AT&T shifted from closed AI models to open-source alternatives to cut costs, with open models now accounting for 40% of its AI use and projected to reach 60%. Chief Data and AI Officer Andy Markus said the company is saving up to 80% on AI expenses. Similar trends are seen at Airbnb and Deloitte, with open models making up 58% of AI use across U.S. users via Open Router, up from 10% a year ago.

Nvidia's $12.9 billion acquisition of Hugging Face underscores the growing importance of open AI. Open models come in open-source and open-weights forms, with many powerful models from Chinese firms like Moonshot AI, Deep Seek, and Alibaba offering 80-90% of the performance of Open AI and Anthropic models at just 20% of the cost. Jerry Tang of Atlas Cloud likened closed models to Maseratis and open models to Mazdas—practical and cost-effective.

The rise of open AI challenges the business models of Open AI and Anthropic, which rely on subscription fees amid high R&D costs. The trend reflects a broader debate on AI control, with figures like Dario Amodei advocating regulation and Jensen Huang and Mark Zuckerberg supporting open-source accessibility.