Apple's M7 Ultra targets enterprise AI with 1.5TB memory to rival Nvidia's Blackwell accelerators

2028 Could Bring the Most Mind-Bendingly Expensive Apple Product of All Time

It's been previously reported that Apple will be skipping expected variants of its M6 chip and speeding up production of the M7. But the writing on the wall suggests very, very expensive M7 Ultra-powered desktop Macs coming in 2028 -- and the reason may (not) shock you. Anonymous sources are apparently continuing to inform Bloomberg's scoop-getter Mark Gurman about Apple disrupting its traditional chip rollout process. M6-powered Apple products don't even exist yet -- though they are expected later this year -- and that whole generation of chip is already basically obsolete as far as Apple is concerned, according to Gurman. But you're never going to guess what Gurman now claims the reason is. Haha yes you are because it's AI: "The takeaway is that AI is no longer just another feature Apple's chips need to support. It is now shaping how those products are designed and when they are shipped. That's a shift from the days when the main concerns were things like processing speeds, graphics, battery life and thinner designs." That's a little depressing because I happen to like things like processing speeds, graphics, and battery life. Hell, thinness is even pretty cool compared to AI. But anyway, Gurman says we shouldn't expect M6 Pro, Max, or Ultra products. Finalization of the M7 started just six months after the M6 was finalized. Which suggests a weird product timeline: We'll basically say hello and goodbye to the M6 at the same time at the end of this year, and the first M7 products will materialize at the start of next year. Then things will be relatively normal as M7 Pro, and M7 Max arrive at the end of 2027. Then M7 Ultra products will come along in 2028. If you're saving your pennies for a high-end desktop, that M7 Ultra is something to keep in mind. Gurman writes: "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. Widespread memory-chip shortages have made the component harder to find and more expensive." Indeed, it's worth pausing here to really mull over the pricing implications a bit. There are no M5 Ultras yet, but rumors suggest a high-performance desktop release in the form of the Mac Studio. Last month, when Apple famously raised the price of products like the MacBook Neo, which went from $600 to $700, it also raised the price of the base Mac Studio by $500 to $2500. But the price of the higher-end 96GB Mac Studio climbed $1,300 to $5,299. Are you sweating yet? It gets a lot worse. Last time Apple shipped a Mac with 1.5TB of RAM was in 2019. At the time that much RAM cost $25,000 -- just for the RAM. You could conceivably pay $53,000 for your entire computer. And that was seven years ago. Before historic inflation. And an all-out crisis in the price of memory. And remember, the M7 Ultra is reportedly built for AI. Gurman claims that it will approach "the class of dedicated AI accelerators such as Nvidia Corp.'s Blackwell." And how much does it actually cost to buy an Nvidia Blackwell? Currently the cheapest Blackwell I can find on Newegg dot com is priced at $12,499.99. Just for the processor. Obviously you can't buy this imaginary Mac right now, and the consumer market will shift in predictable and unpredictable ways over the next couple of years. But if we imagine it's 2028 and you're thinking about buying a maxed-out M7 Ultra-powered Mac Studio (or whichever model is the top of the line in almost two years), then presumably you just -- and I mean this literally -- took out a second mortgage on your house. This truly could be a computer at a real-estate-level price point.

Apple's M7 Ultra could take on Nvidia Blackwell with a staggering 1.5TB of memory

Bloomberg says Apple's next-generation AI chip is being built for far more than just future Macs. Apple's next flagship chip may not just be another performance upgrade. Instead, it could be the company's biggest AI leap yet. According to Bloomberg's Mark Gurman, Apple is developing the M7 Ultra with one clear goal: dramatically boosting AI performance. Expected to arrive in 2028, the processor is reportedly being designed to handle workloads on a scale that brings it closer to dedicated AI accelerators like Nvidia's Blackwell than traditional desktop processors. A desktop chip with server-class memory The biggest headline is memory. Bloomberg reports that the M7 Ultra is being designed to support up to 1.5TB of unified memory, which is roughly double the capacity currently planned for Apple's upcoming M5 Ultra. That's an eye-watering amount even by workstation standards, and far beyond what today's consumer Macs offer. The idea is simple: larger AI models demand larger memory pools, and Apple wants its future silicon to handle them without constantly relying on external storage or cloud processing. There is one catch, though. Bloomberg notes that whether Apple can actually ship Macs with the full 1.5TB configuration will depend on the state of the memory market. Ongoing memory-chip shortages continue to make high-capacity modules difficult to source and significantly more expensive, meaning the maximum configuration may ultimately depend as much on supply chains as on engineering. Apple Intelligence isn't the only target The M7 Ultra isn't just being built for future Macs, either. According to Bloomberg, Apple also plans to use the chip as the backbone of its next-generation AI servers. While an M5 Ultra-based server platform is expected to arrive first, engineers are already developing a more powerful M7 Ultra-powered architecture targeted for deployment around 2029, helping power Apple Intelligence both on-device and in the cloud. More importantly, the M7 Ultra highlights a major shift in Apple's silicon strategy. Instead of focusing solely on faster CPUs, better graphics, or improved battery life, Bloomberg says AI is now dictating how the company designs its chips. That's reportedly why Apple accelerated the M7 family, with the Ultra model expected to deliver AI performance much closer to enterprise accelerators like Nvidia's Blackwell. In other words, Apple is no longer building chips that happen to support AI -- it's building chips around AI. If Bloomberg's roadmap proves accurate, the M7 Ultra won't just be another annual silicon refresh. It could mark the point at which Apple's AI ambitions begin to compete with those of some of the biggest names in enterprise computing.

M7 Ultra To Offer Double The Maximum Unified Memory As The M5 Ultra, Potentially Opening Doors To Run 1 Trillion-Parameter Models Locally

* 0-20%: Unlikely - Lacks credible sources * 21-40%: Questionable - Some concerns remain * 41-60%: Plausible - Reasonable evidence * 61-80%: Probable - Strong evidence * 81-100%: Highly Likely - Multiple reliable sources Apple is planning to skip the M6 Ultra launch, instead choosing to launch the M5 Ultra later this year, followed by the M7 Ultra after it unveils the M7 Pro and M7 Max sometime in 2027. Fortunately, the new workstation-class silicon will be worth the wait, especially for those who aspire to run multi-billion-parameter AI models on their systems but are limited by memory constraints. This is because a new report states that the M7 Ultra will feature double the unified memory of the M5 Ultra, and you'll want to know how much RAM you can pack for the top-end configuration. With the M5 Ultra reportedly limited to 768GB of unified memory, the M7 Ultra could overshadow it with a 1.5TB memory option The ability to run advanced AI models depends on CPU speeds, with the heavy lifting being undertaken by the memory count and bandwidth. Fortunately, the M7 Ultra will address all three limitations as it's expected to be the fastest Apple Silicon when it launches. While no CPU or GPU core counts are mentioned, Bloomberg's Mark Gurman previously stated that the M5 Ultra will offer up to a 36-core CPU and an 80-core GPU, paired with 768GB of unified memory. The M7 Ultra, on the other hand, will offer up to a mammoth 1.5TB of unified memory. Also, for those worried about running trillion-parameter models, we'll remind you that the M3 Ultra featured 819GB/s of memory bandwidth, so we shouldn't be surprised if the M7 Ultra can cross over to the 1TB/s bandwidth territory. "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. Widespread memory-chip shortages have made the component harder to find and more expensive." The only obstacle that remains is how much the Mac Studio equipped with the M7 Ultra will cost. Given that the California-based titan is testing CXMT DRAM, the company should minimize its supply problem, but consumers should expect to pay around $20,000 for the maxed-out version. Just compare the top-end configuration of the Mac Studio, which is kitted with an M3 Ultra coupled with 96GB of unified memory. The cheapest option retails for $5,299, assuming you're willing to select fewer CPU and GPU cores, not to mention outfit the machine with just a 1TB SSD. For those believing that the 1.5TB RAM on the M7 Ultra can't handle one trillion-parameter models, the M3 Ultra with its 512GB RAM can run DeepSeek's R1 AI model with 671 billion parameters. It should be possible for the M7 Ultra to run 1.2 trillion-parameter models with 8-bit quantization, so let us keep our fingers crossed for these future benchmarks. News Source: Bloomberg Follow Wccftech on Google to get more of our news coverage in your feeds.