US corporations are increasingly adopting open-source and open-weight AI models to reduce operational costs. According to Financial Times, mentions of these models by US company executives in public statements have increased sixfold year-over-year between August and September 2026.
Data from Vercel indicates that the share of open models in token usage reached 56% by August, compared to 7% in December 2025. Chinese AI laboratories are now among the top ten most active in terms of token usage intensity, according to OpenRouter statistics.
Tinder is shifting secondary internal requests to open-weight models to curb rising costs. Match Group reported that annual AI expenses reached $10 million by July, a tenfold increase since January. The company aims to reduce reliance on premium models like Astra and Claude.
AT&T currently handles 45 billion tokens daily and uses open models for 40% of its AI workloads. The company plans to increase this share to 70% within a year. Internal tests show that open models can match or exceed the performance of proprietary alternatives.
The shift toward open models is driven by the need for cost efficiency and enhanced data security through local infrastructure deployment. In response to this trend, OpenAI and Anthropic have introduced more affordable versions of their proprietary models.
