Woofun AI reports that Apple is redirecting its artificial intelligence operations toward NVIDIA GPUs, following performance shortfalls of its proprietary M2 Ultra chip during advanced AI workloads. The company’s attempt to utilize internal servers for Google’s Gemini model to enhance Siri capabilities proved insufficient, necessitating the use of NVIDIA hardware hosted on Google Cloud.

Furthermore, development of the next-generation "Baltra" server chip has encountered delays, and the M7 Ultra, designed to compete with NVIDIA’s Blackwell architecture, is not expected to arrive before 2029.