Businesses are facing significantly higher costs for AI usage, with some companies like Uber exhausting their entire 2026 AI budget by April and others like OpenAI's Sam Altman noting that budgeting has become a "huge issue" this year. This surge in expenses is partly due to a shift from flat-fee structures to usage-based pricing by major AI providers, as well as the increasing popularity of computationally intensive AI agents.
In response to these budget-busting bills, companies are increasingly exploring alternatives to expensive frontier models. Open-source AI models, which can be hosted locally and are free to use, are gaining traction, especially among companies like banks and telecom firms for whom data security is paramount. A study supported by the Mozilla Foundation estimates that open-source models achieve about 90% of the performance of closed alternatives and could cut AI costs for businesses by up to 70%. Many companies are now benchmarking different models and building routing tools to select the best and most cost-effective option for specific tasks.
The competitive landscape is further shifting with the emergence of powerful and affordable open-weight models, particularly from Chinese firms such as Tencent, Xiaomi, DeepSeek, MiniMax, and Z.ai. These models are undercutting the economics of proprietary AI from U.S. firms. For instance, Z.ai's GLM-5.2 is an open-weight model that competes with Anthropic's latest offerings. Data from platforms like Hugging Face shows that open-source models are handling a significant portion of AI application volume, while closed models remain a higher-cost, premium layer. On OpenRouter, the top six most popular models are now open models from Chinese firms, with Anthropic's Claude Opus 4.7 trailing in seventh place.
Industry leaders are also warning against vendor lock-in and the concentration of power in a few AI providers. Microsoft CEO Satya Nadella has emphasized the importance of firms controlling their own "learning loop" rather than relying solely on single providers. Clem Delangue, CEO of Hugging Face, argues that keeping powerful models closed creates an "asymmetry of power" and that transparency through open models is crucial for safety and a level playing field. He notes that half of Fortune 500 firms are already using Hugging Face to deploy their own private and open-source models, indicating a growing trend towards ownership and customization.
