AT&T Aims to Raise Open-Source Model Share to 60%-70% to Control AI Costs

AT&T is relying on open-source large models to reduce AI operating costs. Vice President Mark Austin said the company plans to keep spending flat over the next several years while controlling costs tied to employee use of closed-source models from providers such as Anthropic and OpenAI, mainly by adopting more open-source models like Nvidia's Nemotron. Austin oversees the AI tools that AT&T makes available to 100,000 employees, covering code writing, financial analysis, human-resources support tools and customer-service support systems. Apart from some consumer-facing AI functions, the vast majority of its AI uses are internal business operations.