Apple is Reportedly Facing a 'Massive Dilemma' With the MacBook Neo

[Jeremy3292]

So they don't need to 'run out' of Neo-grade 18 Pros, they can just take otherwise perfect chips and set them to behave at the same level as the 'broken' ones. It's that simple.

I'm not sure you read the article bc they discussed this exact point. The binned chips are free for the Neo production as they would otherwise be scrapped as a byproduct of manufacturing the 6 core chips. If they have to use perfectly good 6 core chips and "bin" them on purpose, well that doesn't cost $0 bc you're now using perfectly good chips that would have been used elsewhere.

So yeah it is "that simple" but it costs Apple profit margin to do so, hence the entire point of the article.

 

[deevey]

These so-called "binned" chips with a 5-core GPU are effectively "free" to Apple, given that they otherwise would have been discarded.

Seeing that these "free" chips are no longer a throwaway product and now have (considerable) value, how likely is it that TSMC would continue to give them away to Apple regardless (even it's A18/19/20).

 

[Zackmd1]

So you think:

1) Apple prefers to sell a $599 device instead of a $999 device?

2) Neo has high demand. Since it's so high, it needs an update to A19 Pro sooner rather than later?

Well considering the Air hasn't sold the numbers that Apple was expecting.... And considering the fact that this whole article is devoted to the issue that A18 pro is OUT of production and that Apple would either have to restart said production to fill demand or switch to another processor, I'm thinking its probably cheaper to accelerate products already in motion and utilize existing stock than to restart production of a older chip. The Air likely has a higher profit margin than the Neo so they could likely afford to absorb a higher production cost due to the accelerated timetable...

 

[I7guy]

Seeing that these "free" chips are no longer a throwaway product and now have (considerable) value, how likely is it that TSMC would continue to give them away to Apple regardless (even it's A18/19/20).

“Free” chips as in “free” beer are not really “free” the way I read it. Apple still pays for the volume and gets the defective ones as well. Now we don’t know if there is an underlying contractual thing about defective chips or not. But the Neo sales just have caught Apple off guard.

 

[cobioguy]

Or add a non binned chip option to upgrade to full cores. Maybe Apple can tempt people to pay a bit more for full chips off setting some demand and cost

 

[CrysisDeu]

I hope Apple can bring a proper MacBook using non binned chip, or an M chip. Then make the AIR a premium option with M chips, but super slim and light.
The current air is serving as the standard non pro macbook, limiting what a true air can be.
The Macbook neo really should be a macbook E below the normal macbook

 

[mindctrl]

Sounds sus. Why would Apple launch a new product at a low price to attract many more buyers and not have a plan for the limited supply they know they have? Doesn’t make sense.

 

[Reason077]

The obvious thing to do? Stoke up rumours that a new Neo with the A19 Pro and 12GB RAM are in the pipeline, thus dampening demand for the current Neo and ensuring that stocks of 8GB A18 Pro SoCs last the rest of the production cycle.

 

[sdf]

Well, now we know why they were able to get the price point so low on these. Here, have a discarded chip in your machine that we paid $0 for. We think you're going to love it!

No, you've misunderstood. Apple paid for those chips, they were just without purpose before.

 

[bodonnell202]

The problem is explaining why Apple waited so long to package and fully test the A18 Pro wafers. You don't know the full count of chips until the wafers have been cut and encapsulated because those steps further reduce yield. Why would Apple wait so long to do that?

The simplest explanation is these chips are brand new chips made for Neo.

So what do you think they did with the binned chips then? Don't try to tell me they don't use them - there's really no other good explanation as to why A and M series chips with disabled GPU and CPU cores show up in a number of Apple products. Also why do you think them not packaging the chips until November 2025 is a gotcha? Why would they package the chips far in advance of using them? Then they just have extra money sunk on RAM waiting a possible more than a year for product release. The Apple supply chain is much more efficient than that.

 

[ajkoeltz]

I don't know that they're all that worried about profit margins on the hardware. I think they're playing the long game, getting more people into Apple subscriptions (ie Creator Studio, AppleOne, Apple TV, Music, News, Apple Care+ etc.) I think this is a non story, just looking for something potentially newsworthy. I haven't read any other comments and I'm sure someone else has already mentioned this. All just my opinion, I could be way off.

 

[teh_hunterer]

When a chip is binned what actually happens is that the chip has failed one or more tests - an automated decision-tree then chooses what happens to that chip, if it's decided to 'kill' a core or feature then that is done, irreversibly, to the chip by disabling fully the offending component. This can be done to perfectly healthy chips, and this route is routinely taken as yields increase over time during the production-run of a particular chip. So they don't need to 'run out' of Neo-grade 18 Pros, they can just take otherwise perfect chips and set them to behave at the same level as the 'broken' ones. It's that simple.

"It's that simple"

No it's not. If you or the 22 people who liked this had actually read the article, you'd know they used binned chips left over from the iPhone 16 Pro production cycle.

If they run out of those chips, they have to start making new a18 Pro chips. Which would be more expensive. Which is the whole point of the article.

 

[Spec360]

Well, now we know why they were able to get the price point so low on these. Here, have a discarded chip in your machine that we paid $0 for. We think you're going to love it!

Apple is all about saving the environment

 

[DerekK]

The Neo and the iPhone have the same size SoC - it is not a redesigned SoC. I cannot explain the package date, but it wouldn't make sense for Apple to have just thrown away chips with only 5 working GPU cores.

No wafers are perfect, so there is always an expectation of a yield less than 100%. Being able to use the 5 GPU core chips just means that there is a higher usable yield from each wafer. As many have pointed out, Apple would pay for the wafer either way, so these chips aren't free.

Apple would have known their product lineup, and likely had a clause with TSMC to produce extra chips.

Given the high success rate of the Neo (maybe even higher than Apple anticipated), it is very likely that TSMC is actively disabling GPU cores from the wafers to give Apple their desired SoC for the Neo. I would not expect Apple to release a Neo plus with extra cores, as this would just cause more confusion and not give much value.

It's possible that Apple has/had other uses for binned SoCs (iPad 12? Apple TV refresh?), but given the success of the Neo, I could see them focusing on more Neo production.

 

[deconstruct60]

"These so-called "binned" chips with a 5-core GPU are effectively "free" to Apple"...you'll have to take this up with the author of this article.

The wafer has a total overall cost whether the dies on the wafer are 'prefect' or have minor defects. If the wafer costs $30K and have 300 perfect chips then the cost per chip is $100. If can use some of the one defect chips, (example 50 more ) then there is a cost of $85.72 per chip. More chips sold for money, the lower the average cost of the chip gets.

The only time the defect chips are 'free' is if saddle the perfect chips with 100% of the overall wafer costs. Can do that but in effect robbing Peter (lower iPhone Pro margins) to pay Paul ( artificially rising the margin of the Neo).

"free" here is a bit of a misnomer. Apple had no other product to put these minor defect chips into. So they had to 'eat' that cost overhead because the A18 Pro was a one product chip that discontinued in 12 months . It doesn't make any sense to create a chip and toss it in the trash can in 12 months. Apple does not do that at all for the regular iPHone. The leading iPhone model would sell for another 2-3 years. The A-series used would be 'handed down" to other products like the entry iPad . etc.

Nobody in the chip business throws away a chip in 12 months. Especially bleeding edge and/or larger ones that are very expensive to develop.

 

[mannyvel]

this is a great problem to have. Does Apple make more money or more money?

 

[deconstruct60]

Seeing that these "free" chips are no longer a throwaway product and now have (considerable) value, how likely is it that TSMC would continue to give them away to Apple regardless (even it's A18/19/20).

TSMC charges per wafer , not per chip. It is only "free" if the iPhone Pro pays for all the chips. That is just a shifting of overhead.

They were in part "throaway" by Apple's choice because they didn't have a product to put them in more so that TSMC doing one thing or the other. Once the iPhone Pro started using a different die than the 'plain' iPhone there was a problematic issue of premature retirement of the that die. Apple is just aligning the A-pro die up with the same policy they use on the rest of their product. Use the die in multiple products spread over time. Don't want only 12 months of production. Want multiple years and spread the development costs over more aggregate units. That lowers the unit development costs for each chip.

[ Where Apple doesn't do that as well. e.g. M Ultra chip you can see products with slower than 12 month upgrades ... because some product needs to extend the timeline of the SoC being sold. ]

 

[dampfnudel]

When a chip is binned what actually happens is that the chip has failed one or more tests - an automated decision-tree then chooses what happens to that chip, if it's decided to 'kill' a core or feature then that is done, irreversibly, to the chip by disabling fully the offending component. This can be done to perfectly healthy chips, and this route is routinely taken as yields increase over time during the production-run of a particular chip. So they don't need to 'run out' of Neo-grade 18 Pros, they can just take otherwise perfect chips and set them to behave at the same level as the 'broken' ones. It's that simple.

Or they could just upgrade the Neo with an extra core?

 

[dampfnudel]

It may push some Neo potential buyers to consider buying the Air, maybe it was all planned from the start. 😉

I’m sure upselling was part of the plan all along. Maybe a small part, but every upsell helps.

 

[deconstruct60]

Or they could just upgrade the Neo with an extra core?

No good reason to do that. Selling binned cored chips means they effectively increase the wafer yield. The whole point here is to sell more dies over more wafers for a longer amount of time. It is also about predictable flow from the fab also.

The A14X sold with a binned down die for it whole lifetime. Apple came out with the A14Z about 1.5 year later. It was the exact same die. The fab was running at higher yields at that point so the average die costs were about the same.

"

A12X: October 30, 2018 A12Z: March 18, 2020

"

https://en.wikipedia.org/wiki/Apple_A12X

The root cause real problem here is that the iPhone Pro product churns its now seperate die prematurely. The plain iPhone keeps getting sold for 2-3 more years. For a while Apple did a hand me down to the iPhone or binned down some GPUs for the plain model (when shared the same die). Apple Silicon default policy is to sell a SoC die for several years at a time.

The Neo trailing a year behind the iPhone Pro will allow the 'Ann Pro' production for that generation to easily go 24 months , rather than just 12. If you make more of a product then can amortize more fixed development costs over more units. $500M over 100M units is $5 per chip. $500M over 10M units is $50. Over 1M units is $500 per chip. Sell more of the units lower the costs.

 

[Alameda]

Seeing that these "free" chips are no longer a throwaway product and now have (considerable) value, how likely is it that TSMC would continue to give them away to Apple regardless (even it's A18/19/20).

The original article is pure speculation. He cites a source who assumes the chips are binned and “free,” which is not true. TSMC makes wafers, which Apple purchases.

After the wafers leave TSMC, there are many steps which turn the wafer into a chip. They have to be tested, packaged, tested some more, baked, and shipped. It’s not likely that Apple left them in a partially assembled state, so if they’re binned from the prior production run, Apple paid for all of this over the course of a year or two, with the intention of using them in the future. They were not free and in fact cost exactly the same as the fully functioning parts.

 

[jz0309]

The alternative theory is those are binned chips cut from wafers back in 2024 and stored as loose dice until November 2025 for the Neo. But every move, inspection, or re‑sort of those loose dice increases damage risk. Unpackaged dice also risk contamination from oxidation. So why the heck would Apple tell TSMC to delay packaging them? All of this also ignores common sense that a $599 MacBook is probably going to have huge demand, so you can't rely on scrap chips for a new product line.

and that doesn't make a while lot of sense either ...
wafers get probed, gross failures are marked, then the wafer gets thinned and diced, put on a tape and stretched so there is some space between dies. Then the wafer on a round tape, with an outside metal ring, is run through the assembly machine and the good die is picked, and then packaged. After it's packaged it's tested, 100% of all tests nd then they can determine if a GPU passes or fails, if 1 GPU fails, the chip gets binned accordingly.

Never ever have I seen assembly houses storing millions of wafer carcasses, that just doesn't make sense. Plus there is a shelf life of the tape being used on those carcasses that is certainly playing a role, you cannot pick a die from a year old tape easily.

That Apple had some "leftover" "binned" chips is certainly plausible, but the qties certainly never were as high as some people here believe.

This whole story is made up, nothing else ...

 

[Detnator]

On one hand, I'm disappointed they're cramming "faulty" chips into these and selling them for hundreds, but on the other I'm happy to see so many people experiencing macOS for the first time. I hope Tim can handle a slight bump to profit margins and get it sorted out.

“Faulty”. What a ridiculous thing to say about the entire chip.

If you’ve got 6 bananas in a bunch, and one of them is bad so you pull that one off and throw it out are the remaking five bad? And if you then advertise and sell it as a bunch of five bananas are you doing something wrong?

As someone else explained the part that is faulty gets completely disabled and what’s left (the 5 cores) is as good as if it was built that way (with 5 cores) in the first place.

If you’re going to participate in the rampant Apple-shaming here, at least do it with something they’ve actually done wrong (and there’s plenty of that, but this isn’t it).

 

[demodave]

Well, now we know why they were able to get the price point so low on these. Here, have a discarded chip in your machine that we paid $0 for. We think you're going to love it!

This is actually good for both companies. Free chips for Apple are an obvious win. For TSMC it means they can maintain a higher internal Quality metric by reducing scrap.

And if new customers buy a Neo at a price that is attractive to them and love it, that is an additional benefit to Apple and also to the new customer.

Seems like a win-win-win-win to me.

Apple is doomed.

 

[subjonas]

Isn't that a good problem to have though? Yes, the margin might drop, but the overall profit would increase, right?

Maybe the margin doesn't scale if the overhead increases more than the sales. Or more likely, it's probably an opportunity cost issue, where more resources spent on a lower profit per unit product means less resources for higher profit products. I'm not a business major, but something like that sounds plausible to me.