Large on-device AI models, particularly large language models (LLMs), require substantial memory to store model weights and intermediate data. A 1.5TB unified memory pool would enable professionals to run models that previously required dedicated AI accelerators or cloud instances. Gurman has noted that the M7 Ultra's upgrades bring the chip close in performance to dedicated AI accelerators, including Nvidia's Blackwell .
Whether Apple actually ships a 1.5TB configuration depends on supply chain conditions. Gurman reports that the industry faces a memory chip shortage that makes such large modules expensive and difficult to source . Apple has reportedly already postponed the Mac Studio refresh because of memory chip supply issues and price increases
. The company has tested support for up to 768GB of unified memory in the M5 Ultra, but supply constraints could prevent it from launching with an option for that much memory
.
The M7 Ultra is expected sometime in 2028, following the base M7 chip in early 2027 and the M7 Pro/Max in late 2027 . Apple is reportedly skipping the M6 Pro and M6 Max entirely to fast-track the AI-focused M7 generation
.