models, Large Language Models (LLMs), AI engineering, AI agents, AI applications, AI use cases, Infrastructure, Self-supervised learning, AI supercomputing, Multimodal generative AI, AI TRiSM, Agentic AI, Embodied AI, strategic investment

## Authors

Arun Chandrasekaran

## Ton

**Profile:** Professional-analytical | Research-institutional | Analytical-prescriptive | Expert

Chandrasekaran (Gartner) adopts an authoritative market-research tone combining trend analysis with strategic recommendations. The Hype Cycle methodology (peak of inflated expectations, trough of disillusionment, plateau of productivity) provides a systematic framework for assessing technology maturity. Specialized terminology (trigger innovations, mainstream adoption horizons, transformational impact) targets executive leadership and enterprise architects. Precise time-based quantifications (2-5 year, 5-10 year adoption horizons) anchor the predictions. A measured, professional tone avoids both hype and excessive conservatism. Typical of Gartner research publications, positioning the firm as a trusted advisor on enterprise technology strategy with proprietary frameworks recognized across the industry.

## Pense-betes

- **By 2028**: **more than 95% of enterprises** are expected to use GenAI APIs, models, or GenAI applications deployed in production
- **4 critical technology areas**: GenAI models, AI engineering, AI agents/applications/use cases, infrastructure/enabling techniques
- **LLMs = cornerstone**: the most mature technology, highly customizable
- **Open-source LLMs, domain-specialized models, large reasoning models** are emerging as alternatives
- **AI engineering is critical** for scaling: tools/techniques to build, govern, customize GenAI applications
- **AI TRiSM** (Trust, Risk and Security Management): ensuring safe and effective use of AI
- **Shift from passive chatbots to IA agentique**: perceives, decides, acts autonomously/semi-autonomously
- **Infrastructure**: self-supervised learning, AI chips, specialized tools for efficiency and cost reduction
- **Moving beyond proofs of concept**: focus on building, governing, customizing for production
- **Multimodal generative AI, Embodied AI, AI supercomputing** = highlighted sample technologies

## RésuméDe400mots

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.

## GrapheDeConnaissance

- Gartner —publie→ Hype Cycle for Generative AI 2025 (DOCUMENT, 0.99)
- Arun Chandrasekaran —a_créé→ Hype Cycle for Generative AI 2025 (DOCUMENT, 0.98)
- Arun Chandrasekaran —travaille_chez→ Gartner (ORGANISATION, 0.97)
- Hype Cycle for Generative AI 2025 —prédit→ 95% des entreprises utiliseront la GenAI en production d'ici 2028 (AFFIRMATION, 0.96)
- LLMs —fait_partie_de→ GenAI models (CONCEPT, 0.95)
- AI TRiSM —améliore→ sécurité et gouvernance IA (CONCEPT, 0.93)
- IA agentique —remplace→ chatbots passifs (TECHNOLOGIE, 0.91)
- self-supervised learning —réduit→ besoin de données étiquetées (CONCEPT, 0.9)
- AI engineering —améliore→ déploiement à l'échelle GenAI (CONCEPT, 0.92)
- Multimodal generative AI —fait_partie_de→ GenAI models (CONCEPT, 0.94)
- Embodied AI —fait_partie_de→ AI agents (TECHNOLOGIE, 0.9)
- AI supercomputing —réduit→ coûts entraînement et inférence modèles (CONCEPT, 0.87)
- large reasoning models —concurrence→ LLMs (TECHNOLOGIE, 0.88)

---
Canonical: https://www.thekb.eu/en/fiches/gartner-hype-cycle-genai-2025-critical-innovations-2025-07-29/