Apple will release its own server processor for “Apple Intelligence”

Currently, the Apple Intelligence AI service uses a server infrastructure based on Apple’s own M2 Ultra chips. According to rumors, they will be replaced by M4 series chips next year. However, The Information now reports that Apple is developing a new server chip that will provide even higher performance for AI functions, planning to phase out the use of Mac chips in servers.

Image Source: MacRumors

According to sources, the new chip contains “multiple copies” of the Neural Engine to provide faster performance for processing AI workloads. It is believed that at least three companies are involved in the project to create a new server chip. While Apple is handling general design tasks for the chip, Broadcom is providing the networking technology for it, sources said. TSMC will be responsible for production – mass production of the new chip will start in 2026 using the third-generation 3-nm N3P process. The same technology is expected to be used to produce the A19 chips for the iPhone 17 next year.

Apple’s server chip will likely support some Apple Intelligence features that rely on server-based generation, such as Image Playground. Recall that Apple is pursuing a strategy in which core AI tasks are performed locally on a smartphone or computer. But complex tasks, which are usually in the minority, are processed on the server.

«Apple will likely use the new AI chip for inference, where the chip processes new data, such as a user’s description of a possible image, and applies it to models to generate output, such as creating the image itself,” writes The Information.

Apple announced that iOS 18.2, iPadOS 18.2 and macOS Sequoia 15.2 will launch today with additional Apple Intelligence features including Genmoji, Image Playground, Image Wand and ChatGPT integration in Siri. All iPhone 16 models also feature a new AI visual search feature called Visual Intelligence, which lets you identify objects in the real world using your smartphone’s camera.