Gartner: More than half of generative AI projects will exceed budgets due to poor design

Gartner says more than half of generative AI projects could run over budget because of poor architecture and weak operational execution.

The firm expects many companies to abandon custom model builds as costs and complexity rise, especially for in-house fine-tuning efforts.

Coding assistants and content-generation apps are moving into growth, while domain-specific models and AI agent protocols like MCP and A2A are still early.

Longer-term bets include misinformation security and world models, and Gartner warns open-source AI competition may shift toward Chinese players.