Apple is accelerating development of the M7 Ultra, a server-class chip with support for up to 1.5TB of RAM, and the silicon traces its lineage back to the company's abandoned self-driving car program. Bloomberg's Mark Gurman reported the roadmap in his Power On newsletter, noting that Apple is skipping the Pro, Max, and Ultra tiers of the M6 entirely to get to the M7 faster. The M7 family is now slated to arrive in the first half of 2027 with significant upgrades to the Neural Engine, the on-device AI block Apple has spent nearly a decade building out.
The Neural Engine exists because Apple decided, early in the car project, that a self-driving platform would need serious on-device AI processing. That chip was never finished. The IP was. Apple redirected the work into a general-purpose neural accelerator that first shipped inside the A11 Bionic in the iPhone X in 2017, initially powering FaceID, Animoji, and augmented reality features rather than anything resembling autonomy.
From there the Neural Engine migrated to the desktop as part of the M-series chips, giving Apple a foothold in on-device inference at a time when almost every competitor was routing AI workloads to the cloud. That hardware position is what lets Apple market privacy as a product feature — the less data leaves the device, the less exposure the company has to carry.
Key facts
- 01Apple's M7 Ultra will support up to 1.5TB of RAM and anchor a new server product for the company.
- 02Apple is skipping Pro, Max, and Ultra variants of the M6 to accelerate the M7, due in the first half of 2027.
- 03The Neural Engine, now central to Apple's on-device AI, originated in R&D for the shelved self-driving car program.
- 04The Neural Engine first shipped in 2017 inside the A11 Bionic, powering FaceID, Animoji, and AR on the iPhone X.
The M7 Ultra changes the story again. A 1.5TB memory ceiling is not a laptop specification. It is a server specification, and Gurman reports it will be the basis for a new Apple server product. That puts Apple silicon in direct competition with the data-center parts Apple's own suppliers and rivals are shipping into the AI buildout, and it aligns the hardware roadmap with the company's push to run more of its AI stack on infrastructure it controls.
The strategic tension is well known. Apple's AI hardware is strong. Its AI software has lagged. Apple Intelligence has landed unevenly, and the company recently sued OpenAI alleging that former staff carried hardware trade secrets out the door — a case AI Chat Daily covered earlier this month. The M7 program is Apple's clearest signal that its answer to the software gap is to double down on the silicon layer.
Skipping the M6 Pro, Max, and Ultra tiers is a notable concession. Apple has historically staggered its M-series releases, giving Mac Pro and Mac Studio buyers a high-end refresh in the back half of a chip cycle. Compressing the roadmap means some pro-tier Mac customers will wait longer between meaningful upgrades. It also means Apple believes the M7 Neural Engine gains are large enough to justify the gap.
The bet has risk. Getting to a first-party AI server is not just a silicon problem — it requires networking, cooling, orchestration, and a software stack that hyperscalers have spent a decade refining. Apple has no public track record operating merchant compute at data-center scale, and Gurman's reporting does not detail how the company plans to package, sell, or deploy the new server product. Until Apple ships, the 1.5TB number is a spec on a roadmap, not a shipping SKU.
The shape of Apple's AI strategy is finally legible. The car program that never launched left behind the Neural Engine, the Neural Engine carried Apple through the iPhone AI era, and the M7 Ultra is the piece that tries to extend that advantage from the pocket to the rack. Whether that translates into a competitive AI product line depends on software Apple has yet to demonstrate — but on the silicon side, the company is no longer playing defense.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
