Introduction

On May 10, 2025, multiple reports from supply chain analysts and industry insiders suggested that Apple is preparing to license its custom M4 Neural Core to select OEM partners. Originally designed for Apple’s upcoming Mac and iPad Pro models, the M4 Neural Core integrates a 60 TOPS NPU with secure memory enclaves, optimized for real-time, on-device AI inference.

If confirmed, this would mark Apple’s first foray into silicon IP licensing—traditionally a model used by firms like ARM and Imagination Technologies—and signal a dramatic shift in Apple’s hardware strategy.

“This could be Apple’s Quiet AI Play: turning its silicon into a standard across premium devices without opening iOS,” said Ben Thompson of Stratechery in a post on X.¹

The rumored deal would allow premium laptop and embedded device makers to integrate the M4 Neural Core as a co-processor, enabling edge inference compatible with Apple’s developer tools and privacy framework.

Why it matters now

• Apple has stayed out of the AI platform wars—this move could create a third ecosystem beyond OpenAI and Google.
• Licensing AI silicon expands Apple’s reach without compromising OS exclusivity.
• Secure, on-device AI is now a differentiator in health, finance, and industrial markets.

Call-out: Apple may disrupt AI from the edge, not the cloud

The M4 Neural Core reportedly uses 30% less power than Qualcomm’s Hexagon NPU while supporting 128K token windows and multimodal fusion on-device.

Business implications

OEMs seeking to offer Apple-like user experiences, particularly in privacy-sensitive contexts, could now do so without becoming iOS vendors. Apple would gain leverage in AI infrastructure while maintaining platform control.

This also pressures traditional NPU providers to improve software stacks. Apple’s licensing model reportedly includes access to CoreML-compiled runtime and Apple’s encrypted model distribution format.

Looking ahead

Apple has not confirmed the reports, but analysts expect a June announcement at WWDC 2025. If true, M4 Neural Core adoption could begin with 2026 devices from select Japanese and Korean laptop makers.

Gartner notes that by 2029, 50% of enterprise AI endpoints will require secure, tamper-proof inference. Apple’s move could preempt regulation while setting new expectations for performance and privacy at the edge.

The upshot: If Apple licenses its M4 Neural Core, it won’t just be selling chips but a philosophy of trusted, local intelligence. And that could be the most disruptive play of all.

––––––––––––––––––––––––––––
¹ Ben Thompson, Stratechery commentary via X, May 10, 2025.