TSMC “Apple-first” 3nm policy leads to AMD and Qualcomm mutiny
club386.com
club386.com
An early example.
https://www.apple.com/newsroom/2005/11/21Apple-Announces-Lon...
A more recent example.
https://www.reuters.com/article/us-apple-corning/apple-award...
If you go back to the time when Apple was looking at single sourcing all their SOC production at TSMC, you'll see TSMC's CEO publicly saying it would make sense to dedicate a Fab to a single customer.
>The world's leading foundry chip maker Taiwan Semiconductor Manufacturing Co. Ltd. is considering operating single-customer wafer fabs, according to chairman and CEO Morris Chang.
"I think that they are going to be larger customers, and now it makes complete sense to dedicate a whole fab to just one customer and hold that – to hold fabs in fact to just one customer."
https://web.archive.org/web/20120728040723/https://www.eetim...
I think the reason that Apple is always first in line at TSMC is that they bought TSMC a Fab.
Actually I would put the blame squarely on AMD and Qualcomm for making themselves dependent on TSMC - especially AMD who have turned themselves into a "fabless manufacturer" and are now experiencing the consequences...
I'd argue that they need to pay significantly more than other customer in order to get such preferential treatment.
In a short-sighted view of things, you only need to pay slightly more. If AMD is willing to pay $1 and Apple is willing to pay $1.01, you make more profit selling to Apple. However, as this article shows, you might end up losing the other business if you offer Apple such preferential treatment. If Apple isn't going to use all of your capacity year-round, you don't want to alienate the other companies whose orders you rely on.
I think Apple likely has to pay a significant premium for getting access to the latest and greatest to the detriment of competitors. It's not in TSMC's interests to become too dependent on Apple. If AMD and Qualcomm mutiny and their orders start boosting Samsung's foundries with more money for R&D, TSMC could find itself 1) competing against a better-funded Samsung foundry; 2) with one customer that now has leverage over TSMC and getting paid less.
If AMD and Qualcomm move all their orders to Samsung, it provides Samsung with the money to reinvest in its chip business. If they're able to make long-term commitments to Samsung, that's bad for TSMC since it will allow Samsung to invest knowing it will make a profit (just as TSMC has been able to do that with Apple's commitments).
Likewise, if the two other giant chip design companies move to Samsung exclusively, that leaves TSMC in a tough negotiating position with Apple. Before, Apple would have to compete against AMD and Qualcomm for capacity. Now if AMD and Qualcomm have made long-term commitments to Samsung, TSMC becomes really reliant on Apple and Apple will know that TSMC has capacity they can't sell elsewhere. Sure, MediaTek and others exist, but it swings the power away from TSMC and toward Apple. Let's say that Apple was using 40% of TSMC's capacity, AMD 25%, Qualcomm 25%, and MediaTek 10%. Now AMD and Qualcomm make long-term commitments to Samsung. Apple knows that TSMC's orders have dropped 50% and that gives Apple a lot of power.
Giving Apple the best to the detriment of AMD, Qualcomm, and others is a risky play for TSMC. They'll definitely want to be getting very well compensated for it, not merely slightly more. They'll want to make sure that what Apple is offering is enough to offset the substantial risk of angering competing chip design companies who might look for fabs elsewhere.
It's hard to predict the future; and the whole point of a foundry and why TSMC has been so successful is because they refused to play that game. They kept their options open and flexible rather than betting on a few key chips, and they've repeatedly pointed out how the very diversity of their customers has meant they always have some customer willing to be a guinea pig for some new tech.
Losing diversity is a cost; it undermines their very business model. From a business perspective you'd hope they really made apple pay through the nose for that cost, and not just acquiesce because they're a significant customer.
I'd say it's at least plausible that TSMC miscalculated here, and didn't think AMD and Qualcomm (and perhaps others) would dare.
To be clear: they might still get lucky (and AMD/Qualcomm unlucky) - it's not exactly news that new process nodes sometimes fail or are delayed, and if Samsung turns out not to be able to deliver, they'll be able to reacquire that diversity with little trouble, surely. And if they can replace AMD/Qualcomm with other leading edge customers they might not even really be losing much of that diversity they value so highly.
https://en.wikipedia.org/wiki/Advanced_Micro_Devices#Litigat...
"n 2005, following an investigation, the Japan Federal Trade Commission found Intel guilty of a number of violations. On June 27, 2005, AMD won an antitrust suit against Intel in Japan, and on the same day, AMD filed a broad antitrust complaint against Intel in the U.S. Federal District Court in Delaware. The complaint alleges systematic use of secret rebates, special discounts, threats, and other means used by Intel to lock AMD processors out of the global market. Since the start of this action, the court has issued subpoenas to major computer manufacturers including Acer, Dell, Lenovo, HP and Toshiba.
In November 2009, Intel agreed to pay AMD $1.25bn and renew a five-year patent cross-licensing agreement as part of a deal to settle all outstanding legal disputes between them."
Unfortunately by that point the damage had been done. Fabs are an extremely expensive, normally lower margin, business. Risk averse companies probably can't get into that game. Even with Apple they appear to have focused on something of a partnership with TSMC rather than their own fab.
But at the same time, they were then unshackled as you said; As time went on and they could move more volume to TSMC it wound up helping them out immensely.
It's kinda worth remembering too though, that even at their 'peak' in the early 2000s AMD was 6-12 months behind Intel on process tech if you go by releases. Given the trouble Intel has had keeping up one could only imagine where AMD would be now.
They don't have to, but buying production equipment for their manufacturing partners (in return for guaranteed pricing and production levels) is the norm at Tim Cook's Apple.
Apple dont buy production equipment for TSMC, or Samsung Foundry. Zero.
To add to the other examples given here, Apple actually owned the manufacturing equipment for the first generation retina panels, even though it was housed in factories owned and operated by Sharp.
Undoubtably that's why Apple gets such preferential treatment from TSMC and other suppliers.
The consequences being if you don't like what your fab is up to, you can switch fabs. Doesn't sound too bad. GlobalFoundries, the former AMD fab, has all but given up on smaller nodes at this point, but AMD switched to TSMC for CPUs and a mix of TSMC and Samsung for GPUs. On the other hand, Intel had problems with their fab for several years, and is only now starting to consider using other fabs, when their process seems to be starting to work.
Yes, designing chips to fabricate on different lines is more work, but it's something AMD has intentionally done and it has benefits over running your own fab, especially when your own fab has trouble with node shrinks.
It's not as simple as high margins but controlling most of the stack. If all Apple was doing was charging more, we'd see comparable quality devices at lower cost but we don't and it's not uncommon for e.g. a Samsung flagship to cost more despite being slower (or narrowly beating the previous generation, depending on launch timing). What we're seeing instead is that Apple has both consistent goals on their execution (e.g. not Google and Qualcomm working at cross-purposes which make sense for their individual businesses) and captures more of the revenue.
That's a lot harder to catch up with and gives Apple consistent enough revenue that they can do things like effectively floating TSMC a great loan or pay more per unit since they don't need to justify it based on the profits from a single layer. Dell can't pay for their processor R&D using the money you spend on Spotify or Steam.
AMD is not the savior here, nor any other TSMC customer. Intel completely squandered its lead, and now we really need to hope they can catch up and remain competitive. We shouldn't allow TSMC to take total control of the market.
It's like a restaurant reservation for a very regular "VIP" customer. They may get really special treatment, because they're probably spent enough to earn it. And everyone else schleps in line.
Really there's nothing wrong with that. It may be annoying, but someone is paying way more and consistently for that table.
I know the few places I frequent often, I get a table mostly right away if anything is available, regular staff knows what I want, more or less, and I get what I expect, mostly because I have a relationship with the vendor.
I tip well. Chat everyone up, etc.
You don't think Apple or TSMC does this?
I've been in that type of relationship in that industry. That's really how it kind of works. It's not even the most $$$$, but relationship and longer term leverage.
They don’t even have to do that. They just have to place orders that are sufficiently larger than any of TSMC’s other customers’.
Also large minimum orders, and probably a bunch of the payment upfront or somesuch.
And uh yeah: AMD and Qualcomm do have the order sizes to do that.
On MacBook laptops from ~2011, a green light would indicate when the webcam was in use. But the magical part of this was that when it was off, you couldn't see the LED at all, so it looked like this light was shining through the metal bezel at the top. This was achieved by drilling extremely small holes in the frame above the LED, small enough that they're barely visible without magnification but having enough of them that light can easily pass through. It's a striking effect, if you notice it.
To do this, they apparently bought so many high-precision lasers just for this feature that it created a shortage for other companies that needed them: https://www.bloomberg.com/news/articles/2011-11-03/apples-su...
(Presumably this would help by giving them access to the IP so they could contract other factories to manufacture them as well, since Apple has the up-front cash to do it)
Again, I can't find the article so who knows if it's true, but it sounds very, very Apple to do that.
http://www.asymco.com/2011/10/16/how-much-do-apples-factorie...
https://www.engadget.com/2019-05-06-apple-sapphire-glass-sup...
https://www.youtube.com/watch?v=Tj5NNxVwNwQ Arthur W. Astrin, sort of father of WiFi, hired by Apple to develop and incorporate 802.11b into their products. Spend most of 1998/99 flying West Coast-Taiwan twice a week (coordinating iBooks/iMacs manufacturing).
Made HP very happy:
>so I called up HP and I said "I need a spectrum analyzer. It needs to be at least three gigahertz,", and it would be nice if it was programmable. So, they whipped up something and next day the salesman shows up with this.. they call it HP Basic Language Computer. So we built 1, 2, then we ordered 10 then 50. All of a sudden HP started coming. The salesman became my best friend. <Laughs> In fact, I think I was the largest order that year for the spectrum analyzer at HP
We are talking small car unit prices here.
Afaik they did the same for iPhone camera modules with Primax, who went from small time accessories and gadgets manufacturer to one of Taiwan titans.
Apple used to buy up the flash memory in advance. Leaving competitors to pay more for smaller yield or inferior quality.
It doesn't always work. Apple once invested a huge amount in making sapphire glass for their watches. I can't remember the company name but they failed completely at delivering to Apples required spec and the deal was scrapped (leaving the supplier to face massive debts).
Apple says something along the lines of “we guarantee to buy x million chips this year, y million next year. We know it’s expensive to build that capacity, so here’s a few billion up front”.
That’s an offer few can make and nobody else is willing to make.
Maybe a more fair version would be for the federal government to socialize each of these players, and decide that they will pay a premium for slow, hot, American-made parts.
If that sounds ridiculous, meditate on why you think capitalism/markets have a negative connotation when they are in fact producing great outcomes - amazing chips (M1 series) and competitive pressure on Intel, AMD, and Qualcomm to eventually deliver similarly competitive chips.
Unbridled capitalism gets a bad rap, but that’s a relatively new innovation. You can have a well-regulated market produce even more fabulous outcomes; don’t blame corporations for those outcomes, blame politicians and the voting public.
Edit: a theme I see repeated over and over, on hacker news of all places, is people making excuses for businesses operating in competitive markets. Making excuses is bad for consumers, it’s bad for competition, it’s bad for society. Stop making excuses and ask business to work harder. Android users deserve fast processors, and it is 100% Qualcomm’s fault that they don’t have them. Intel PC users deserve fast, low-power processors, and it is 100% Intel’s fault that users don’t get them, or have to go to AMD/Apple for them. Apple’s success is proof positive that AMD, Qualcomm, and Intel have been insufficiently vigorous and innovative in their competition, which in manufacturing products also involves the surrounding business practices needed to ensure access to fab capacity.
>> You can have a well-regulated market produce even more fabulous outcomes...
Can you cite an example of this?
At a much smaller scale, I have had contracts with various vendors to provide “stuff” associated with services they provide to us.
In one case, they had a facility dedicated to my company’s needs, with tooling maintained by my company. The contractor has an SLA to achieve different levels of operational readiness.
What Apple is doing is no different, except they are buying 8-9 figure tools or loaning lots of cash at favorable terms.
I certainly wouldn’t shed many years for Qualcomm, which has an actual monopoly on modems. Their lack of strategic competence isn’t Apple’s sin.
I'm not sure that's true these days, but, even if it is true, plenty of other companies could _raise_ it easily enough. It might make less sense for other companies, though. Like, realistically, if AMD is beating Intel already (and, for the moment, they largely are; Alder Lake is niche for now), what does being a little earlier with 3nm buy them? Is it worth the money? Perhaps not.
It's even less clear that it would be worth it for Qualcomm. For them, it's hard to see that there'd be any return; they have a captive audience already.
How many companies can easily raise $200 billion?
https://www.cnbc.com/2021/01/27/apple-q1-cash-hoard-heres-ho...
This extraordinary claim that it's easy to raise 200BN is backed up by no argument or evidence?
The claim that it's more worthwhile for Apple than it is for a CPU company to have CPU production exclusivity is backed up by no argument, logic or evidence?
What can be asserted without evidence can be dismissed without evidence.
Although most would argue that lower end architectures are perfectly fine replacements, so Apple gets an edge here but can't be said to be a monopoly.
Whether it's unibody aluminum milled frames, bleeding edge injection molding, glass, silicon...
It's done nothing but good for Apple to be aggressive and as vertical without owning the manufacturer as possible. Some of their processes are _years_ ahead of what anyone else can get their hands on, because they buy all the equipment, lease all the floors, and just throw money around like it's nobody's business. AMD and certainly Qualcomm can't touch em.
They are about a year ahead on CPUs. But pretty much everything else (screens, camera sensors, battery) has been on a par with, or behind what everyone else is doing.
When they moved away from unibody laptops they did so by introducing friction-stir welding. That was hot tech on Boeing airplanes at the time. Some of those Boeing patents are still active.
But Apple products tend to have some really interesting manufacturing or technology. Machining steel for the iPhone 4, FSW, putting an Xbox Kinect into the notch, LiDAR, making edge to edge screens without a chin, custom silicon.
Most of this is easily replaced or omitted because it doesn’t really matter much to the end user.
Apple are able to provide unique experiences through their polish and proper concepts. Other manufacturers can even have the features, but they usually don't know what to do with them, make the experience bad and so nobody will care.
Cool! I mean hot.
You mean the "unibody" made of two sheets of aluminum glued together in a way that the heat unsticks the adhesive?
The bit I scanned through shows cracking due to dumb placement of holes. And that’s on the lid, not the laptop body.
Drilling holes through the edge of load bearing elements is still a classic failure mode for home construction, so I’m disappointed but not surprised. There’s a lot of bad blood about their hinges, and I’m not gonna fight anybody on that. They wanted the thinnest laptops, they got them, but not without consequences.
But part of that thinness was actually pretty smart, and that’s what we are discussing here. Using billet aluminum and CNC fabrication for complex shapes instead of gluing (gluing takes large contact points which means more material). Switching to airframe construction techniques was cooler and probably faster. Carving them out of solid aluminum was innovative mostly because it sounded so crazy. Stir welding sounded positively sober by comparison, even though it was hot shit at the time.
No. The unibody was the second gen redesign: https://en.wikipedia.org/wiki/MacBook_Pro#2nd_generation_(Un...
Rossman is handling a first gen MacBook Pro in this video, pre-unibody, and discontinued in 2008.
Not suprised to see Rossman succeeding in mining anti-Apple rage clicks.
Rossman's video is about the pre-unibody, but the first-generation unibody indeed had a heat-unsticks-the-adhesive issue. The EM209 issue (<https://randyzwitch.com/broken-macbook-pro-hinge-fixed-free/>) affected me twice, the second time not covered by Apple (<https://news.ycombinator.com/item?id=21542522>).
Competitor is using new unproven tech. Apple decided to wait.
Apple can't use new tech due to supply constraint (new tech supplier can't make enough to supply Apple).
Apple decided that new tech doesn't offer enough competitive advantage vs profit made selling old tech (720p camera vs 1080p camera).
Whilst the party lasted for apple it can buy the world. But would it last. We have to see. Hence saying it is a monopoly is totally off.
Should you accept, they'll give you a shot of capital in the arm and you'll grow faster than you ever could have to meet their schedules. Soon, your engineers are tied up in daily afternoon stand-ups to go over the latest data with Apple's engineers, and they're dictating your R&D schedule. Your company is effectively dependent upon Apple because investors expect revenue growth. Inevitably, however, they'll discard you when a cheaper competitor comes along, or they decide to take the work in-house.
Should you refuse, they'll poach away your employees, or enable your competitors to do the same.
I haven't thought of a scenario where a vendor can actually rebuff Apple and stay intact.
Which is not to excuse some of Apple’s less moral behaviour.
But to characterise Apple as somehow abusing their relationship with TSMC as a result of what you’ve said is not right. It’s just a normal commercial relationship between a customer and a large supplier.
Also bear in mind that TSMC has a large number of mobile SoC customers - maybe they think that’s their strength and focus on that rather than say photonics. Seems to be working!
[1] https://www.breakinglatest.news/business/samsung-3nm-technic...
Their 2nd Generation 3nm which should be what AMD and Qualcomm are using should he little better than this but we dont have any data. ( yet )
One can explain TSMCs preferential treatment for Apple based on purely commercial terms. Apple is after all the biggest foundry services consumer and they usually demand the advanced nodes which tend to be more expensive.
But there is no such explanation for TSMCs preference for Mediatek over Qualcomm. Qualcomm is generally as large as Mediatek, in fact they used to be significantly larger before TSMC hobbled them.
Well the first explanation that leaps to mind is patriotism (mediatek is a taiwanese company like TSMC, Qualcomm is American). But if that plays a significant factor then perhaps the chipmakers of the world should not so eagerly trust TSMC to make their chips.
Qualcomm has a longstanding reputation as a bad actor; they're the Oracle of hardware. It's easy to believe that other companies would prefer to avoid doing business with them.
Qualcomm as incumbent could've taken too much time trying to negotiate a lower price (e.g. "if you don't lower your price, we'll go to Samsung!").
If everyone else had access to cutting edge TSMC I’m sure Apple’s chips would still be good but I don’t think we’d be quite as impressed.
Eh? The vast majority of Apple chips that will be made on this process will go into phones, and Apple has been doing its own phone chips for about a decade. The M1 won't be quite a rounding error, but it won't be far off; they just don't sell that many Macs.
Microsoft can't supply ARM Windows to Qualcomm's competitors. Windows can't be ported to anyone else's ARM chips because of an exclusivity agreement with Microsoft.
Same goes in phones. The A14 is around 10-20% faster in all metrics than an 888 and has around that much of a process advantage.
The SPECInt suite lasts from 1000 to 3000 seconds. Therefore the 5980HS was at 35W for the vast majority of the test.
That cannot be presumed based solely on the nominal TDP and the total duration. SPECInt is a suite of a wide variety of sub-tests. It is not at all a consistent sustained test like the Prime95 results you are using as the basis for the turbo behavior. Actual power consumption during SPECInt is highly variable, because the workload itself is highly variable. Whether it truly averages out to 35W over the full duration is something that must be measured, not assumed. And you definitely cannot generalize your assumptions to apply to benchmarks or workloads with durations that are not far longer than the 5 minute turbo duration observed under Prime95.
Qualcomm uses the exact same process that Apple uses for the 888 and TSMC produces it, yet it's inferior in every way to the A14
Despite an inferior process, the 888+ and 888 are only 10-20% behind the A14 in most metrics including energy efficiency, despite being on a process with 35% lower density.
Everything I can find suggests new model numbers too.
You mean the exact same LPDDR4X that everyone else uses now? Maybe not in huge quantities but there are a lot of AMD Renoir and Intel Tiger Lake laptops with the same 16GB setup.
If you compare the M1 that uses LPDDR4X it loses or ties in multicore and is barely faster in single core.
6 times?! What? No generational improvement in the last decade (if not two) was six-fold.
https://www.samsung.com/semiconductor/dram/lpddr5/
Samsung marketing says 1.5 times.
400/60 = ~6
Anything using massive linear algebra.
I answered your question with a joke because your question is a joke, what doesn't benefit from a 10-fold increase in processing power? Why would no new use case arise from a widely available 10-fold increase in processing power when the last 40 years have shown that new tech always materializes?
Or maybe you just wanted to make a cynical statement, that all this tech from the last 20 years was pointless, that social media are a net negative for the world, that better video game graphics are pointless because only gameplay matter, that smartphones (only possible because of a previous wave of 10-fold increases) are only addictive little screens and not actually useful in everyday life?
Yes there are genuine new use cases that would appear with a ten-fold increase in processing power. It's a certainty.
You absolutely need high refresh rates and frame rates for VR, or else you get motion sick and/or lose immersion.
You’d pretty much always benefit from higher resolution for VR since the pixels are being placed much closer to your eyes and spread across your entire peripheral vision.
VR reduces the resolution your GPU can handle at the same level of performance/fidelity because two separate images are drawn.
The increase in performance in graphics cards is enabling entire industries to exist and accelerating scientific research.
Phones in your pocket are performing on-device ML and AR in ways previously thought impossible. They’re being used to shoot actual movies that are shown in actual theaters.
Low power tech like smartwatches wouldn’t be possible without these breakthroughs because ultimately faster processors also imply low power devices that can actually do decent amounts of computation.
So to answer your question, new use cases have already been unlocked by simply having more processing power to play with. It’s never been the case that all possible use cases are crystal clear before the tech that enables those use cases exists.
If you want to know why Facebook changed their name to Meta, it’s actually because they see a near-future of VR/AR devices getting a lot less clunky to the point where they can offer a seamless virtual social network and/or truly next generation video conferencing where everyone feels like they’re in a room together. While the exercise may appear to be damage control from an arrogant billionaire, I can see the argument and business case for their vision.
I know I know, they bought NUVIA (which is widely believed to become the savior of Qualcomm CPUs), but we won't see any results of that acquisition until 2023.
Can't imagine their N3 process will be much better, but its good they found customers to finance their latest node with I guess.
Lots of people were asking why TSMC wasn't charging a higher premium for their 7nm when it was the best process in the world, my theory is that they understand the above and need to maintain relationships for the possibility that Samsung (and Intel as they adapt their business model) can come back around with a little bit of luck and a lot of investment.
Secretive Giant TSMC’s $100 Billion Plan To Fix The Chip Shortage https://www.youtube.com/watch?v=GU87SH5e0eI
Apple gaining marketshare because of this should be a matter of concern for many nations.
[1] https://community.cadence.com/cadence_blogs_8/b/breakfast-by...
One day it may not play out. Bombing by china say. But the risk of betting is in the blood of a lot of successful IT firms. For startup easy. For big firm like ibm then and apple now.
It is crazy. Sort of.
ASML can only build so many machines... IIRC, there is also a loop here because ASML needs advanced chips to make those chipmaking machines.
They demand staggering amount of power, purest water, liquid H, liquid He.
And if one of those are missing, they will refuse to work. So you better have backup.
In the end, the cost of the ASML machine won't be that high.
Edit: apparently, TSMC alone is responsible for nearly 5% of the entire energy consumption of Taiwan. Holy. https://english.cw.com.tw/article/article.action?id=2766
A 300mm wafer holds 150ish chips, and TSMC makes 13 million wafers a year. Or roughly 2 trillion chips per year.
Now let's say TSMC expands to 2.2 trillion chips/year with a new fab.
If there is no customer for the 200 billion chips / year, you need to keep building newer and newer warehouses constantly. That's just not a number that can be solved with one or two warehouses.
The extreme scenario you define is a fantasy.
It's a genuine question - because they're in such a crucial industry, and they're taking sizes based on order volume, damaging other businesses?
There are plenty of crucial industries. Lots of them are in US, Europe, and Japan. Would they open their IP? No. The only way to break that situation is starting the war commercially or physically just like how US bullied Japan in 80s and China right now.
This is next level: Without building his own fabs, Cook has dominated microprocessors with a combination of an industry leading design and dominating state of the art manufacturing. This is even better than what Intel had in their dominant period because this strategic pattern can be shifted into different areas of products and their supply chain. The car industry better watch out.
Natural monopolies are not illegal and Apple is selling hundreds of millions of devices that use these chips every year. Apple silicon is in every Apple Watch, Apple TV, iPhone, iPad, Macs of all sizes (soon), headphones, and so on.
Placing limitations on a single company itself would be anti-competitive because we'd be restricting natural monopolies from growing and placing favoritism on other competitors, which itself wouldn't be fair either.
Wouldn't Apple customers suffer just as well because they can't get Apple Silicon SoC due to supply restrictions enforced on them?
Apple isn't stocking chips to prevent others from taking over, they're buying every supply they can to ensure they have enough to sell all year long and they're planning several years in advance.
Put it another way, if AMD was richer years ago, AMD would be doing the same thing to get larger share.
There's nothing stopping AMD and others paying TMSC several tens of billions of dollars to build a fab for them.
Also, let's remember that AMD had their own fab but sold it off (GlobalFoundries).
Like imagine not regulating a natural monopoly that provides electricity. They'll serve only areas they want and will not serve the rest. Good idea? Nope. Requiring them to serve areas equally is a proper method exactly because it's a natural monopoly.
https://en.wikipedia.org/wiki/IPod
https://9to5mac.com/2021/10/23/20-years-ago-today-ipod-chang...
They either needed that many because they were selling that many.
Or bought way more to stop others from buying.
So how much you need is a forecast, or more precisely, a range of forecasts that depend on what you and your competitors will sell. Your order will say "I commit to buying x and want an option on y more" and a larger x gets you a lower unit price for the first x you buy.
If the vendor's production capacity is within your forecasts, or near it, why shouldn't you just commit to order everything and get a nice low unit price? You'll lose if your forecasts are very wrong, but then that's true whatever you do. It sucks for your competitors, who did not order according to their own forecasts.
These are separate things (getting to the point of large scale production and blocking your competitors), e.g they could have played fair with the same initial success without excessive overstocking, allowing their competitors to continue competing and letting market forces decide, and letting suppliers gradually match current demand.
In short, there is no advantage to overstocking to such an excess other than to block your competitors. But you can only do so once you have a foot in the door and enough money to do so.
What I recall is them being production limited over and over again because their best wasn’t always good enough to stay ahead of demand. Nobody has talked about them hoarding.
Blocking competition was a benefit, but not the goal.
Apple is obviously not going to buy less so their competition can have more when Apple needed the supply.
The fact that it's one of severel effects and one of several motivations does not make it unintentional, or even necessarily just a side effect.
It could just as well be simply one of multiple fully considered effects, and surely was.
General business intelligence: You do NOT want one of your 'verticals' (companies that you need to buy supplies from, or companies that you sell your product to) to turn monopolistic; after all, once they are a monopoly they can really squeeze you out.
This holds even if you are also nearing a monopoly or are an effective monopoly (in that you're the only one that can supply it at the quantity and quality required).
There are mitigating circumstances, but most of them don't look good for apple:
* TSMC is stupid. Don't knock it - I've seen companies go for the quick buck, then get killed by the monopolistic monolith they enabled and be surprised.
* TSMC laughs in the face of this and simply isn't taking an apple near monopoly on chips particularly seriously. I can definitely see that - the risk is presumably that apple gains massive increases in marketshare of PCs/laptops and smartphones, but TSMC presumably doesn't think it'll be so high as to risk the situation that TSMC can produce more chips than non-apple buyers could buy. In other words, that apple's total 'TSMC fab capacity' needs won't go anywhere near 50% of what TSMC is likely to ever be able to manufacture.
* Apple is putting pressure on TSMC. TSMC knows this is a risk and doesn't want to, but apple does want this to happen and is making some shady deals so that TSMC does the math and decided that the gains in dealing with apple exceed the losses*odds-it-will-happen of the squeeze if apple is a monopoly buyer for TSMC (as in, of dubious legality and certainly of questionable morality, but then, it's apple, a company. Waiting for companies to act morally is silly, you write laws and set up societal systems to incentivize them to do so instead, companies are amoral (not immoral - they just don't really know what it is, by design).
Your comment seems to suggest it's a smart move for TSMC to just sign one giant deal with apple and be done with it: You can send the lawyers and salesfolk to early retirement, get some money upfront, guarantee 100% sales of all your fab capacity with a party that is unlikely to renege or declare bankruptcy.
It's not. It's a dumb move. Either the C-level execs at TSMC are missing something pretty fundamental or more likely we don't know enough to realize that there's sufficient weights on the other side of the balance to counteract the downside of enabling the monopoly on verticals.
Then, Apple gets a temporary exclusivity on the new node (which they helped finance), but this leaves a lot of capacity on the older nodes. These processes are still really good, the vast majority of TSMC customers don’t care about being on the bleeding edge for the sake of it.
Let’s be realistic: what would have been the alternative for TSMC? Leave Apple’s investment out of the table and waste time bringing their new process online without it? AMD or Qualcomm would not be in a much better position right now. Or TSMC, for that matter.
What is the end game you’re afraid of? Apple takes its business elsewhere? To whom? Wouldn’t this leave TSMC with paid-for facilities they could now use to produce chips for other customers at a discount?
From TSMC's perspective, the entry level market isn't where they make their money. You can make entry level devices on old processes that have a lot of competition. It would be really bad for TSMC if Apple were to monopolize the high end of the market, because that's what uses the advanced process nodes TSMC makes its money from.
> Then, Apple gets a temporary exclusivity on the new node (which they helped finance), but this leaves a lot of capacity on the older nodes. These processes are still really good, the vast majority of TSMC customers don’t care about being on the bleeding edge for the sake of it.
Except that they do care, because they're competing in the same market. Apple is currently making a lot of hay out of the fact that the M1 is more power efficient than PC laptops, which is attributable in no small part to the fact that they're on TSMC 5nm when everyone else is on TSMC 7nm or worse.
If nobody was on the newest node, or Apple was making products that didn't compete with AMD/Qualcomm/Nvidia/Intel in the market, the others not having it wouldn't matter. When the device customer is going to prefer the best one, and the one on the best node has an advantage, it does.
> what would have been the alternative for TSMC?
If Apple isn't signing some kind of exclusivity deal requiring TSMC to give all of its new capacity to Apple, build more of it using their own money so that when it comes online, there is enough for more than Apple. If they are demanding exclusivity, inform antitrust authorities of this.
> What is the end game you’re afraid of? Apple takes its business elsewhere?
Suppose Apple monopolizes the high end market. They take 75% of it using the process advantage, then take the rest as the network effect from that crushes alternatives. They're now your only customer for the newest process nodes, so they squeeze your margins to zero. At the same time, they buy Global Foundries and pour money into it until their process is better than yours, and you no longer have the money to compete because you let all your other high end customers be destroyed.
It's what Apple did to Imagination Technologies and tried to do to Samsung.
Similarly intels process issues showed up when they started trying to make xeon before the desktop processors.
Because your angle completely miss capacity and resources planning. As the 30 years old joke goes, most expensive Fab in the world isn't the leading edge Fab, it is the Empty Fab. And then there is the initial capital and order guarantee. As a matter of fact, it is nearly the same across all industry supply chain. Insert Pizza Doll, Sausage Rolls, Paper Towels, or the recent pandemic event Toilet Paper, and your skill set from production line are all the same.
It is not Apple is without risk, so to speak. They will have to fill the Fab orders, even if somehow no one buys any iPhone. It just happens Apple has never had this problem. Compared to Qualcomm, AMD, Nvidia which all have their fair share of flops and market did not react as they expected. Or you could end up like Nvidia where they had three quarter worth of GPU stocks sitting channel during the BitCoin crash.
Getting Apple ( or the largest customer )'s order filled with a price premium is and will always be the simplest and effective way for the business.
And there is a huge market in GPU, HPC, AI, and Cloud Computing along with forever increasing chips in all segment. TSMC isn't really beholden to Apple like Dialog or PowerVR IMG.
That said, Apple probably just invested money in their 3nm in exchange for priority (or similar) and maybe TSMC is being a bit short sighted.
Apple is very far from monopolistic market share of smartphones and even much further away with Laptops. Apple with their intentionally overpriced products caters to a branded premium segment. They are not even aiming to take the mass market of low to mid level smartphones that makes up most of the market outside the US
I would also not be afraid if I were TSMC, especially as this is just about the new 3nm process that is catering to the highend market and doesn't eliminate the existing business.
Your comment reads a bit like there could be a distopian world where a shortage of 3nm results in Apple becomeing the only option for phones and laptops
Obviously TSMC doesn't care about VC's, but the same reason why VC's prefer a large number of small customers still applies.
This is intriguing and I would appreciate if possible to get any sources to read up on this?
I think the caveat/variable is "all else being equal or close to equal."
If having 10 customers will net you $10 total per month, and having one large customer will net you $100 per month, it is really hard to justify going with the $10 customers. But if those 10 customers will net you $10 EACH per month, then the 10customers is a better situation.
When people see small number of large customers, the implication is you are basically just running an outsourced dev/consultant/staff aug shop. Less product vision and ability to "scale".
Especially for a new untested company no customer is sure they actually need yet
This depends on your business strategy: having a single or few customers makes you much more dependent on their whims and their negotiation power. If you have more customers you are in a better position for pricing negotiations.
To me this says as much about Intel as about Apple. When Intel has a new chip or a bin that has small yields, they can’t supply a large customer. But MacBook had maybe a third of the market share it has now, so that’s a relatively small niche you can supply while you boost your yields.
And AMD and Qualcomm want more access to it than they have so they complain to the press. This is the supply chain folks.
Arguably, bleeding edge silicon fabs are natural monopolies. Pushing transistor sizes down to atomic scales entails non-linear increases in costs to get decent yields. That's why there are very few fabs at the bleeding edge any more.
I’m not claiming Apple respects TSMC more than Qualcomm does (but does Qualcomm respect anybody? They are practically the bad guys in any number of stories), but this is what respect would probably look like.
Speaking of Qualcomm, Apple is trying to compete with them on radio chips, so leaving TSMC probably works to that goal.
We're in the middle of a chip shortage, though.
In all, you're characterising Apple, competing fiercely in an open market, as preventing other people competing by being too good at it. That is a bit ridiculous. If AMD and Qualcomm die out as a result, it will be because they couldn't keep up, which is capitalism functioning correctly. Even under those very broad laws, the only party you can remotely fault is TSMC, and even then, they are literally just taking from the highest bidder, which is again capitalism functioning correctly.
AMD didn’t run a monopolistic business with 90%+ profit margins for years, and couldn’t have done so.
Apple has never had a monopoly on anything.
They just use the same silicon for a lot of products and have good profit margins, so they can place large orders with TSMC.
> Apple has never had a monopoly on anything.
We'll see what the EU antitrust commission says about that, so far there's been no legal decision in an independent jurisdiction yet.
Independent of what?
But from a physical perspective, they've never had the kind of monopoly Intel enjoyed for decades.
The courts have ruled that the iOS App Store is not a monopoly.
They do on thier secoundary markets, (esp App Store) but they are also being monopolistic with replacement parts and repair.
Apple sells tools and parts to anyone who wants them, so it’s not really clear what you are talking about when it comes to repair.
But even if there was some merit to that point, it’s silly to suggest that their repair policies have anything to do with them buying a lot of silicon from TSMC.
- 24/7 worldwide availability, instant payment/app download
- Easy re-install after deletion, you still own the app even if deleted
from the device
- Region restriction
- Separate app pricing by region
- Revenues paid to developer from multiple region currencies without
conversion fees
- Tax calculation and collection
- Instant customer refunds
- User rankings and reviews
- App store advertising in category listings
- Video previews of the app in operation
- Packaging of media content allowing developers the ability to load game
levels as needed. This allows a user to start playing your game quickly.
- App sales statsIn other words, they couldn't move to 5nm because they wouldn't be able to procure any significant volume from it. Why bother with 5nm if you wouldn't be able to ship products?
That's exactly why NVIDIA ditched TSMC in favour of Samsung despite their inferior process.
https://www.youtube.com/watch?v=8MxC6eVjPFQ
So, Apple CPUs are not that great at the end.
By the way, the guy on the video forgot to turn off power saving mode on S21 (-> less performance) and S21 still beat iPhone...
Here's the reverse video where the iPhone beats the S21 ...
TSMC also seems to be taking a big gamble, pissing of existing good clients.
Edit: (1) https://semianalysis.com/apple-cpu-gains-grind-to-a-halt-and...
"somehow delay their competitors from launching new and better chips"
Both AMD and nvidia have designed and built on older processes long after newer process capacity was available to them. They aren't being delayed at all.
Apple committing to production capacity is neither "Apple first", nor is it Apple "delaying" competitors. There is finite production capacity, and customers spread across the available options. There is literally nothing notable or interesting in that.
But once you add words like "Mutiny", you pander to a very gullible crowd.
NVidia and AMD do not have access to the latest processes. There is not enough capacity left at TSMC because of the Apple orders getting priority, so they have to stay on older processes. NVidia Ampere was not designed for Samsung 8nm, it was designed for TSMC 7nm but they couldn't get by with the leftovers so they had to put the consumer line on Samsung 8nm.
Apple is absolutely first. They are using their warchest to ensure their competitors do not have access to high quality processes.
Reality: Apple had designs ready and committed to production. Indeed, the first 5nm production was Apple chips, and HiSilicon. Yes, HiSilicon with stock ARM cores somehow managed what AMD and nvidia (who are barely even Apple "competitors") couldn't.
"They are using their warchest to ensure their competitors do not have access to high quality processes."
Again, nvidia and AMD both tend to hang two processes behind (making their chips cheaper, higher yield, and because they have a slow design iteration for new processes), but let's pretend that they were really trying for the newest process (you know, the one that HiSilicon managed to get). Apple's evil here is apparently simply making products. Because you know that Apple isn't just reserving unused capacity -- they are literally making A15s and M1s and soon 5G models in the hundreds of millions. It's pretty amazing that simply making your products is now nefariously blocking competitors (if you're ignorant enough of actual fact to think that was actually happening).
And AMD/Nvidia are direct competitors to the entire M1 line.
By the way, AMD die size production only for the next-gen consoles in less than a year is equivalent to ~100 million iPhones as they use 4x more silicon, and that's in supply constrained conditions. AMD and NVidia absolutely can move more wafers than Apple, they just got outbid by a company that can buy them both.
The issue is that we're in a market where Apple can afford to pay suppliers so much that other companies literally cannot afford to compete. Their positions in the market is slowing down technological advancement.
Don't manufacture arguments and then soundly beat them down. They didn't prioritize the new process so they ended up at the back of the line. They made a choice (to virtually no market detriment, and improving their financials in the process). As mentioned, both companies often run on older processes because power simply didn't matter, and wafer yield just mattered more. nvidia often has monster chips because...eh. If your GPU or giant AMD processor uses 300W, eh, that's life.
This may come as a shock to some, but despite all of the nonsensical rhetoric posted on HN (by Apple naysayers who have to dismiss anything Apple does with One-Simple-Trick nonsense that has zero association with reality -- I recall when it was the magical "big cache", as if Apple was sneaky having big caches, cheating the system), process improvements are not as big as they are held on here. They are of course a benefit, but a process improvement yields power or performance or density benefits (yup, even that last one is conditional on other factors), but the fiction spinners declare that no really it's everything all at once. It doesn't work like that. So they opted not to prioritize it.
Yet read your other posts and not only would AMD and nvidia be world's better than Apple in every dimension (see above about the Apple naysayers and their ignorance), but also they couldn't afford to because I guess the $1500 GPUs and $5000 CPUs just can't afford the big bucks Apple can spend making a smartphone CPU (where apparently they're outbidding everyone on every component, yet also simultaneously having by far the highest profit margin in the industry...my normal brain cannot even comprehend the lengths of how ludicrous and contradictory this nonsense is). Oh, and HiSilicon making chips for discount, very low end smartphones also apparently has more money to blow on this.
"The issue is that we're in a market where Apple can afford to pay suppliers so much that other companies literally cannot afford to compete"
So the bill of materials for Apple products must be enormous, right? Oh wait, they're absolutely rolling in profit, with some of the highest profit margins in the business. No, it isn't that whatsoever. Absolutely nothing indicates that Apple used its "warchest", or that it is paying a penny more than anyone else. Can you point out a single authoritative source claiming that? Because actual economics say no, that's utter nonsense.
But sure, Apple is "slowing down technological advancement" in the same post where you declare that they're paying more for that technological advancement. The desperate lengths this rhetoric has to go, with laughable self-contradiction, is embarrassing.
??
> [...] somehow delay their competitors from launching new and better chips while Apple scrambles to find new talents in the limited pool available.
Nonsense, Apple buying capacity because they actually need to manufacture that much stuff. Anything else would be completely idiotic.
> Apple has bet their future on Apple Silicon, and it badly needs to retain a lead on other competing CPU for a year or two, to capture sufficient market share (the perception of Apple Silicon leading over other CPUs is very important for this).
Share of what market are they supposedly wanting to capture? Share of PCs in general? They care about share of premium devices, where they are very well represented.
Share of Macs themselves? That'll solve itself, since Apple won't be offering non-Apple Silicon Macs at some close point in the future.
> TSMC also seems to be taking a big gamble, pissing of existing good clients.
They take the offer with the most money attached, no one's (neither AMD nor Qualcom) surprised by that.
Taking the offer with the most money attached is not a good idea in the long term. They're currently providing unprecedented leverage to Apple.
Apple Silicon needs the performance crown as the comment you're replying to mentioned. Having a good market share in the premium sphere is not good enough - if instead of trading blows but still ending up ahead of AMD, AMD processors also were on 5nm with DDR5 memory they'd be plainly slower. Then Apple would be in the position where their laptops are slower and have worse battery life than the competition and also Docker doesn't work right for developers and some of your VSTs don't work at all, and few people would take the pain of changing architecture for worse performance.
Their SoCs are pretty good. But they're only good because Apple has almost exclusive access to TSMC 5nm.
I'd love to hear about how the M1 isn't competing against AMD and NVidia.
You’re making claims about hypothetical AMD / Nvidia performance at different nodes which are impossible to verify.
And Apple is clearly taking a huge risk by investing massively in the latest node. At the minimum they deserve credit for that.
In any case, if someone said that the M1 was unquestionably faster than anything else on Earth it was not me. It does not change the point that asserting that better is somehow worse is still stupid.
> I'd love to hear about how the M1 isn't competing against AMD and NVidia.
To whom are they selling their GPUs?
> We believe Apple had to delay the next generation CPU core due to all the personnel turnover Apple has been experiencing ...
https://semianalysis.com/apple-cpu-gains-grind-to-a-halt-and...
> Chief Architect for all Apple CPU and SOC development. For CPU, lead the Cyclone, Typhoon, Twister, Hurricane, Monsoon, Vortex, Lightning and Firestorm architecture work. Chief Architect for Apple MAC hardware platform M1 Pro and M1 Max.
I guess they were in the works long before he left.