Apple is preparing an AI M8 Ultra server for professional clients

The demand for Macs for AI applications is growing, and revenues from this range are rising. In this context, Apple is developing an inference server equipped with M8 Ultra chips, aimed at developers, businesses, and governments. The project, still in development, is not expected to launch before 2029 and could potentially bypass Nvidia technologies if necessary.

Mac Purchases by AI Labs and $10.4 Billion in Quarterly Revenue

AI laboratories such as OpenAI and Anthropic are currently purchasing Mac Minis and Mac Studios in large quantities for tasks related to artificial intelligence. Revenues generated by Macs increased by nearly 29% in the last quarter, reaching $10.4 billion. According to sources close to the project, the development of the server has recently gained momentum.

An Inference Server Based on M8 Ultra, in Two Configurations

Apple is developing a server for enterprises equipped with its own chips, intended for AI developers, businesses, and governments. This server is designed for inference, meaning the execution of already trained AI models. Two versions are being considered, incorporating either two or four M8 Ultra chips.

NVLink Considered, Timeline Beyond 2029, and Uncertainties

Apple is considering using Nvidia's NVLink Fusion technology to connect the server's chips, which would allow for rapid communication between components in data centers. NVLink was originally designed for Nvidia chips but is now accessible to other hardware. The server is not expected to launch before 2029, and it is possible that the project could be halted or continued without relying on Nvidia technology. John Ternus, who recently took the helm at Apple, had supported this project about a year ago while still serving as head of hardware, believing that the AI computing market presented an opportunity.