Apple is reportedly developing Apple’s own AI servers powered by its custom chips, according to The Information. The company has also discussed using Nvidia networking technology to connect them. Apple reportedly plans to sell the system to AI developers, businesses and governments, therefore marking a potential return to the commercial server market after more than a decade.
Up to Four M8 Ultra Chips
Two versions of the server are reportedly under consideration. The smaller configuration would combine two upcoming M8 Ultra chips, while a larger version would use four. Additionally, Apple has discussed connecting the chips through Nvidia’s NVLink Fusion, a high-speed interconnect technology that Nvidia also licenses to Qualcomm, Arm, Fujitsu and Marvell.
However, the project remains under development. The server is not expected to reach the market before 2029, and Apple could still cancel the project or launch it without Nvidia technology. Meanwhile, shares in Apple and Nvidia showed little movement following the report.
Apple Has Not Sold a Server Since 2011
Apple already builds servers for its own operations, and its Private Cloud Compute infrastructure initially relied on Apple silicon. In June, the company expanded some of its AI infrastructure through Google Cloud using Nvidia GPUs and Intel processors. Still, Apple has not sold a dedicated server product since 2011, when it discontinued the Xserve after nearly nine years on the market. This reported system would therefore represent a return to selling server hardware directly to outside customers.
Governments Could Become an Important Market
The inclusion of governments among Apple’s potential customers comes as Europe increases its focus on technological sovereignty. Furthermore, this interest aligns with broader shifts in how governments approach AI infrastructure procurement.
Hardware Falls Outside the Proposed Cloud Rules
The European Commission’s proposed Cloud and AI Development Act focuses on cloud services rather than server hardware. Instead, a direct Apple server sale would not itself fall under those cloud-service requirements, though governments can still impose their own procurement conditions. That distinction separates sovereign technology from sovereign operation, since a government-owned server running inside its own facilities may face a different sovereignty assessment than an outside cloud provider.
For Apple, this could make locally operated enterprise hardware appealing to governments seeking greater control over AI infrastructure. The European Commission proposed the Cloud and AI Development Act in June, introducing four sovereignty assurance levels for public-sector cloud services. At the highest level, providers must maintain extensive control over their software supply chains.
Europe has already shown that non-European technology does not automatically block a service from meeting sovereignty requirements. In April, the Commission awarded a six-year sovereign cloud framework worth up to €180 million to four European providers, including a group led by Proximus. For now, Apple’s server project remains unconfirmed, with no commercial launch expected before 2029.
