Apple not only discontinued the Mac Pro in March 2026 but also quietly developed and then scrapped an entire generation of “Extreme” processors and a surprise Intel-based successor. Bloomberg’s Mark Gurman revealed the details on July 20, 2026, citing sources inside the company.

Behind the scenes, Apple had created functional M2 Extreme and M3 Extreme chips, planned an Intel Mac Pro with model identifier J170, and even readied an M3 Ultra Mac Pro (J190) for a 2025 launch. None of them reached customers. The company ultimately decided that the economics of niche silicon and the overwhelming success of the Mac Studio made the tower workstation impossible to justify.

The secret projects Apple never shipped

The report lays out four cancelled efforts, all more ambitious than the Mac Pro that actually shipped in 2023 with the standard M2 Ultra.

M2 Extreme and M3 Extreme

Apple’s Ultra chips combine two Max dies using an UltraFusion interconnect. An “Extreme” design would have taken that to the next level—essentially linking the equivalent of four Max dies. On paper, that could have delivered up to 48 CPU cores, 152 GPU cores, and drastically more unified memory bandwidth.

These weren’t just PowerPoint slides. Gurman reports that Apple developed and tested actual M2 Extreme and M3 Extreme silicon. The results were impressive enough to be considered for a flagship Mac Pro, where they would have provided a clear performance advantage over the Mac Studio.

But they were killed for two hard reasons: cost and demand. Each Extreme chip would have required complex, multi-die packaging that was vastly more expensive than an Ultra. With the Mac Pro already selling in tiny volumes, the per-unit development cost became astronomical. Apple would either have to absorb it, price the machine into the stratosphere, or walk away. They walked.

The Intel Mac Pro that almost lived (J170)

Perhaps the most surprising detail is that Apple prototyped a new Intel-based Mac Pro even while aggressively transitioning the rest of the lineup to its own silicon. Codenamed J170, this machine was designed to “address certain uses” where Intel compatibility remained critical.

Professional environments like broadcast, scientific computing, and industrial control often depend on certified software stacks that can’t simply be ported overnight. An Intel Mac Pro could have run legacy x86 apps natively, supported traditional discrete GPUs, and maintained Boot Camp or virtualization workflows that Apple silicon can’t replicate.

Ultimately, Apple decided that maintaining a parallel Intel product line was too costly and confusing. It would have slowed the Arm transition and undermined the message that custom silicon offered better performance per watt. But the fact that J170 existed at all shows Apple understood the pain some users would feel.

The M3 Ultra Mac Pro that never launched (J190)

Alongside the M3 Ultra Mac Studio that did ship in early 2025, Apple had built an M3 Ultra Mac Pro (J190). It would have been the natural successor to the 2023 M2 Ultra Mac Pro, adding Thunderbolt 5, up to a 32-core CPU, an 80-core GPU, and support for as much as 512GB of unified memory.

But the same problem that plagued the 2023 model persisted: the Mac Studio was already offering nearly identical performance in a smaller, cheaper package. The tower’s internal PCIe slots added expansion value, but without discrete GPU support, the central selling proposition had collapsed. Apple killed J190 before announcement.

What this means for you

If you’re a Windows power user or workstation buyer

The Mac Pro’s death doesn’t make towers obsolete—it just means Apple is no longer competing in that category. Windows workstation vendors (Dell, HP, Lenovo, and custom builders) still offer full modularity: replaceable processors, discrete GPUs, ECC memory, multiple NVMe drives, and specialist PCIe cards.

That flexibility is especially valuable for:
- CUDA-based workloads (machine learning, rendering, simulation)
- Studios that need to upgrade GPUs without replacing entire systems
- Scientists and engineers who depend on certified ISV hardware
- Enthusiasts who build and evolve a machine over years

For Windows users, this widens the performance gap in AI and graphics-heavy tasks where Nvidia dominates. Apple’s unified memory can’t match a multi-GPU system running CUDA, and the lack of modularity makes long-term cost of ownership higher for evolving workloads.

If you’re an IT professional supporting mixed environments

Enterprises with fleets of 2019 Intel Mac Pros face a ticking clock. Apple will continue macOS support for Intel for a limited time, but all new development targets Apple silicon. You should begin an immediate audit:
- Inventory any 32-bit apps, kernel extensions, or drivers that require Intel
- Test Rosetta 2 performance for critical line-of-business software
- Evaluate whether Thunderbolt 5 external PCIe enclosures can replace internal cards
- Compare Windows 11 workstations for users who need CUDA, x86 virtualization, or replaceable GPUs

The J170 revelation may sting: Apple had the option to extend Intel compatibility and chose not to. Don’t expect another Intel Mac.

If you’re a general Windows user or gamer

The demise of the high-end Mac tower won’t affect your daily computing directly, but it solidifies a long-term trend. Apple no longer sells a desktop with user-upgradeable RAM, storage, or graphics. The entire lineup—from Mac mini to Mac Studio—is now sealed at the factory.

If you value repairability, upgrade paths, or mixing and matching components, Windows remains your only mainstream option. It’s a reminder that the two ecosystems have diverged irreversibly: Apple offers integrated appliances; Windows offers build-it-yourself flexibility.

How we got here

From Power Mac to dead end

Apple launched the Mac Pro in 2006 as the Intel successor to the Power Mac G5. It was a classic tower: drive bays, PCIe slots, dual Xeons. Professionals loved it.

The 2013 cylindrical redesign was a mistake. It bet on dual GPUs and external Thunderbolt expansion just as GPUs got bigger and hotter. Apple admitted the thermal design painted them into a corner.

In 2019, the tower returned—spacious, modular, and expensive. It restored PCIe slots and Intel Xeons, but the starting price of $5,999 (before a display) limited its audience.

The Apple silicon transition broke the formula

When Apple moved to its own chips in 2020, the Mac Pro’s value proposition fell apart. Apple silicon integrates CPU, GPU, memory, and I/O on one package. That delivers incredible efficiency but eliminates the discrete GPU and RAM upgrades that tower workstations exist to offer.

The 2023 Mac Pro with M2 Ultra had PCIe slots, but they couldn’t fit an aftermarket graphics card. Suddenly, the big box seemed hollow. The Mac Studio with the same chip started at $1,999—thousands less—and did nearly everything the Pro could.

The Extreme gamble that never paid off

Apple’s Extreme chips were the answer: give the tower a unique, massive performance lead. But the economics didn’t work. A chip designed for a $6,000+ machine selling in the tens of thousands can’t amortize the same R&D as an iPhone chip selling in the hundreds of millions.

Add to that the difficulty of scaling multi-die interconnects, thermal density, and packaging yields, and the project became a financial liability. Meanwhile, Mac Studio sales kept growing. The tower had lost its internal champion.

What to do now

For current Mac Pro owners

  • Inventory your PCIe cards: Figure out which ones are essential and whether Thunderbolt alternatives exist.
  • Test App Rosetta translation: Some apps may work fine; others (especially those tied to hardware) won’t.
  • Consider a Mac Studio now: The M3 Ultra model is already workstation-class, and Thunderbolt 5 enclosures can handle many internal-expansion tasks. But accept that you’re buying sealed hardware.
  • If you must switch to Windows: Prioritize software compatibility first. Benchmark your actual workflows, not just CPU scores. Factor in the cost of replacing macOS-specific tools.

For Windows workstation buyers eyeing switchers

  • Highlight modularity in RFPs: Emphasize replaceable GPUs, user-upgradeable RAM, and 5–7-year lifecycle support.
  • Address the reliability gap: Mac users switching to Windows will expect quiet operation, consistent drivers, and minimal bloat. Pre-configured, certified workstations from major OEMs can close that gap.
  • Prepare for Arm-based competition: Windows on Arm is maturing. Qualcomm’s Snapdragon X Elite and future Nvidia/AMD Arm chips could challenge Apple’s efficiency while preserving more hardware choice. Keep an eye on that space, but don’t let it delay immediate decisions.

For the curious enthusiast

Even if you’re not a workstation buyer, the Extreme chip saga is a window into how chip economics shape product lineups. It also reinforces why building a PC makes sense: you’re not betting on one company’s willingness to keep a niche alive. If you’re on Windows, you can keep upgrading your GPU, storage, and memory as your needs evolve, without praying for a company to bless you with a compatible tower.

Outlook: What to watch next

Apple is expected to release a Mac Studio with an M5 Ultra chip as early as fall 2026, according to Gurman. That machine will likely cement the Studio as the professional desktop. The question is whether Apple can push unified-memory capacity high enough and accelerate local AI performance sufficiently to make a tower seem irrelevant.

Thunderbolt 5 will improve external expansion, but it will never match the latency, bandwidth, and simplicity of internal PCIe. As AI workloads demand more memory and compute, the ability to drop in a new Nvidia GPU or add terabytes of RAM may look increasingly attractive.

Windows on Arm will also mature. If Qualcomm or Nvidia can deliver workstation-class Arm chips with strong GPU support and broad ISV certification, the market could see systems that combine Apple-like efficiency with PC-like modularity. That would be the true long-term challenge to Apple’s strategy—and a direct beneficiary of the Mac Pro’s exit.