AT&T targets 60-70% open-source AI to cap OpenAI/Anthropic spend
AT&T processes 45B tokens daily internally and aims to shift 60-70% of AI workloads to open-source models (Nvidia NeMo, Llama, Gemma) running on its own data centers to keep found…
Meta's Llama models power 40% of AT&T's internal AI usage
Llama is a core component of AT&T's open-source AI strategy, with the telecom giant customizing and running Meta's models on its own data centers to replace paid API calls to fron…
Google's Gemma models adopted by AT&T for cost-sensitive AI tasks
Gemma is being evaluated and customized by AT&T alongside Llama and NeMo, showing Google's open models gaining enterprise traction as a cost-effective alternative to closed-source…
AT&T customizes Nvidia NeMo models for internal AI workloads
Nvidia's NeMo framework is being adopted by AT&T for customizing open-source models, demonstrating enterprise traction for Nvidia's software stack beyond GPUs as companies build s…
OpenAI rushes enterprise privacy features to counter open-source threat
OpenAI and Anthropic are racing to launch confidential computing and data privacy controls as a direct response to enterprises like AT&T shifting workloads to self-managed open-so…
Anthropic quickly followed OpenAI's enterprise privacy announcements, suggesting competitive pressure from open-source alternatives is forcing both frontier labs to prioritize dat…