According to The Information, Apple is constructing an enterprise-grade server platform based on its upcoming M8 Ultra processors, with discussions underway about incorporating Nvidia's NVLink Fusion technology to bind multiple chips together. The target market spans artificial intelligence developers, corporate buyers and government agencies, with commercial availability pencilled in around 2029.

The timing intersects with Europe's regulatory evolution. In June, the European Commission unveiled its Cloud and AI Development Act proposal, which would categorise public cloud procurement based on how insulated a provider remains from external state control.

The Hardware Architecture

Two distinct configurations are under consideration: a smaller variant housing two M8 Ultra chips and a larger model containing four. Nvidia's NVLink Fusion interconnect—already licensed to Qualcomm, Arm, Fujitsu, Marvell and others—would serve as the glue between processors.

However, significant uncertainties remain. The project could be shelved entirely, or Apple might pursue development without Nvidia's networking layer. Market reaction to the report proved muted, with share prices for both companies barely shifting.

Apple's Server History and Current Operations

Apple already manufactures servers for internal use. Private Cloud Compute initially ran exclusively on Apple silicon until June, when the company expanded the service to encompass Google Cloud infrastructure running Nvidia GPUs and Intel processors.

The company has not sold servers to external customers since discontinuing the Xserve in 2011, ending a product line that had lasted nearly nine years.

Europe's Sovereignty Framework

The Commission's June proposal establishes a tiered approach to sovereignty assurance in public contracts. The uppermost tier, designated for defence and national security applications, mandates a European cybersecurity certificate and comprehensive control over every software element. Third countries cannot exercise authority over a provider's design, development or maintenance operations.

Yet Europe's actual procurement decisions reveal a more pragmatic stance than the regulatory language suggests. In April, a six-year, EUR 180M sovereign cloud framework was awarded to four consortia. One group, led by Proximus and encompassing both a Thales-Google partnership and Mistral AI, included American infrastructure.

At that time, the Commission stated: "Non-European technologies, operated within a strict and appropriate framework, can meet the minimum level of sovereignty required." Evaluation criteria spanned eight dimensions, ranging from legal exposure to supply chain transparency.

Sovereign Operation Versus Sovereign Technology

A crucial distinction emerges between sovereign operation and sovereign technology. A machine that a government owns, controls and operates within its own facilities presents a stronger sovereignty argument than a service delivered by an external operator.

The Cloud and AI Development Act itself targets cloud services rather than hardware, meaning a server sale would technically fall outside its scope. Nevertheless, the procurement philosophy underpinning the legislation shapes buyer expectations, and organisations establish their own contractual conditions.

Europe confronts a fundamental constraint that regulation alone cannot resolve. When the underlying silicon originates from American manufacturers regardless, no legal framework can conjure genuine technological sovereignty.

Source: The Next Web