9to5Mac confirmed the discontinuation and highlighted that the Mac Pro was the last Mac with PCIe expansion slots. The reporting frames this as the end of user-expandable professional computing at Apple, noting the M2 Ultra model had gone nearly three years without an update.
The editorial argues this is not just a product line cleanup but an 'architectural admission' — Apple Silicon's unified memory design has a hard ceiling because you cannot add GPU compute after purchase. The Mac Pro's existence used to paper over this limitation, and its removal makes that ceiling explicit.
The editorial acknowledges that for many workflows — video editing in Final Cut, music production in Logic, and moderate ML inference — Apple's unified memory and custom silicon genuinely delivered workstation-class performance. A Mac Studio with M4 Ultra outperforms the old Intel Mac Pro in most creative benchmarks at a fraction of the power draw.
The editorial notes Apple is pointing pro users to the Mac Studio, which maxes out with the M-series Ultra chip and 192GB unified memory but offers no PCIe slots, no expansion, and no tower form factor. For studios and ML shops that historically slotted in discrete GPUs to scale compute over time, the message is stark: 'this is the ceiling.'
Apple has pulled the Mac Pro from its lineup, officially discontinuing the tower workstation that has anchored the company's professional hardware story for two decades. The move was first spotted on Apple's website and confirmed by 9to5Mac on March 26, 2026. The last Mac Pro — the M2 Ultra model released in June 2023 at a starting price of $6,999 — had already been looking like a dead product walking, receiving no updates while the rest of the Mac line moved to M3 and M4 silicon.
The Mac Pro was the last Mac with PCIe expansion slots, and its death marks the end of user-expandable professional computing at Apple. The machine shipped with seven PCIe slots in its stainless steel tower enclosure, but under Apple Silicon, those slots became a peculiar artifact — you could add storage controllers, networking cards, and certain I/O devices, but not the one thing workstation buyers historically wanted most: discrete GPUs.
Apple is pointing professional users toward the Mac Studio, which maxes out with the M-series Ultra chip, up to 192GB of unified memory, and Thunderbolt connectivity. No expansion. No PCIe. No tower. The message is clear: this is the ceiling.
The Mac Pro's discontinuation isn't just a product line cleanup. It's an architectural admission. When Apple transitioned from Intel to Apple Silicon, the promise was that unified memory and custom silicon would deliver workstation-class performance without the complexity of discrete components. For many workflows — video editing in Final Cut, music production in Logic, even moderate ML inference — that promise delivered. A Mac Studio with an M4 Ultra genuinely outperforms the old Intel Mac Pro in most creative benchmarks at a fraction of the power draw.
But the Apple Silicon architecture has a hard ceiling that the Mac Pro's existence used to paper over: you cannot add GPU compute after purchase. In the Intel era, studios and ML shops could buy a Mac Pro, slot in AMD Radeon Pro W6800X Duo cards, and scale their GPU memory to 96GB or beyond. That expansion path is gone. The M-series Ultra chip's GPU cores are fixed at manufacturing time. What ships is what you get.
This matters acutely for machine learning practitioners. The ML community's brief flirtation with Apple Silicon for training — driven by the unified memory advantage that lets you load larger models than VRAM-limited NVIDIA cards — has always bumped against a throughput problem. A single M2 Ultra's 76-core GPU delivers roughly 27 TFLOPS of FP32 compute. An NVIDIA RTX 4090 delivers 82 TFLOPS. An H100 delivers 756 TFLOPS. The Mac Pro with PCIe slots at least held open the theoretical possibility that Apple might someday support multi-GPU or accelerator cards. That door is now closed.
The Hacker News discussion (211 points and climbing) reflects genuine frustration from developers who had been holding out hope. Several commenters noted they'd been waiting for an M4 Ultra Mac Pro before committing their teams to Linux workstations. Others pointed out the irony: Apple spent $50+ billion on its self-driving car project before canceling it, but couldn't justify updating a $7,000 workstation that professional developers actually wanted to buy.
The Mac Studio is a genuinely excellent machine for 90% of professional workflows — but the 10% it can't serve are disproportionately the highest-value, most compute-intensive workloads in the industry. Video effects houses running Houdini simulations, ML teams doing distributed training, audio professionals with PCIe DSP cards, scientific computing labs — these users don't have a Mac migration path anymore.
If you're a developer or engineering manager making workstation purchasing decisions, the calculus just got simpler in an uncomfortable way.
For ML/AI work: The Mac is now firmly an inference and experimentation platform, not a training platform. If your workflow involves fine-tuning models larger than what fits in 192GB of unified memory, or if you need multi-GPU parallelism, you're looking at Linux workstations with NVIDIA hardware or cloud compute. The MLX framework and Apple's investment in on-device ML are real, but they're optimized for deployment, not for the kind of brute-force training that requires expandable GPU infrastructure.
For general software development: Nothing changes materially. The MacBook Pro and Mac Studio remain the best development machines for most teams. The Mac Pro was already overkill for compilation, container orchestration, and IDE work. If your bottleneck is build times, a Mac Studio with an M4 Ultra will compile faster than the discontinued Mac Pro with an M2 Ultra — the product line's death doesn't reduce your available compute, it just removes the illusion of future expandability.
For creative professionals with PCIe dependencies: This is the hard one. If your workflow depends on PCIe cards — Avid Pro Tools HDX cards for audio, RED Rocket cards for video, specialized I/O cards for broadcast — Apple has effectively told you to stay on your existing hardware, move to Thunderbolt alternatives where they exist, or leave the platform. Some of these Thunderbolt alternatives are mature (UAD audio interfaces, for instance), but others don't exist at all.
For IT departments standardized on Mac: Budget the Mac Studio as your high-end option and adjust expectations accordingly. The good news is you're saving $3,000-$4,000 per seat versus Mac Pro pricing. The bad news is you're losing the ability to service or upgrade machines over a 5-7 year lifecycle — the Mac Studio is a sealed unit with a 3-year practical lifespan before it falls behind.
There's a strategic reading of this move that goes beyond hardware economics. Apple's fastest-growing revenue drivers are Services and iPhone. The Mac business, while healthy at roughly $30 billion annually, is not where Apple's strategic attention lives. The Mac Pro served a market that was small in unit volume but outsized in influence — the professionals whose workflows defined what "pro computing" meant.
By discontinuing the Mac Pro, Apple is conceding that it no longer aspires to serve the most demanding professional computing needs — it's optimizing for the broad middle of the professional market where the Mac Studio and MacBook Pro already dominate. This is a rational business decision. It's also the kind of decision that, over a decade, erodes the "halo" effect that kept Apple relevant in enterprise and developer mindshare.
For now, most developers won't feel this loss directly. The Mac Studio is genuinely excellent. But the signal matters: Apple's professional hardware ambition now has a defined ceiling, and that ceiling is lower than it was yesterday.
The question isn't whether Apple will reverse this decision — they almost certainly won't. The question is whether the Mac Studio line will absorb any of the expandability features that made the Mac Pro relevant. Could a future Mac Studio offer a single PCIe slot? An external GPU enclosure protocol over Thunderbolt 5? A multi-chip-module Ultra variant with more GPU cores? These are the product moves that would soften the blow. But Apple's track record suggests they'll simply push unified memory capacities higher (256GB, 384GB) and call it sufficient. For a lot of us, it will be. For the users who needed more, the answer is now Linux, and Apple seems fine with that.
I think that's an expected thing.G5 was the thing. And companies were buying G5 and other macs like that all the time, because you were able to actually extend it with video cards and some special equipment.But now we have M chips. You don't need video for M chips. You kinda do, but truthf
As someone who came from the SGI O2/Octane era when high-end workstations were compact, distinctive, and sexy, I’ve never really understood the allure of the Mac Pro, with the exception of the 2013 Mac Pro tube, which I owned (small footprint, quiet, and powerful).For me, aesthetics and size ar
Apple really dropped the ball here. They had every ability to make something competitive with Nvidia for AI training as well as inference, by selling high end multi GPU Mac Pro workstations as well as servers, but for some reason chose not to. They had the infrastructure and custom SoCs and everythi
The Ultra variants of the M series chips had previously consisted of two of the Max chips bonded together.The M5 generation Pro and Max chips have moved to a chiplet based architecture, with all the CPU cores on one chiplet, and all the GPU cores on another.https://www.wikipedia.org/w
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
I bet there’s gonna be a banger of a Mac Studio announced in June.Apple really stumbled into making the perfect hardware for home inference machines. Does any hardware company come close to Apple in terms of unified memory and single machines for high throughput inference workloads? Or even any DIY