> Thirdly, the success of an Apple phone would not have been hard to predict. In 2006 Apple sold nearly 40 million iPods.
I don't like how this whole article is framed as how obvious it was that Intel made a huge mistake. Hindsight is 20/20, there was no such data available.
It was not predictable that the iPhone would be a success, especially not to a degree of Intel sacrificing every profit calculation in mere "hope" of huge volume later.
The numbers were there for the iPod. Intel could have just pitched to become the SoC-supplier for the iPod and didn't do it. They surely crunched the numbers several times and it didn't work out.
The first iPhone was then built on the iPod platform with an Infineon modem, so the better question to ask is probably: Why wasn't Intel interested to supply the CPU for the iPod?
No, it wasn't obvious. But it coming from Apple, who choose to go after the portable music player market ignoring the new hotness PDA and that ipod segment becoming the majority of its market cap. When that happened, as the CEO of #1 cpu maker, you have to spend more than a cursory effort into what is Apple's bet on this. Even to me, a nobody, had guessed at the iphone rumor, before the magically reveal, that it will be become Apple's main product, way surpassing even the ipod. Apple choosing to ignore the PDA market gave me an almost certain conclusion that they only ignored it due to technical challenges to produce a rivaling product that they were proud of and after years of wait, they were finally able to produce it.
So yes, hindsight is 20/20, but as CEO of intel, he failed miserably from not putting more effort into the decision making.
> Even to me, a nobody, had guessed at the iphone rumor, before the magically reveal, that it will be become Apple's main product, way surpassing even the ipod.
At the time of discussion with Intel, it was not an explicit phone project, it was _A_ Apple project. Over the years, Apple spent millions on projects to disrupt some markets which either didn't materialize at all or ended up not disruptive/competitive in the market.
There are countless examples of this. Some Moonshot where Apple thought they knew better what the market want to just figure out that it doesn't work.
The project Steve Jobs was talking about could as well have been one of THOSE projects, and Apple tried to get a established and well-working component at a price which would potentially destroy Intels business with its other customers.
> but as CEO of intel, he failed miserably from not putting more effort into the decision making.
But who said that he didn't put effort in the decision making? According to Otellini himself Apple was demanding a price lower than Intel's COST.
At the end of the day, there was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more and that price was below our forecasted cost. I couldn't see it.
Which was already the case at that time, with Blackberry, Palm and HTC dominating that market using Intel XScale.
Intel tried to corner that market by deviating from ARM with their custom MMX architecture. It worked with Windows PocketPC as a OS-supplier, but HW-vendors with their own OS didn't want to limit their supply-chain to a single CPU-platform.
But yes, their belief in x86 being superior surely clouded their judgement on XScale's future potential
And they would have been just as shitty as the Blackberry. Hell, even Google had to go back to the drawing board with their Android prototype because they still thought a physical keyboard on a phone was king....right until they caught a glimpse of the first iPhone.
i loved the blackberry keyboard and would take it anytimes over this fucking iphone keyboard. The iphone keyboard is a step backwards. people like haptic feedback.
The middle of the bell curve, which dominates the market, has a need for drool-proof interfaces that many of the sort who frequent HN and similar interest communities do not.
I think the gap was market size estimation. Intel was (is?) famously not interested in businesses that can’t generate $1B of incremental revenue in the first year or two.
iPhone was predictably huge, but I don’t think it was a sure thing it would be this huge. And Intel’s finance-driven culture probably did the math and decided it was better to pass because of the uncertainty. Finance hates uncertainty.
And of course by the time a market exists for that dollar figure, Intel has been beaten hopelessly, such that there is no hope for as stodgy and apathetic Intel is culturally.
This is what monopolies do, it breeds laziness in the good days, and as a company backslides on the morass of apathetic incompetent mbas that settled into their middle management Machiavelli fiefdoms, it becomes a clinging game to make retirement before layoff
yeah, but according to the article it wasn't even clear at that point that it's a phone project. It was too early to even make such predicitions.
At that time, it could as well have been Apple's attempt to a build an Apple TV (which never materialized), a Set-Top-Box (which was reworked for years to then end up as a iPhone-in-a-box AppleTV), a Premium Wi-Fi network, a game console,...
Don't forget that Apple's previous phone effort in 2005 was a collaboration with Motorola on the ROKR, aka "iTunes Phone", which was a normal motorola phone except it could sync with iTunes and play media like an iPod
I agree. Apple was ascending again and everything they did seemed to be a hit. I think the real question is how large of a bet did Intel really need to make. No it might not have made sense on immediate ROI alone, but Intel had only to defend its position.
Intel misjudged how much the part would cost to make. That mistake led to a huge missed opportunity.
> Even Otellini betrayed a profound sense of disappointment over a decision he made about a then-unreleased product that became the iPhone. Shortly after winning Apple's Mac business, he decided against doing what it took to be the chip in Apple's paradigm-shifting product.
"We ended up not winning it or passing on it, depending on how you want to view it. And the world would have been a lot different if we'd done it...
At the end of the day, there was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more and that price was below our forecasted cost. I couldn't see it...
And in hindsight, the forecasted cost was wrong and the volume was 100x what anyone thought."
> Intel misjudged how much the part would cost to make
I just don't buy it. If Intel realized in hindsight that they could price out a competitive chip, why didn't they sell it to Android manufacturers? So much effort goes into making us lament a world where Intel iPhones existed, but I don't even see the appeal. Intel didn't have a vested interest in ARM (an attitude persisting to this day), they weren't going to acquiesce to Apple's desired margins then or now, and Apple would always have the logical opportunity to cut out their chip middleman. It's about as appealing as a shit sundae sat on a trap door.
> That mistake led to a huge missed opportunity.
Again, was it really that huge? After licensing cost to ARM, core design fees and the bulk deals worked out, I can't see Apple allowing Intel room to breathe. From a certain perspective, it almost seems like Intel never had any desire to genuflect for Apple and only signed the Mac deals to proliferate their existing design catalog. I can't say for sure what the financials worked out to be, but considering Mac market share at the time it makes sense why Intel wasn't super motivated. If we could look at a modern iPhone bill-of-materials, I don't think either of us would feel bad for Intel.
Arguably, TSMC being Apple's preferred fabricator is why it has a current market cap 10x larger than Intel.
Apple probably ends up getting chips only slightly above cost from TSMC. (It pays $18 B a year, which is pretty low for the CPUs in roughly 400 million phones, watches, tablets, Macs, AppleTV, etc. it sells every year) but it also pays well in advance the costs for building the fabs for those chips. This gave TSMC the scale needed to be the world's leading semiconductor manufacturer, which means it simultaneously has lower costs from economies of scale, as well as more pricing power, from being able to offer leading edge nodes first.
Sure, but also let's not pretend like Apple would pay the same amount for 14nm++++. Even if TSMC didn't exist, Samsung could have beaten Intel on a density roadmap. The looming possibility (if not inevitability) of Apple designing in-house was casting a shadow even then, and today it only makes less-and-less sense mourning the separation of both companies. They wanted different things, and Intel didn't have to rely on TSMC-level density to make their case.
It's not like the Core products that went into Macs - making iPhone chips is a bespoke, ground-up production. It barely makes sense for Apple to insist on using Intel fabs unless they're heavily insecure about outsourcing their supply chain. It makes absolutely no sense at all for Intel to design ARM chips when they don't own an architecture license and would ostensibly only purchase one to make Apple happy. Again, I feel like this entire "lament" is manufactured by Apple fans eager to rewrite history with Apple making generous offers to domestic American businesses. In reality, Apple's endless desire for hardware margins drove them to dictator-states and third-world manufacturing hubs to meet their quota. It's okay for us to pragmatically admit Apple never had any business with Intel, and Intel was right not to convince them otherwise.
If Apple went with Intel from the beginning I doubt they would have ended up developing their own chips and they'd probably be just stuck with x86 on Mac XSScale/whatever on mobile.
> ARM chips when they don't own an architecture license and would ostensibly only purchase one to make Apple happy
They did certainly own one. XScale was a custom core and not licensed from ARM. IIRC pretty much all high-end PDAs used them so it would have been the default/safe choice for Apple had Intel decided not to abandon it.
Is there any information about those terms? Could they really have been worse than whatever deal with Apple has with ARM these days?
The main downside with ARM/XScale is that Intel would be forced to compete with all other manufacturers and wouldn't have been able to charge their monopolistic/oligopolistic margins on their products. Of course Intel being Intel they could have just added enough proprietary extensions to solve that.
> Apple's endless desire for hardware margins drove them to dictator-states and third-world manufacturing hubs to meet their quota
Sure. But Apple wasn't really in a position to do that back in 2005-2007. It took years for them to build that because they could do it iteratively, I doubt Intel would have allowed them as much freedom as Samsung though.
>I don't think either of us would feel bad for Intel.
Except TSMC? It is not like Intel wont have its margin in making a chip for Apple. It will just be much smaller margin, or margins of modern day Fabless Soc Design company. The problem is Intel misjudged the volume, which ultimately meant they also miscalculated the whole OPex and Capex equation. And Intel dont want to do low margin business ( by their standards ).
Most of your statement suggest Intel would have to make a loss in working in Apple. And yet TSMC has been enjoying their ride purely because of Apple.
TSMC doesn't have to pay ARM licensing costs or design chips for Apple. Their relationship is markedly different from what people are suggesting here, and is profitable because they more or less require Apple to make upfront investments to reserve upcoming nodes.
Intel's margins made sense when they were selling pre-designed x86 CPUs for use in Macs. It expressly does not make sense when a company notorious for loving low margins asks you to design and manufacture a chip you won't be legally allowed to use anywhere else. Again - Intel was absolutely correct to recognize Apple's attempt to cuck them here.
Intel already had an ARM license they had used to create their XScale line.
They decided to sell that line of business off to Broadcom instead of fabbing a chip for Apple.
Later they decided their failure to enter the mobile device market was a big enough error that they burned through about a billion dollars in a failed attempt to push x86 as a viable alternative to ARM for mobile devices.
Pretty much all PDA's (even devices like the first gen Kindle) were using Intel's ARM chips. There is no reasons to believe that whatever Apple needed would somehow have been extremely different from what their other clients used (and it's not like Apple had that much bargaining power those days anyway).
> Intel was absolutely correct to recognize Apple's attempt to cuck them here.
I see it the other way around. Unless Intel somehow severely fucked it up (not a distinct possibility, just Intel being Intel) Apple would probably be stuck with their chips both on mobile and laptop/desktop to this day.
>
I just don't buy it. If Intel realized in hindsight that they could price out a competitive chip, why didn't they sell it to Android manufacturers?
Intel later attempted to bring x86 SoCs to (Android) smartphones (for example the ASUS ZenFone 2 was x86-based), but at that time Arm processors were already quite established in the Android ecosystem, so this attempt was a commercial failure.
Which really further proves Intel's point - they don't want to design phone chips. They wanted to (and did) ape off old nodes for nearly a decade, sat on the same designs and never really competed for the mobile market. There wasn't a world where Intel pays for an ARM architecture license, designs Apple's cores, and then gets cheated out of a fair price in the dealmaking room. It's not worth it for them, had Intel signed onto such an agreement they'd only be guaranteed to cuck themselves in the future.
We can argue for or against Intel acquiring an ARM license in an alternate reality, but truth is they don't care. Apple wanted to turn them into something that would never work, and even if they got what they wanted, Apple's current supply-chain squeeze would have obviated any chance of Intel sticking around.
> There wasn't a world where Intel pays for an ARM architecture license, designs Apple's cores,
They had an ARM license and they had their own high-end cores that pretty much all PDAs used. In the early 2000s if you wanted a high-end ARM chip Intel was the default choice.
> We can argue for or against Intel acquiring an ARM license in an alternate reality, but truth is they don't care
Well as I said they already had one in our reality and could have easily ended up in the position Qualcomm is now if they played their cards right (well realistically in a better position because Qualcomm historically struggled keeping up with their own core designs and ended up switching to Cortex while Intel had a significant head start).
You're quoting all the stuff TFA tries to debunk. Intel simply didn't have anything to offer at that point. XScale already lost edge and about to be sold, a matching x86 chip not on the horizon yet.
Intel was perfectly capable of fabbing the ARM chip Apple wanted.
They just screwed up the cost to manufacture estimate which led them to make a bad call.
That's straight out of their CEO's mouth in an interview, so there isn't anything to "debunk".
Later, they would attempt to correct the error by selling x86 chips way below cost in a failed effort to break into the mobile device market.
> In order to make these platforms viable for the low-end and mid-range of the tablet market, Intel provided system vendors with "contra-revenue" subsidies. These platforms simply required an expensive set of surrounding components to work, so Intel "made up the difference," so to speak, to customers who used its platforms with subsidies.
This program likely cost Intel, and its stockholders, in excess of a billion dollars.
The PDA/Tablet market was a reasonable size, the XScale CPUs were the highest performance ones in this sector, Intel could have seen that this market would grow.
In 2006, Intel supplied their XScale design to Palm for the Treo 700, to Blackberry for the Pearl and Curve AND to HTC for use in the XDA/MDA Windows Mobile line, in 2006 arguably the much bigger wins as all those were established constantly-growing product-lines already on sale globally.
Apple demanding a (probably much) lower price-point to have the SoC applied in a mysterious unknown project without a volume-commitment could as well have destroyed the business and profit-margin for the entire Xscale business for Intel and its other customers...
There's no reason for other customers to strangle Intel or move (again) to TI OMAP platform when they find that Apple as a new entrant in their market got favorable terms from one of their suppliers...
In 2006, Intel has just won back Palm from TI and were under incredible pressure from Qualcomm who had their MSM7200 SoC in the pipeline (which integrated both CPU and Modem in a single low-power package).
> Intel often cuts prices to favored customers to win business, for example they did so for the original Xbox CPU, to prevent an AMD from being used.
There's "cutting prices" and then there's selling BELOW COST, without a volume-commitment from the customer for long-term break-even...
The Atlantic article on the interview with Otellini (which this article linked to), was very clear that Otellini wanted to say yes to the iPhone project but was swayed by data:
> "The lesson I took away from that was, while we like to speak with data around here, so many times in my career I've ended up making decisions with my gut, and I should have followed my gut," he said. "My gut told me to say yes."
It's not even sacrificing profit; it's actively reducing profits by selling at a loss.
Did you miss the part of the quote where he shares that Intel's cost estimate was wrong?
Also, after the magnitude of their error to get in at the ground floor sank in, Intel had no problems with selling x86 chips at a loss trying to break into mobile devices.
I didn't miss that part of the quote. And no it would make no difference to Otellini or to the discussion in this thread because he also didn't know the estimate was wrong when he made the decision.
Yes, but that's just an outcome of Intel's/Otellini incompetence (in that specific case) rather than anything else. He used bad data to make an (arguably) bad decision.
The iPod Classic when the iPhone came out had a CPU estimated to run around 80 Mhz vs. a 600 Mhz chip underclocked to 400 Mhz in the original iPhone. The software is also completely different. They didn't build it on the iPod platform at all, though I believe that was considered in the early stages of the design.
Apple used a newer generation of Samsung's CPU used in the iPod to develop both the iPod Touch as well as the iPhone, both introduced in 2007.
That's also a reason why the iPhone was not designed around a more power-efficient SoC which combined the CPU and the cellular Modem into one package: The Modem was the add-on to the architecture which turned the iPod Touch into that exclusive Phone they built for AT&T.
> Apple used a newer generation of Samsung's CPU used in the iPod
I mean, that's true only in the same way as your phone uses a newer generation of the CPU in your washing machine (these days, both are likely to contain ARM cores of some sort). They were both ARM SoCs made by Samsung, but the relationship kind of ends there; totally different ARM core, totally different peripherals, totally different everything.
That's not what I'm saying. They developed the next generation iPod on the next generation of their iPod CPU supplier platform. And then they added a modem and made it a phone.
They literally utilized their existing iPod supply chain to build the core BOM of the product.
The pre-Touch iPods (the iPod Touch only came out after the iPhone) used very small chips; this just wasn't an area Intel had an offering in at all, and it would have been very low margin.
> The first iPhone was then built on the iPod platform
Eh? No it wasn't. The iPod that existed when the iPhone came out was an 80MHz ARM7. The Touch (a 400MHz ARM11) came out a few months after the iPhone, and was essentially an iPhone with the cellular equipment stripped out.
> I don't like how this whole article is framed as how obvious it was that Intel made a huge mistake. Hindsight is 20/20, there was no such data available.
Possibly because Otellini framed it as such?
The latter part of the article answers your other points.
Otellini's gut feeling was to go for it, the numbers didn't pencil and he decided not to and then when it went on the market the volumes were two orders of magnitude higher than estimated. Should've went for the reality distortion field... as you said, hindsight.
In 2006, Intel supplied their XScale design to Palm for the Treo 700, to Blackberry for the Pearl and Curve AND to HTC for use in the XDA/MDA Windows Mobile line, in 2006 arguably the much bigger wins as all those were established constantly-growing product-lines already on sale globally.
Apple demanding a (probably much) lower price-point to have the SoC applied in a mysterious unknown project without a volume-commitment could as well have destroyed the business and profit-margin for the entire Xscale business for Intel and its other customers...
> Apple demanding a (probably much) lower price-point to have the SoC applied in a mysterious unknown project without a volume-commitment could as well have destroyed the business and profit-margin for the entire Xscale business for Intel and its other customers...
One of the key premises of the article is that Apple likely didn't know it was bidding for the iPhone and the decision was taken on that basis.
Why is why saying 'Intel turned down the iPhone' is problematic.
Hindsight is also where you can fudge the numbers to justify your mistakes. Whenever you're calculating things, there's always more optimistic and more pessimistic assumptions. Given that competitors were able to supply Apple, it seems slightly dubious for Intel to say that there was no way it'd be profitable no matter what the volume.
Given Intel's fat margins at the time, it seems like a more likely explanation is that it wouldn't meet their margin expectations and that they wanted to keep showing their investors the great results they'd been used to in the desktop/server/laptop markets that Intel was truly dominating at the time.
Also, when you're a company, you have limitations on how much you can do. You can hire more people and build more fabs, but it can be hard to grow your capacity. Even looking at TSMC, Apple has been using all of their 3nm capacity for the past year. If Intel was thinking about where to allocate its scarce resources (whether that be engineers, fab capacity, etc.), it might make more sense to concentrate on the areas where you're going to have a significant advantage and higher margins - unless you really know that Apple is coming out with something world-changing, which you wouldn't know in advance. Intel was still making enormous gains in the x86 market at this time and giving users big reasons to keep upgrading their machines. 2008-2010 Intel doubled their performance. The point is that there was certainly an opportunity cost to betting on Apple's new mystery device where they wanted a low-margin chip and at the time Intel's x86 business was booming.
I think there can be other issues as well. Why didn't Intel become the iPod SoC? Maybe Intel didn't have such a low-spec'd part for that purpose. The XScale processors were higher spec'd and Apple certainly didn't need that kind of speed for an iPod. Even with the iPhone, the article notes that the Cortex A8 designs were closing the gap with Intel's XScale, but the iPhone didn't use a Cortex A8 until the iPhone 3GS in 2010 (the original and 3G used older ARM designs). So part of the issue might have been that Intel's XScale processors were powerful beyond what Apple wanted to pay for at the time and it would have cost Intel money to make what Apple wanted (a low spec processor) while others already had that available.
I think it's also hard to say that this decision meant that Intel lost out on the iPhone. Realistically, if Apple was going with an ARM architecture, Intel would have lost the iPhone in a few generations anyway. It's not like Samsung got to keep Apple's iPhone CPU business. Maybe Intel could have kept Apple's business with heavy investment in XScale, but Apple bought PA Semi in April 2008 (less than a year after the original iPhone launch). So it seems like Apple was looking to build their own processors even before the iPhone became huge (the original iPhone only sold around 7M units while iPods were selling around 50M/year at the same time). Even if Apple had chosen XScale, Intel probably would have lost the iPhone business.
And Intel probably wouldn't have made up for it on the Android side. Qualcomm's control of CDMA (important for the US) and subsequent domination of high-end modems was used to reinforce their CPU business. Maybe Intel could have overcome that, but we've seen that modems are a hard business - Intel failed in its modem business and while Apple has seen huge success in their CPU business, they haven't had the same success in modems. To put it in perspective, Apple bought PA Semi with 150 employees in 2008 and 4.5 years later had new custom Swift cores in their A6 CPUs. Apple bought Intel's modem business with 2,200 employees in 2019 and 5 years later they'll be introducing the iPhone 16 still on Qualcomm modems.
So did Intel truly miss out on the iPhone? Maybe somewhat. However, it kinda sounds like Intel's XScale business was too high-end and even if Apple had selected XScale, they were still going to be making their own CPUs given that Apple could license the ARM architecture (unlike x86). There certainly was some space for Intel, but it would have been a difficult fight even if Intel were committed to it. Qualcomm's strategy of tying together its patents, CPUs, and modems to reinforce each other is hard to overcome.
In hindsight, sure: the mobile market would have been worth fighting for. But it seems like it would have been a tough market to crack into given what Intel had (a processor too high-spec'd for what the market wanted at the time) and the fact that vendors could easily switch away from Intel to any number of other ARM manufacturers, including ARM's reference cores. If Intel had bet big on mobile, they probably could have made it a good business for them, but it would have been a big gamble for a market without good barriers to entry and where competitors like Qualcomm might have their own barriers.
> So it seems like Apple was looking to build their own processors even before the iPhone became huge
So but possibly only because they had no alternatives. After all building your own chips is a huge long-term investment with all of risk. Intel fully committing to XSCale would probably have significantly altered that calculation.
> license the ARM architecture (unlike x86).
They still had to design their own cores which isn't cheap or straightforward. Something even Qualcomm struggled a lot with (and ended up failing at).
> So it seems like Apple was looking to build their own processors even before the iPhone became huge (the original iPhone only sold around 7M units while iPods were selling around 50M/year at the same time).
This is almost certainly true once they decided that iOS was going to be based on MacOS. Those software design choices had to have been based on a very aggressive hardware roadmap
To be clear the article doesn’t frame it as being clear it was a huge mistake at the time - rather that it’s clear now - and seeks to explain why that decision was made.
>I don't like how this whole article is framed as how obvious it was that Intel made a huge mistake. Hindsight is 20/20, there was no such data available.
Intel is a big reason things are so great, but also so crap, in computing.
Try to remember, the computing revolution didn't start in the 21st century. It's just got incredibly insane, in the 30 years of personal computing.
>Why wasn't Intel interested to supply the CPU for the iPod?
Because Intel and Apple had beef, long-since, already. The iPod was just another mp3 device, until - suddenly - it wasn't.
The writing on the wall for personal computing has been there for all of the computing pioneer companies .. the fact that Intel and Apple didn't get connected, is as much about the fact that these people were really in competition with each other from the beginning.
Apple was always going to own its own fabrication capabilities. From the outside of the box, all the way in.
Intel wanted all the other boxes, not just Apples, too ..
> Apple was always going to own its own fabrication capabilities
I'm not sure that was a foregone conclusion back in 2005-2007. If anything Apple was forced to design their own chips because nobody else was able to offer what they wanted. M-series is a thing because Intel couldn't build the low-power/higher-performance CPUs that Apple wanted.
And Apple only decided to own its chip design-capabilities after the iPhone became a smashhit and Steve Jobs felt betrayed because Samsung used the same ARM CPU-architecture to build the first Galaxy S Smartphone...
It's anyway a big claim that Apple was always somehow competing with everyone in the market. That's the company which rebranded HP Laser Printers, made an HP-branded iPod, made an iTunes phone with Motorola, just moved from IBM PowerPC to Intel for their PCs (!), decided for a while to not make a 5K display but let LG build an official one instead, pushed LaCie as an external HDD partner, etc.
Apple always had ambitions to "wow" with its customized thoughtful products, but before the "iPhone money" they barely ever had enough order-volume to dominate a supplier...
If it wasn't the iPhone, it would have been something else. The logic that drove the decision and not the end result was the undoing of Intel and the logic would have remained the same as long as the culture remained the same.
I don't like how this whole article is framed as how obvious it was that Intel made a huge mistake. Hindsight is 20/20, there was no such data available.
It was not predictable that the iPhone would be a success, especially not to a degree of Intel sacrificing every profit calculation in mere "hope" of huge volume later.
The numbers were there for the iPod. Intel could have just pitched to become the SoC-supplier for the iPod and didn't do it. They surely crunched the numbers several times and it didn't work out.
The first iPhone was then built on the iPod platform with an Infineon modem, so the better question to ask is probably: Why wasn't Intel interested to supply the CPU for the iPod?