📡 AT&T now routes 40% of employee AI requests through open source models, with a target of 60 to 70% in the next few yea...

📡 AT&T now routes 40% of employee AI requests through open source models, with a target of 60 to 70% in the next few years.The pivot cut coding and other advanced-task costs by up to 56% while output quality slipped only 2%, using a cache-aware router that sends each query to the cheapest capable model.The open models are American: Nvidia, Google's Gemma, and Meta's Llama. AT&T also fine-tuned OTel 2.0, a telecom-specific model built on Gemma 4 31B.Meanwhile US hyperscalers are on track for roughly $700 to $800 billion in AI capex this year, about 7 to 8 times China's estimated $105 billion.Read more:https://www.pymnts.com/news/artificial-intelligence/2026/att-slashes-ai-costs-by-adopting-model-routers-and-open-source/https://www.fierce-network.com/cloud/open-models-are-driving-atts-ai-tokenomics-strategyhttps://www.scmp.com/tech/big-tech/article/3352068/chinese-firms-face-pressure-ai-investments-us-peers-spending-keeps-soaring#AI #opensource #ATT