1.5 TB of unified memory would push Apple’s rumored M7 Ultra far beyond today’s Apple Silicon ceiling and into territory normally associated with high-end workstations and entry-level servers.
That is the headline signal from a new report on Apple’s chip roadmap. The M7 Ultra is tipped to support up to 1.5 TB of unified memory, according to Bloomberg’s Mark Gurman, as reported by Notebookcheck. If accurate, Apple is not just preparing a faster Mac chip. It is designing a much larger shared-memory platform for professional workloads that need capacity as much as compute.
The caveat is important: support does not guarantee shipment. Apple previously discontinued the 512 GB configuration of the M3 Ultra Mac Studio even though the processor supported it. So the M7 Ultra rumor is best read as a design target, not a confirmed Mac configuration.
“The new Ultra is designed to support as much as 1.5 terabytes of memory — roughly double the capacity planned for the M5 Ultra — though whether Apple ultimately offers that configuration will depend on the state of the industry,” Gurman wrote, according to 9to5Mac.
Apple’s M7 Ultra memory target points beyond the desktop Mac
The reported 1.5 TB ceiling would be three times the 512 GB maximum supported by the M3 Ultra, currently the highest memory capacity offered by any Apple Silicon chip in the supplied source material. Gurman also claims the eventual M5 Ultra will support about 768 GB, roughly half the M7 Ultra figure.
That makes the roadmap jump unusually sharp. Apple is reportedly skipping M6 Pro, M6 Max, and M6 Ultra, while the base M6 is still expected later this year alongside a new MacBook variant. The next high-performance Apple chips are instead said to arrive under the M7 family.
MLXIO analysis: that shift suggests Apple may be treating the Ultra tier less like a normal annual Mac upgrade and more like a larger architectural reset. The memory ceiling is the tell. CPU and GPU gains matter, but a move from hundreds of gigabytes to 1.5 TB changes the class of workloads Apple can plausibly target.
Unified memory is central here. Apple’s model gives the CPU, GPU, and other accelerators access to one shared pool rather than splitting system memory from discrete GPU memory. That can reduce duplication and make large workloads easier to keep resident, provided the software is built to take advantage of the architecture.
The clearest number is also the biggest uncertainty
The rumored memory ladder looks like this:
| Apple Silicon chip | Reported / known unified memory ceiling | Status in supplied sources |
|---|---|---|
| M3 Ultra | 512 GB | Highest Apple Silicon capacity to date; 512 GB Mac Studio configuration was later discontinued |
| M5 Ultra | About 768 GB | Claimed by Gurman for eventual launch |
| M7 Ultra | Up to 1.5 TB | Reported design support; actual configuration unclear |
The difference between “supports” and “sells” matters. Apple’s discontinued 512 GB M3 Ultra configuration shows that technical capability can be overruled by product, supply, or pricing decisions. The additional context from 9to5Mac notes Gurman’s warning that memory-chip shortages could affect whether Apple offers the full 1.5 TB option.
That supply angle should not be treated as a footnote. A maximum-memory M7 Ultra Mac would require Apple to commit scarce, expensive memory to a niche configuration. For separate MLXIO coverage of Apple’s memory-cost pressure, see Apple Bets on Blacklisted CXMT as Memory Costs Bite.
Pricing is also unresolved. 9to5Mac estimates that, based on Apple’s current RAM pricing of roughly $25 per additional gigabyte, moving from 128 GB to 1.5 TB could cost over $35,000. That is not a confirmed M7 Ultra price. It does show why a maximum-memory configuration, if sold, would likely sit far outside mainstream Mac buying decisions. MLXIO has also covered the broader upgrade-cost issue in New $10,149 MacBook Pro Reveals Apple’s Upgrade Trap.
The Mac Studio becomes the most likely proving ground
Notebookcheck reports that Gurman previously said the next Mac Studio with M7 Ultra is expected in early 2028 with a better cooling system. That pairing is logical if the chip’s memory and compute envelope grows materially.
A larger unified-memory pool alone does not make a workstation. Bandwidth, latency, CPU resources, GPU resources, thermals, and software behavior still decide real performance. The source specifically notes that if Apple also expands the CPU and GPU resources available on the M7 Ultra, the chip could compete more directly in the high-end workstation segment.
The named comparisons are telling. Notebookcheck says the M7 Ultra could line up against AMD’s upcoming Medusa Halo and Intel’s Nova Lake-AX processors. At the upper end, it may also chase workloads traditionally handled by entry-level AMD Epyc, Intel Xeon, and Qualcomm Dragonfly C1000 processors, especially as AI and memory-intensive applications become more common.
MLXIO analysis: Apple’s opportunity is not to copy the traditional workstation formula. It is to sell a tightly integrated box with unusually large shared memory. The risk is that fixed memory cuts both ways. Buyers must choose the right capacity at purchase, with no post-purchase upgrade path implied by Apple’s unified-memory design.
AI is the cleanest explanation for a 1.5 TB ceiling
The source material does not prove Apple’s exact workload target. But it does explicitly connect the M7 Ultra’s potential to AI and memory-intensive applications. That is enough to frame the strategic logic.
AI workloads can be constrained by memory capacity before raw compute becomes the only issue. A larger shared pool gives Apple more room to position future Ultra-class Macs and possibly server-adjacent systems for models and datasets that do not fit comfortably inside smaller configurations.
Technobezz’s related summary says Apple sees the M7 Ultra as part of its AI server strategy and that engineers are reportedly working on an M7 Ultra-based server product that could arrive by 2029. Treat that as a separate report, not a confirmed Apple plan. Still, it aligns with the memory number: 1.5 TB is excessive for ordinary desktop use, but not excessive for machines expected to chew through heavy local workloads.
The software caveat remains the biggest practical issue. A massive unified-memory pool only matters if developers can use it efficiently. Apple needs applications and frameworks that treat the M7 Ultra as more than a bigger Mac Studio spec sheet.
The manufacturing split hints at a complex M7 generation
The M7 roadmap also appears technically fragmented. Notebookcheck says the base M7 is rumored to be manufactured on Intel’s 18A-P process, while the flagship M7 Ultra is expected to stay with TSMC.
Given the expected 2028 debut, Notebookcheck says the M7 Ultra will likely use a member of TSMC’s N2 family, though the exact process node remains unknown. That uncertainty matters because process choice affects power, density, cost, and cooling requirements. It also reinforces why this report should be treated as a roadmap signal rather than a final product sheet.
MLXIO analysis: if Apple is willing to split manufacturing across the M7 family, it may be optimizing different chips for different constraints. The base chip can serve thinner, broader Mac products. The Ultra chip can prioritize memory scale, thermal headroom, and professional throughput.
The real test is whether Apple ships the full configuration
The M7 Ultra rumor changes the conversation because it puts Apple Silicon back near a symbolic high-water mark: 1.5 TB, the maximum RAM configuration of the 2019 Intel Mac Pro, according to 9to5Mac. For years, Apple Silicon’s integrated memory design delivered speed and efficiency, but constrained maximum capacity versus traditional expandable systems.
If Apple ships the full 1.5 TB option, the Mac Studio or a Mac Pro-class machine becomes a stronger candidate for buyers who want a compact workstation-style system with very large shared memory. If Apple only supports the capacity on paper, the story weakens.
The evidence to watch is specific: whether the M5 Ultra actually arrives with about 768 GB, whether Apple restores high-memory Mac Studio configurations after discontinuing the 512 GB M3 Ultra option, and whether the early 2028 M7 Ultra Mac Studio launches with the rumored cooling upgrade.
Until then, the safest read is this: Apple appears to be engineering the M7 Ultra for a much more memory-heavy future. The open question is whether supply, pricing, and product strategy let that future reach buyers.
Impact Analysis
- A 1.5 TB unified memory ceiling would move Apple Silicon closer to workstation and entry-level server territory.
- The jump from 512 GB on M3 Ultra to a rumored 1.5 TB on M7 Ultra signals a major shift toward capacity-heavy pro workloads.
- Apple may support the memory technically without actually selling that configuration, so availability remains uncertain.
