According to technology media The Information, Apple has encountered challenges in developing its self-developed AI server chip. The AI server chip, codenamed Baltra, may not be launched as scheduled in 2026. The chip is based on Apple's M2 Ultra chip, but the latest information indicates that it lacks sufficient performance in handling AI tasks and has been delayed.

Chip Technology (2)

There are fundamental contradictions between mobile device requirements and server needs

According to insiders, Apple's chip team has long focused on developing low-power designs for mobile devices, while AI server chips usually require high power consumption, high concurrency, and large bandwidth. These are entirely different technical paths, leading to numerous challenges during the transition. This "genetic mismatch" has become the core problem facing Apple's self-developed server chip.

This year, Apple tested its self-developed server chip, but it was unable to handle the massive Siri AI computing load. Eventually, Apple decided to transfer the high-load tasks of the new iOS 27 version of Siri AI to Google Cloud Platform, using NVIDIA GPUs to complete these tasks. This means that Apple will still rely on external infrastructure for AI computing power in the short term, and the breakthrough path for its self-developed server chip will take time.