Apple is working with Broadcom to develop its first AI-specific server chips.
According to a report from The Information, Apple is working with Broadcom on the chip’s networking technology.
The chip, codenamed Baltra, is expected to be ready for mass production by 2026 and is intended for internal use only.
Citing three people with direct knowledge of the project, The Information said Apple is planning to develop the chip using TSMC’s N3P advanced manufacturing process, the same process OpenAI and Nvidia are expected to use for their own AI chips.
While Apple currently does AI processing on high-performance chips originally designed for Macs, those chips were not expressly designed for that purpose and therefore aren’t as fast or energy efficient as those that have been, posing a challenge for the company as it plans to continue rolling out and scaling up its AI features and services.
Broadcom and Apple did not provide comment to The Information regarding its report.
Apple is not the only company that has opted to develop its own AI chips rather than relying on third-party GPUs from the likes of Nvidia.
Amazon, Google, Microsoft, and Meta have all developed their own AI chips and it has previously been reported that OpenAI was also working with Broadcom and TSMC to build its own AI inference chip. Earlier this month, Apple revealed it was also exploring using Amazon's Trainium2 chips for AI model pretraining.
According to a report from October 2024, the generative AI giant has secured manufacturing capacity with TSMC and hopes to have its first custom AI chips manufactured by 2026, although that timeline is subject to change.
