They're not ditching anything. They're switching RAM modules from M3U to M5U. Mark my words, there will be a 512 GB option if not larger options.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Says Mac Studio and Mac Mini Will Be in Short Supply for Months
- Thread starter MacRumors
- Start date
- Sort by reaction score
Not sure about this; WMCM isn’t optional
The information that I am following is this;
Apple's A20 chip for the base iPhone 18 will continue using TSMC's InFO packaging instead of the advanced WMCM technology due to ongoing DRAM shortages. This decision reflects supply chain challenges impacting hardware innovation and limits the chip's architectural flexibility for now.
This shift underscores how supply chain constraints stemming from AI companies' supply squeezes are forcing even industry leaders like Apple to scale back ambitious hardware plans. Previously, the A20 was expected to transition from TSMC's InFO (Integrated Fan-Out) packaging to WMCM, a modular approach enabling flexible CPU/GPU core configurations.
The upcoming iPhone 18 Pro may well be a WMCM, but doubts remain for the base iPhone 18
Yeah, I did as well.
Interested in upgrading to an M5 Max Studio, or might splurge on an M5 Ultra as my last work-related desktop before retirement, but I shudder at the cost.
artifex
macrumors 6502a
Don't mistake what "influencers" and other status-seekers do as normal.
tenthousandthings
Contributor
Sorry, I guess I wasn’t being clear — WMCM isn’t available for 3nm and earlier nodes, and InFO-PoP is not available for 2nm nodes. If the A20 exists and it does not use WMCM, then it will be because it is on a 3nm node (like N3P). It is not possible for it to be on a 2nm node (like N2) and not use WMCM.
The source you are quoting seems to believe that an A20 on N2 with InFO-PoP packaging is possible. It is not.
Most likely the iPhone 18 will use the A19 — Apple did this with the iPhone 14 and iPhone 15, both of which used the A16.
Just as the A17 Pro didn’t have a corresponding A17, the A20 Pro may not have a corresponding A20.
There is from the M3 Ultra to the M5 Ultra.
Not true. See next comment.
RDMA support which is designed for AI has clearly been in progress for (years?).
I kept reading headlines that said the price increased. Later they clarified that the base model disappeared and the 512 GB model was still the same price. That's a little disingenuous.
During today's earnings call for the second fiscal quarter of 2026, Apple CEO Tim Cook said that the Mac mini and Mac Studio could be hard to get for months to come.
"We think, looking forward, that the Mac mini and Mac Studio may take several months to reach supply demand balance," Cook said.
Apple underestimated demand for the Mac mini and the Mac Studio. "Both of these are amazing platforms for AI and agentic tools and the customer recognition of that is happening faster than what we had predicted, and so we saw higher than expected demand," Cook said.
Shipping delays for the Mac mini and the Mac Studio have been increasing over the last few months, and the waits for some models stretch into months. Apple stopped selling the Mac Studio with 512GB RAM entirely, and it stopped accepting orders for some models with higher amounts of RAM. As of last week, the base Mac mini was listed as "Currently Unavailable" from Apple's online store because it is out of stock.
Article Link: Apple Says Mac Studio and Mac Mini Will Be in Short Supply for Months
Thanks for your reply... I guess we will find out what the strategy is this fall/winter. But recent reports still indicate a possibility that Apple might skip or delay WMCM for the A20 due to global DRAM shortages. If these shortages persist, Apple may fall back on more established packaging methods while still utilizing the 2nm node for the silicon itself.
I know the above statement contradicts your position.
apple4everOG
macrumors member
No it wasn't. Without the AI crazy it was highly likely a spring or summer release.
apple4everOG
macrumors member
apple4everOG
macrumors member
For what tasks? I am not talking theoretical improvements quoted by apple but results for real workloads.
apple4everOG
macrumors member
dredlew
macrumors 6502
I doubt it. The M5 Ultra chip has likely already begun production and that chip can only be used in a Mac Studio. Hence, we’ll end up in the same stupid scenario again as when the M3 Ultra launched while the M4 Max outperformed it in various tasks at a much lower price. - Not only is it a bad value proposition for the consumer, it’s also a terrible sales strategy.
Apple really needs to fix the rollout issue and release all new processor models at or around the same timeframe again.
Only soldering memory causes costs for apple. A ram stick would have made it possible to get out and the customers could upgrade memory in the future. This ram shortage might be over in autumn or be a lot bigger problem.
Any examples of performance e figures?
Sadly it won't. Every ai chip manufacturers by nvidia, Google, Amazon or nvidia needs 144-288GB of ram.
More ram + chips + electricity = more ai compute to serve customers = more revenue
Basically there is a direct link between ram and revenue. That makes the need existential to anthropic, openai and Google. They will buy and forward order as much as they can even above market value.
Given the work they’ve put in to things like RDMA I dont think they were ignorant of the potential for AI use, the unified memory arch is perfect for it, Apple’s engineers arent stupid. What I do think happened is they didnt secure enough inventory of parts (or couldnt) to handle what became an utterly massive surge. Remember, Apple (and other large companies) plan their production lines a couple years in advance of a release. And there’s a shortage on both options for production for their CPU/GPU/SoC in general and for memory everywhere right now
And look, we know the iPhone has both higher margins and waaay more sales than the mac mini, and if you have a limited number of parts and you have a choice of course you’re going to prioritize the bigger and more profitable product line to keep supplied
Well said. What's peculiar about AI is that it sucks and is useless for certain tasks, then 6 months later it's magically exceptional at those tasks.
12 months ago, local models were gimmicky for Mac mini. The reasoning wasn't good enough and token per second speed wasn't fast enough. However in the past 6 months all of that has changed. Local models are good and a well equipped Mac mini can hit 60-80 tokens per second, Openclaw has only recently become good. Furthermore, frontier models are now increasingly more expensive and rate limited which has spiked interested in local inference on Mac mini.
Apple couldn't have forecasted all of these catalysts, and as a result didn't buy sufficient amount of RAM to growing demand for Mac mini with 64GB+ RAM which is not the norm for consumer grade devices.
It seems incredibly doubtful that issue is a surge in people buying mac minis for LLMs
That feels more like urban legend then anything else at this point
It’s more likely a shortage of ram and storage due to ai demand in general combined with supply chain constraints in both manufacturering and shipping due to energy shortages brought about the current wars in the Middle East blocking the shipment of energy to Europe and Asia
Not sure if you’ve noticed but it’s not just Mac minis that are in short supply
Last edited:
The M3 to M5 difference is correct. However, the all-new Mac Mini features the game changing M4. In this case, the M5 does not take things to a whole new level. Therefore, the Mac Mini cannot receive the M5.
Not true. See comment above.
It's not.
This what Tim Cook said in the earning call last week.
"on the Mac mini and the Mac Studio, both of these are amazing platforms for AI and agentic tools, and the customer recognition of that is happening faster than what we had predicted, and so we saw higher than expected demand."
Apple is rationing RAM.
Manufacturers like Apple order a fixed allocation of RAM chips from Micron and Samsung. Let say one year most orders are 8-32GB. Then the next year there a surge in orders for 128GB, 256GB and 512GB orders. This drastically affects the number of Mac Minis you can produce.
It's good that the 512GB option has been discontinued because one of those alone, has as much RAM as 16 Mac minis with 32GB RAM. It's more beneficial for Apple to sell more units of Mac mini than disproportionately selling more RAM. Demand from Micron and Samsung is so high that Apple can't just ask for as much more ram as it can afford.
It doesn't matter if not everyone is buying a Mac mini for AI use. A sharp and unexpected raise in big ram configurations across Mac mini and Mac Studio is enough disrupt plans for the year. I wouldn't be surprised is they are rationing and holding back RAM for the M5. You don't want to deplete your RAM application in the first half of 2026 and run into problems later in the year.
it is
Were it not for all of those other factors, and they sold a few more minis than expected, they could just buy more ram and keep selling at the same price
Instead ram (and other components) are harder to get and MUCH more expensive
Apple doesn’t change same generation machine prices. When the fall spec bump to m5 minis and studios come expect to pay alot more
If the Mac mini and Studio didn't have:
-Unified ram for the GPU to maximise and
-Very low power consumption
-AI cores that makes it the most efficient consumer device for AI inference
it would have not have been selling out.
Vincent Hanna
macrumors 6502
WWDC is a month away. If no M5s are released, I very much doubt you’ll still be able to get an M4 in 12-14 days.
you don't know that it has anything to do with local LLM use. it's speculation built on a narrative you happen to believe