AT&T is planning a significant shift in its artificial intelligence strategy, aiming to have open-weight AI models power 70% to 80% of its total AI usage over time. Currently, these open models account for about 25% of the company's AI workloads, which include tasks like network management. The move is part of a broader trend among corporations to adopt open alternatives to proprietary AI systems to reduce costs and maintain control over sensitive data.
The telecommunications giant processes an average of 45 billion AI tokens daily. By moving away from a heavy reliance on large, general-purpose proprietary models, AT&T expects to gain more flexibility and significantly lower expenses. The company has already been focused on efficiency, developing an "AI Gateway" that routes tasks to the most cost-effective models available.