Data centres have increasingly become flashpoints of public controversy. Criticized for gobbling up massive amounts of energy and water, generating constant noise from industrial generators, and occupying vast swaths of land, mega-scale facilities face mounting backlash. Yet, they have long been viewed as the unshakeable foundation of online technology and the artificial intelligence boom.

A recent partnership between Silicon Valley rivals Apple and Meta, however, hints at a potential shift away from absolute reliance on massive cloud infrastructure.

The Rise of Open-Weight Models on Local Machines

Discussed on an episode of BBC News’ AI Decoded featuring AI safety researcher Nate Suarez, consumer technology expert Ben Hatton, and tech author Dr. Stephanie Hair, the tech giants recently launched a powerful new open-weight AI model designed to run directly on desktop hardware—specifically Apple computers—rather than relying entirely on remote cloud servers.

To understand the significance, consumer technology expert Ben Hatton breaks down the model landscape:

• Closed Models: Classic systems like ChatGPT or Claude require users to go through a developer’s website or app, sending prompts to the cloud for compute and processing.
• Open-Source Models: Fully open-source models provide free access to raw training data on the internet, allowing complete user manipulation.
• Open-Weight Models: Occupying the middle ground, open-weight models do not give away raw training data or inner workings, but they allow users to download the final weighted model directly to their computers for localized adjustments.

This new Meta model represents a major open-weight release from a major U.S. lab, bringing local AI capabilities into sharper focus.

A Strategic Business Move

According to Dr. Stephanie Hair, the Apple-Meta collaboration is driven as much by business strategy as it is by technology. With Apple operating under a new technical CEO, John Turnis, the company is looking to catch up in the competitive AI landscape while protecting its core hardware business. Meanwhile, Meta continues its aggressive all-in pivot toward artificial intelligence.

By making models open-weight and pairing with Apple’s cutting-edge hardware, the companies can pitch local execution as a major privacy win for consumers. As Ben Hatton notes, personal AI demands access to private user data, making on-device processing via watches, phones, laptops, and glasses an appealing alternative to uploading personal information to the cloud.

Enterprise vs. Consumer Reality: Is It Cheaper?

While local processing sounds ideal for privacy, running massive models locally comes with steep hardware barriers. Hatton points out that a model with hundreds of billions of parameters requires massive memory capacity—far exceeding standard consumer laptops. High-end enterprise setups require specialized, high-tier hardware running into tens of thousands of dollars. Consequently, for everyday consumers, heavy local AI compute remains a tool tailored for hardcore developers and enterprises rather than casual users.

Moreover, experts agree that data centres are far from obsolete. Trillions of dollars continue to pour into hyperscale data centre buildouts to support frontier models. Rather than replacing data centres entirely, the future likely points to a hybrid model: massive cloud infrastructure for high-power enterprise computing, running alongside local, device-based AI for everyday personal tasks.

Public Pushback and the “Sleep-Racing” Dilemma

Public opposition to data centres is not just about environmental or noise concerns; it also reflects a deeper societal anxiety over the trajectory of AI. As AI safety researcher Nate Suarez highlights, prominent figures in the tech industry have issued stark warnings about the systemic risks of advanced artificial intelligence. Protesting local data centre expansions has consequently become a tangible way for communities to push back against an accelerating and largely unregulated technological shift.

While authoritarian nations barrel forward with uninhibited infrastructure buildouts, democratic societies are using public friction to hold innovation to account. Whether humanity is sleepwalking—or “sleep-racing”—into technological disruption remains a pressing question as the infrastructure race continues.