Apple M7 Ultra: Up to 1.5 TB of memory to power the next generation of AI
Apple is not just preparing a new generation of faster chips. According to the latest information from Bloomberg, the Cupertino-based company is designing the M7 Ultra as a dedicated engine for artificial intelligence.
Expected to debut in 2028, this processor aims for performance that could compete, in certain applications, with Nvidia’s AI accelerators rather than with traditional processors designed for personal computers.
This evolution speaks volumes about Apple’s priorities.
Up to 1.5 TB of unified memory for running massive AI models
The most striking information concerns memory capacity. According to Mark Gurman from Bloomberg, the M7 Ultra could support up to 1.5 terabytes of unified memory, nearly double the maximum expected capacity of the upcoming M5 Ultra.
Even within the realm of professional workstations, such a volume remains exceptional. For Apple, the goal is clear: to enable local execution of increasingly large artificial intelligence models, without constantly relying on the cloud or external storage solutions.
This architecture aligns with Apple’s cherished unified memory philosophy, where the CPU, GPU, and Neural Engine access the same memory space to minimize data transfers and enhance performance.
The global memory shortage could, however, limit ambitions
This maximum capacity remains theoretical. Bloomberg notes that the availability of configurations reaching 1.5 TB will rely directly on the evolution of the memory chip market. Ongoing tensions regarding high-capacity DRAM modules make their supply more challenging and significantly increase their cost.
In other words, even if the M7 Ultra is technically capable of managing such a memory volume, Apple may be forced to offer less ambitious configurations if industrial constraints persist leading up to its launch.
Much more than a chip for Macs
The M7 Ultra won’t be exclusively aimed at future high-end Macs. According to Bloomberg, Apple also plans to use this architecture to power a new generation of servers dedicated to Apple Intelligence. An initial infrastructure based on the M5 Ultra would be deployed before this deadline, while a platform utilizing the M7 Ultra is expected to arrive around 2029 to support AI processing in the cloud.
The objective is to ensure a seamless connection between processes performed directly on Apple devices and those that require more server-side power, while maintaining complete control over the entire hardware stack.
Apple is gradually shifting its philosophy
Since the introduction of the first Apple Silicon chips, progress has primarily focused on CPU performance, GPU capabilities, and battery life. With the M7 Ultra, it seems the logic is evolving. Artificial intelligence is poised to become the main design driver for future Apple chips. Rather than simply adding AI accelerators to a traditional architecture, the company is now designing its silicon around the needs of generative models.
This approach gradually aligns Apple with the world of AI infrastructures, where architectures like Nvidia Blackwell currently dominate the training and inference of the most advanced models.
Apple is likely not looking to compete with Nvidia in the data center market but instead aims to develop a platform optimized for its own ecosystem, capable of running Apple Intelligence effectively on both devices and in its cloud infrastructure.
A new milestone in the race for AI
If this roadmap holds, the M7 Ultra could represent much more than just an annual evolution of Apple Silicon. It would symbolize Apple’s entry into a new phase where hardware is primarily designed to meet the demands of generative artificial intelligence. As models become larger and more complex, the amount of available memory becomes almost as strategic as the computing power itself.
This evolution indicates that Apple is no longer just focused on creating the best chips for Macs. The company is now building the hardware foundations for its next generation of AI services, with the ambition of maintaining complete control over its ecosystem from end to end.


