(2011). And Cook was right. I worked at Intel for a few years during that decade and the foundry efforts were just not set up for success; in my area, they hired a bunch of new people, put up a firewall between us internal folks and the foundry folks, then without any guidance turned them loose. I was not even allowed to talk to them to troubleshoot equipment issues. They also got all of the equipment that we’d rejected for various reasons like poor process control, so they were newbies with worse equipment trying to start up a new group without help beyond what vendors would provide (for $$$)
interesting perspective. Care to elaborate a bit more? Where you in the design department? Why did they put up chinese walls? Was it to enable the foundry to have other customers other than Intel designs? Why did you have similar type of equipment? Were you also manufacturing chipsets? If so, why didn't they expand your division to become a foundry?
I was in a fab module. The firewall part makes sense to not cross pollinate IP internally vs. externally, but it was taken to the extreme and management moved zero internal employees over to external so it was Intel’s tools and recipes but not the talent who knows all the tribal stuff.
This situation has been and maybe always will be common. Those who view themselves as integrated manufacturers of *products* often cannot equally package, deliver, and sell their component *services*. No matter how good those services are at delivering the end product—and sometimes they are completely world-class or genuinely unique—they're almost always highly designed, refined, and evolved over years to map to the specific end products and their one business. They are NOT designed, refined, or evolved to be reusable, retargetable capabilities for all interested parties.
By any rights, the major systems manufacturers of the 2000s—Cisco, Dell, HP, IBM, Sun Microsystems—should have had an enormous leg up in becoming the planetary scale, "red shift" cloud computing providers of today. Let's kick major operating environment vendors of that era in for good measure—Microsoft, Red Hat, VMware. Maybe even Oracle and SAP. They had the technical capabilities, the economies of scale, the vendor ecosystems, the customer relationships, the financial might. "Permission to win." Yet, only a few have made it. Microsoft has made the transition with Azure, but it's been a long hard pull. Maybe a few participation trophies for IBM and Oracle. But as a group, where are they, cloud-wise? Not too impressive that AWS and Google and others that had no classic "required assets" or "permission to win" have wiped them off the playing field or kicked them deep into the "others" bin.
You can call that classic innovator's dilemma, and there's a lot to that. It's insanely, unfathomably hard to disrupt your own current successful business for new delivery models and aspirations. But there's something else as well: services and products are different beasts. They take different mindsets, expectations, business models, and metrics. All the best to Intel in becoming that new thing, a service provider. But it not just a side gig or slight extension. It's a radically different thing, and it's no wonder the highly successful product Intel of 2011 wasn't also a good service provider.
From a technical perspective, App Engine and Compute Engine were built on top of internal infrastructure (borg), but did not expose borg directly. And there were a number of interesting mismatches between the semantics that customers expected of VMs and what borg offered to its containers that eventually resulted in dedicated borg clusters with different configs for cloud. And some retrospectives on whether building on borg was a better option than going bare metal directly.
Org-wise, the App Engine team was first and not part of the internal-focused Technical Infrastructure teams. GCS came next, and it too was not part of the canonical storage org. Then GCE, which was only possible because it was either written off or at least tolerated as an experiment by most, with a few key people providing behind-the-scenes support to make it happen -- especially in networking. It likely also helped that GAE was in SF and the rest of GCP in Seattle/Kirkland initially, so geo provided some insulation too.
The dominant perspective internally was that Google's technical infrastructure was its secret sauce, so why would they give it away to others? It took a long time to change that.
[Disclosure/source: I was on GCE and helped get it launched.]
I'll reply to myself since the parent comment is unalived.
I am not trying to generate a phylogenetic tree of cloud services, but show organizational shift that allowed Amazon (the whole company) to be successful by enforcing cell walls.
We would agree that Amazon is a services platform? Both in retail as well as cloud?
GCP itself is an application that rides on Borg, but Google itself does not use GCP. So it never got the recursive self improvement effect that Amazon did.
I killed my reply because it was repetitive with my other reply and not well thought out in hindsight.
But I will agree that Amazon is a services platform internally and externally.
But Amazon for the most part doesn’t run on AWS either and never at any point has Amazon Retail (CDO) and AWS shared any underlying architecture not even to the point that another commenter here who worked at Google described about GCP. Some of the new initiatives might. But AWS treats Amazon Retail as a customer and I heard ruminations that AWS Commercial Professional Services (the consulting division) even got involved with Amazon Retail projects or migrations.
I worked in the public sector division of ProServe (WWPS).
The Quora link reinforces OP's point though, it doesn't contradict it at all.
The only myth listed in your link is that Amazon made AWS to sell off excess computing capacity during periods of downtime.
This is from the CTO of Amazon himself:
>At Amazon we had developed unique software and services based on more than a decade of infrastructure work for the evolution of the Amazon E-Commerce Platform... The thinking then developed that offering Amazon’s expertise in ultra-scalable system software as primitive infrastructure building blocks delivered through a services interface could trigger whole new world of innovation as developers no longer needed to focus on buying, building and maintaining infrastructure.
Even today, if you look at the architecture and systems used by CDO and AWS, it’s clear that AWS was designed from the ground up to be sold to customers. The architecture used by Amazon Retail was not suitable for third party customers.
There is almost no similarity between AWS and CDO. I’m telling you this as someone who worked on the inside.
At most they took some of the things they learned. But it’s not like they took the code from any Amazon Retail implementation and did the equivalent of a “git checkout -b aws”
S3, SQS and EC2 the early AWS services were not based on the Amazon Retail systems at the time.
The entire control plane of CDO is completely different than AWS and always has been
If you want to see what services look like thst started out on Amazon Retail and moved over to AWS with few modifications - look no further than Amazon Connect. It was a lift and shift of the call center software that Amazon Retail uses without any AWS integrations or even publicly available APIs at first. It’s gotten better since 2020 when I was at AWS.
But none of those claims contradict the point that was made. It's not really clear what you're disputing.
AWS exists because Amazon had already developed the know-how and expertise in how to deliver computing infrastructure as a service and saw that offering this as a service to third-parties would be profitable. The only myth is that Amazon wanted to sell off excess computing.
There is a huge difference between “we have the know how to create AWS” and “we used spare capacity to sell to customers” or “we took code we used from Amazon Retail and created AWS”.
Most people think the former and even now people seem to think that there is any similarity between how Amazon Retail (CDO) and how AWS runs. Again with the caveat that some of Amazon Retail’s workloads have been migrated to AWS or were born on AWS.
Exactly, there is a huge difference, that's the point!
You seem to think the original claim is that someone took code from CDO and built AWS off of it, and no one said that. The simple claim which is reinforced in the very Quora link you provided is that Amazon had developed expertise in delivering computing infrastructure internally, and decided it would be profitable to offer this service externally.
Your claim that most people have a very specific belief about an implementation detail regarding whether literal code was reused or transferred is almost certainly false. The belief is that a company that developed expertise in an area that proved to be invaluable internally managed to leverage that expertise into a new product line that they could sell to other people, and nothing you have provided contradicts that.
AWS was obviously not going into building datacenters and taking a modern service oriented architecture approach cold. But, as you say, the pervasive myth is that AWS started out by using excess retail computing capacity and presumably the same architecture which, by all accounts, is simply untrue.
And GCP more or less started out as a classic PaaS. Azure really had more of an on-prem focus at first.
AWS started from a one pager that two employees in the Capetown office wrote (forget their names, it's been years) about the potential to sell compute as a service from the newly revamped Amazon.com website backend and servers.
Once the commercial viability was clear, Amazon.com created Amazon Web Services. The term 'offshoot' is perfectly reasonable.
This is from working at AWS circa 2013, which was just a few years after it started.
It's rare but does happen. And frankly I'd only include AWS in the counterpoint. Google really struggled with GCP. Outside of Bigquery and Spanner many/most of the services were custom built for GCP and were not used internally. Hell they built a VM service when basically everything ran on Borg internally
>TSMC founder says Tim Cook told him in 2011 that Intel did not know how to be a foundry
That is the Original Title. It could be edited as "Tim Cook told TSMC in 2011 that Intel did not know how to be a foundry" - That would still have been accurate with the date on.
Now the title has been editorialised, and meant or imply something else. Like most of the comments are already suggesting.
Less of a technical reason than the fact that TSMC was willing to go farther in accommodating Apple's demand (which is right). Also, this is Morris Chang relaying, so not totally unbiased.
The implication was that Intel lacked the customer-centric mindset required for a foundry business. Unlike TSMC, which tailors its process technologies to meet customer needs, Intel was used to designing and producing its own chips and struggled to adapt to servicing external clients. By contrast, Apple valued TSMC's ability to listen and respond to specific demands, something Intel historically did not do.
"Intel makes more than enough money that they don't have to cater any part of their design to external partners and historically have done very little of this type of work. Our use case was simply not on their radar."
On the other hand I can appreciate that Intel chips are more of a known quantity than the various ARM designs that are floating around out there.
This is back in 2011, when Intel was first flirting with the Foundry business, and has little bearing to their efforts today. That first effort mostly died. The goal wasn't to become a major player like Samsung/TSMC, but to ensure factories don't sit idle (i.e. just fill in the gaps with small customers). Intel products would get priority.
The current vision is very different: It's to somewhat separate the fabs and Intel's products, and the end goal is that Intel Products will just be another Foundry customer.
Arguably, Intel has already achieved the most important part of the Intel Foundry strategy: letting the CPU designers target whatever fab is best for their needs. Getting most of the consumer CPU moved to TSMC for Meteor Lake a year ago and the rest of it moved to TSMC last fall is hugely important. Intel Foundry cannot count on having a big customer, and the product line isn't forced to stick exclusively with failing fabs.
So, Intel's fault was being focused on selling you the whole widget and telling you what you wanted, instead of something that met your needs. Just like Apple, IMO - the whole "product vision" thing is great when it's about figuring out how to sell a good smartphone and terrible when it's selling iPads with desktop chips in them that can't do any useful professional work outside of drawing.
Given what we know about Apple and Cook, this seems like at least a modestly foolish story for Chang to have told.
Apple’s senior leadership has always for better or worse had long memories and held even longer grudges. And they don’t like partners to speak for them, or create the impression—even if accurate—that Apple has a dependency on a single supplier.
If Intel Foundry starting in 2011 had a non-trivial external customer base, Intel management would not have been so easily able to ignore the signs of trouble with their 10nm process and remain in denial. Their wake-up call would have been Apple jumping ship after a year or so, instead of Intel wasting several years re-releasing Skylake without even considering porting their newer CPU architecture to be built on the process that actually worked.
The first 10nm chip Intel actually got out the door (in tiny quantities) had the entire GPU disabled because their high-density transistor library was broken. That's exactly what Apple would have been using for most or all of an Intel-fabbed iPad chip, so Intel would have not been able to maintain the fiction that their 10nm process was mostly or even somewhat healthy.
I’ve seen that with startups. I’ve seen that with established very successful companies that try to sell to new customers.
You try to do something different, and every business process and management member is there to stop you because of what they learned previously … when working with a different customer or market.
The Steve Jobs story about resistance to developing a mouse in house is a good example.
I've seen it kill more than one company. The company is so bent out of shape for that one special customer, that as soon as the relationship is over or isn't as profitable as they'd hoped, they aren't able to actually support any other customers.
I wouldn't count on that. It's going to be some time before the Arizona plant is fully operational, and it's not a substitute for the production capacity in Taiwan. It's also very possible that TSMC will pause or halt production in the US if they're targeted by trade sanctions.
The TSMC Arizona plant is already operational they started production of TSMC 4nm process node this month, and will be ramping up volume over the next few months.
Even so, the ramp-up to full capacity is going to take a while - and even at full capacity it's not going to be a replacement for the facilities in Taiwan.
They don't have to replace all of the Taiwanese manufacturing, just the chips that are US bound. There are plenty of customers in other countries that need TSMC chips.
Besides the fact that the AZ plant does not produce TSMC's bleeding edge process (3nm) but the older 4nm, the product from that fab can't be delivered directly to a customer. Every wafer at that plant has to be shipped back to Taiwan for packaging and can only then be delivered to a customer.
So a blanket tariff would definitely apply even to the production of that plant. Which is why lots of CEO's (including Tim Cook) are trying to suck-up to Trump right now, to get an exception made for their imports.
Packaging is not trivial. I don't think either of those companies has been near the frontier for packaging in ... more than a decade? Two decades? This would be something where it would take a lot of time and throwing away badly packaged wafers to create the experience necessary. TSMC is probably better off doing it themselves in a new build versus working with partners who are so far from their level.
The point of the foreign trade zone is no tariffs, avoid customs. And the point of the tariff is to get plants built in the US, so tariffing them would make no sense.
> "When the customer asks a lot of things, we have learned to respond to every request," Chang said. "Some of them were crazy, some of them were irrational, [but] we respond to each request courteously. […] Intel has never done that, I knew a lot of customers of Intel's here in Taiwan and all [of them] wished that there were another supplier."
Companies just don't get it, that customer service is almost always the most important aspect of the company. The customer will put up with a lot of bullshit if communication is rock solid.
Handing off your customer service to agents that don't have reading comprehension, who don't have any authority, or who are completely non-understanding is going to hurt business.
And I'm not talking about customer service that can write flully emails thanking and apologizing and butt kissing, I'm talking about good customer service is when the agent understands the request, takes it seriously, runs it up the flag pole if needed, and can act on it.
I have to manage my personal urge to reject or overreact with customers all the time.
Customers do come up with a bonkers ideas / UI that's horrendous or just unworkable internal logic. And yet if I sit with them and we talk about "what can we do to solve the problem / enhance the product / get your goal done" I find they often come up with some good ideas and are happy to discard their initial madness.
You only need to look at the history of Apple with and without Jobs, with and without Cook, to see how replaceable Jobs was and how irreplaceable Cook has been.
Apple’s original crash happened while Jobs was in charge, which was why they fired him in the first place. Jobs’ return brought Cook, and since the second Jobs exit the company has done better than ever.
Cooks biggest blind spot is not playing nice with regulators and not being willing to give an inch. I think Jobs would have been worse. Could you actually see Jobs ever being a trophy for Trump and biting his tongue like Cook does?
Sure they do. They just don't have any capacity for producing the chip density that Apple demands. It's like showing up to a motorcycle factory and complaining that they don't know how to manufacture cars.
If America wanted EULV fabrication, it had to be organized and funded by the state. American fabs already made their business decision.
The article is talking about Intel's capabilities almost 15 years ago. EUVL is not part of the conversation. This was when Intel was bringing up their first FinFET process.
It's pretty widely known and documented that Intel at that time was in a horrible position to be a foundry for outside designers, especially ones that wanted to be able to design for more than one foundry.
> If America wanted EULV fabrication, it had to be organized and funded by the state.
The DoE funded initial research in to EUV via the national labs and EUV LLC back in the 90s. The IP was licenced to ASML, whilst Canon and Nikon (the leaders in lithography at the time) were blocked.
People in government in the 90s made a good call, and we haven't done anything since then to ensure it actually played out, until the post WW2 world order was already extremely unstable.
We have had multiple rounds of "why are we paying for any of this?" In our federal government since then.
> They just don't have any capacity for producing the chip density that Apple demands.
TFA says that the "Intel just does not know how to be a foundry" quote is from 2011 (14 years ago), and referred to Intel's lack of a "customer-centric mindset" rather than specific manufacturing capabilities.
(2011). And Cook was right. I worked at Intel for a few years during that decade and the foundry efforts were just not set up for success; in my area, they hired a bunch of new people, put up a firewall between us internal folks and the foundry folks, then without any guidance turned them loose. I was not even allowed to talk to them to troubleshoot equipment issues. They also got all of the equipment that we’d rejected for various reasons like poor process control, so they were newbies with worse equipment trying to start up a new group without help beyond what vendors would provide (for $$$)
I have no insight into the customer facing side
interesting perspective. Care to elaborate a bit more? Where you in the design department? Why did they put up chinese walls? Was it to enable the foundry to have other customers other than Intel designs? Why did you have similar type of equipment? Were you also manufacturing chipsets? If so, why didn't they expand your division to become a foundry?
There's a good video from Asianometry (Lessons from Intel's First Foundry): https://www.youtube.com/watch?v=-Y9LWYmVQu0
If you're an Intel foundry customer, you don't want your design or IP leaking across to Intel's product team, who might be a competitor.
But the IP is already "compiled", or if you want to isolate even further, separate just the mask making process. What could be learned from masks?
> What could be learned from masks
That's like asking what could you learn from building plans ...
I was in a fab module. The firewall part makes sense to not cross pollinate IP internally vs. externally, but it was taken to the extreme and management moved zero internal employees over to external so it was Intel’s tools and recipes but not the talent who knows all the tribal stuff.
> but not the talent who knows all the tribal stuff
Tribal knowledge doesn't appear on management org charts or in human resource titles.
Sounds like they were following some advice from “The Innovator’s Dilemma” by separating the business groups.
This wasn't something with the potential to become a disruptive innovation, as described in that book. Foundries are an established business model.
This situation has been and maybe always will be common. Those who view themselves as integrated manufacturers of *products* often cannot equally package, deliver, and sell their component *services*. No matter how good those services are at delivering the end product—and sometimes they are completely world-class or genuinely unique—they're almost always highly designed, refined, and evolved over years to map to the specific end products and their one business. They are NOT designed, refined, or evolved to be reusable, retargetable capabilities for all interested parties.
By any rights, the major systems manufacturers of the 2000s—Cisco, Dell, HP, IBM, Sun Microsystems—should have had an enormous leg up in becoming the planetary scale, "red shift" cloud computing providers of today. Let's kick major operating environment vendors of that era in for good measure—Microsoft, Red Hat, VMware. Maybe even Oracle and SAP. They had the technical capabilities, the economies of scale, the vendor ecosystems, the customer relationships, the financial might. "Permission to win." Yet, only a few have made it. Microsoft has made the transition with Azure, but it's been a long hard pull. Maybe a few participation trophies for IBM and Oracle. But as a group, where are they, cloud-wise? Not too impressive that AWS and Google and others that had no classic "required assets" or "permission to win" have wiped them off the playing field or kicked them deep into the "others" bin.
You can call that classic innovator's dilemma, and there's a lot to that. It's insanely, unfathomably hard to disrupt your own current successful business for new delivery models and aspirations. But there's something else as well: services and products are different beasts. They take different mindsets, expectations, business models, and metrics. All the best to Intel in becoming that new thing, a service provider. But it not just a side gig or slight extension. It's a radically different thing, and it's no wonder the highly successful product Intel of 2011 wasn't also a good service provider.
Counter-Point:
GCP and AWS came out of product focused companies that effectively converted to providing the services they used internally.
Also a myth for GCE.
From a technical perspective, App Engine and Compute Engine were built on top of internal infrastructure (borg), but did not expose borg directly. And there were a number of interesting mismatches between the semantics that customers expected of VMs and what borg offered to its containers that eventually resulted in dedicated borg clusters with different configs for cloud. And some retrospectives on whether building on borg was a better option than going bare metal directly.
Org-wise, the App Engine team was first and not part of the internal-focused Technical Infrastructure teams. GCS came next, and it too was not part of the canonical storage org. Then GCE, which was only possible because it was either written off or at least tolerated as an experiment by most, with a few key people providing behind-the-scenes support to make it happen -- especially in networking. It likely also helped that GAE was in SF and the rest of GCP in Seattle/Kirkland initially, so geo provided some insulation too.
The dominant perspective internally was that Google's technical infrastructure was its secret sauce, so why would they give it away to others? It took a long time to change that.
[Disclosure/source: I was on GCE and helped get it launched.]
This is not true as far as AWS and this is a myth that needs to die.
http://www.quora.com/How-and-why-did-Amazon-get-into-the-clo...
AWS was always purpose built and designed as a new product and Amazon Retail has a completely different architecture from AWS.
True, now some Amazon workloads have been moved over to AWS. But Amazon Retail is treated like an AWS customer.
Even internally, AWS employees use a different system to stand up internal sandboxes than CDO (Consumer Division Operations).
Confirm that the majority of Google first party stack is not sitting on GCP. They are parallel.
If Google had self hosted on GCP the way that Bezos dictated that Amazon have hard service boundaries, they would be the top Cloud provider right now.
But they didn't and they aren't, and it doesn't matter anymore.
https://www.phind.com/search?cache=fvljl85u7ll9u4wwncsrilfq
I'll reply to myself since the parent comment is unalived.
I am not trying to generate a phylogenetic tree of cloud services, but show organizational shift that allowed Amazon (the whole company) to be successful by enforcing cell walls.
We would agree that Amazon is a services platform? Both in retail as well as cloud?
GCP itself is an application that rides on Borg, but Google itself does not use GCP. So it never got the recursive self improvement effect that Amazon did.
I killed my reply because it was repetitive with my other reply and not well thought out in hindsight.
But I will agree that Amazon is a services platform internally and externally.
But Amazon for the most part doesn’t run on AWS either and never at any point has Amazon Retail (CDO) and AWS shared any underlying architecture not even to the point that another commenter here who worked at Google described about GCP. Some of the new initiatives might. But AWS treats Amazon Retail as a customer and I heard ruminations that AWS Commercial Professional Services (the consulting division) even got involved with Amazon Retail projects or migrations.
I worked in the public sector division of ProServe (WWPS).
Sounds like Amazon has it figured out way better than GCP and Google.
The Quora link reinforces OP's point though, it doesn't contradict it at all.
The only myth listed in your link is that Amazon made AWS to sell off excess computing capacity during periods of downtime.
This is from the CTO of Amazon himself:
>At Amazon we had developed unique software and services based on more than a decade of infrastructure work for the evolution of the Amazon E-Commerce Platform... The thinking then developed that offering Amazon’s expertise in ultra-scalable system software as primitive infrastructure building blocks delivered through a services interface could trigger whole new world of innovation as developers no longer needed to focus on buying, building and maintaining infrastructure.
Even today, if you look at the architecture and systems used by CDO and AWS, it’s clear that AWS was designed from the ground up to be sold to customers. The architecture used by Amazon Retail was not suitable for third party customers.
There is almost no similarity between AWS and CDO. I’m telling you this as someone who worked on the inside.
At most they took some of the things they learned. But it’s not like they took the code from any Amazon Retail implementation and did the equivalent of a “git checkout -b aws”
S3, SQS and EC2 the early AWS services were not based on the Amazon Retail systems at the time.
The entire control plane of CDO is completely different than AWS and always has been
If you want to see what services look like thst started out on Amazon Retail and moved over to AWS with few modifications - look no further than Amazon Connect. It was a lift and shift of the call center software that Amazon Retail uses without any AWS integrations or even publicly available APIs at first. It’s gotten better since 2020 when I was at AWS.
But none of those claims contradict the point that was made. It's not really clear what you're disputing.
AWS exists because Amazon had already developed the know-how and expertise in how to deliver computing infrastructure as a service and saw that offering this as a service to third-parties would be profitable. The only myth is that Amazon wanted to sell off excess computing.
There is a huge difference between “we have the know how to create AWS” and “we used spare capacity to sell to customers” or “we took code we used from Amazon Retail and created AWS”.
Most people think the former and even now people seem to think that there is any similarity between how Amazon Retail (CDO) and how AWS runs. Again with the caveat that some of Amazon Retail’s workloads have been migrated to AWS or were born on AWS.
Exactly, there is a huge difference, that's the point!
You seem to think the original claim is that someone took code from CDO and built AWS off of it, and no one said that. The simple claim which is reinforced in the very Quora link you provided is that Amazon had developed expertise in delivering computing infrastructure internally, and decided it would be profitable to offer this service externally.
Your claim that most people have a very specific belief about an implementation detail regarding whether literal code was reused or transferred is almost certainly false. The belief is that a company that developed expertise in an area that proved to be invaluable internally managed to leverage that expertise into a new product line that they could sell to other people, and nothing you have provided contradicts that.
This is the original message
“GCP and AWS came out of product focused companies that effectively converted to providing the services they used internally.”
They at most used thier know how. You and I are in agreement. But that’s not what is implied by the parent.
AWS was obviously not going into building datacenters and taking a modern service oriented architecture approach cold. But, as you say, the pervasive myth is that AWS started out by using excess retail computing capacity and presumably the same architecture which, by all accounts, is simply untrue.
And GCP more or less started out as a classic PaaS. Azure really had more of an on-prem focus at first.
I'd never really thought about "How did hyperscale cloud providers become cloud providers?" And it's surprisingly distinct.
AWS: Bezos API memo -> internal infra services that were external ready -> polishing and exposing those one-by-one
GCP: Internal architectural/technical excellence -> new org that attempted to productize v2.0 of those services
Azure: Believe it was mostly ground-up build as a new org/product?
(No idea where Oracle cloud came from, internally)
AWS was never an offshoot from Amazon.
http://www.quora.com/How-and-why-did-Amazon-get-into-the-clo...
That's not true.
AWS started from a one pager that two employees in the Capetown office wrote (forget their names, it's been years) about the potential to sell compute as a service from the newly revamped Amazon.com website backend and servers.
Once the commercial viability was clear, Amazon.com created Amazon Web Services. The term 'offshoot' is perfectly reasonable.
This is from working at AWS circa 2013, which was just a few years after it started.
Again the link I posted was from Werner Vogels. The architecture that the CDO (Amazon Retail) uses is completely different than AWS.
Have you seen the difference between the architecture of CDO and AWS internally?
It's rare but does happen. And frankly I'd only include AWS in the counterpoint. Google really struggled with GCP. Outside of Bigquery and Spanner many/most of the services were custom built for GCP and were not used internally. Hell they built a VM service when basically everything ran on Borg internally
>TSMC founder says Tim Cook told him in 2011 that Intel did not know how to be a foundry
That is the Original Title. It could be edited as "Tim Cook told TSMC in 2011 that Intel did not know how to be a foundry" - That would still have been accurate with the date on.
Now the title has been editorialised, and meant or imply something else. Like most of the comments are already suggesting.
If you haven’t listened to it already, Acquired remastered their TSMC episode. You can find that here https://www.acquired.fm/episodes/tsmc-remastered
If you are interested in getting more context on TSMC, this is a great place to start.
And to be clear, they also flew to Taiwan and interviewed Morris in English for what appears to be the first time in over 15 years:
https://www.acquired.fm/episodes/tsmc-founder-morris-chang
A tiny bit of this (excellent!) interview led to the story we're all commenting on.
Less of a technical reason than the fact that TSMC was willing to go farther in accommodating Apple's demand (which is right). Also, this is Morris Chang relaying, so not totally unbiased.
"Intel makes more than enough money that they don't have to cater any part of their design to external partners and historically have done very little of this type of work. Our use case was simply not on their radar."
On the other hand I can appreciate that Intel chips are more of a known quantity than the various ARM designs that are floating around out there.
In 2011. Seems like that should be in the title.
This is back in 2011, when Intel was first flirting with the Foundry business, and has little bearing to their efforts today. That first effort mostly died. The goal wasn't to become a major player like Samsung/TSMC, but to ensure factories don't sit idle (i.e. just fill in the gaps with small customers). Intel products would get priority.
The current vision is very different: It's to somewhat separate the fabs and Intel's products, and the end goal is that Intel Products will just be another Foundry customer.
Whether they can achieve this is another story.
Arguably, Intel has already achieved the most important part of the Intel Foundry strategy: letting the CPU designers target whatever fab is best for their needs. Getting most of the consumer CPU moved to TSMC for Meteor Lake a year ago and the rest of it moved to TSMC last fall is hugely important. Intel Foundry cannot count on having a big customer, and the product line isn't forced to stick exclusively with failing fabs.
So, Intel's fault was being focused on selling you the whole widget and telling you what you wanted, instead of something that met your needs. Just like Apple, IMO - the whole "product vision" thing is great when it's about figuring out how to sell a good smartphone and terrible when it's selling iPads with desktop chips in them that can't do any useful professional work outside of drawing.
Given what we know about Apple and Cook, this seems like at least a modestly foolish story for Chang to have told.
Apple’s senior leadership has always for better or worse had long memories and held even longer grudges. And they don’t like partners to speak for them, or create the impression—even if accurate—that Apple has a dependency on a single supplier.
Chang is 93 years old. He can say whatever he wants.
Second this. I don't think most people on hn understand how powerful he is.
That's true when Apple has alternative suppliers, ie graphic cards when nVidia blabs before a presentation.
But who is going to replace TSMC? Sure someone will in 15-20 years, but for now...
they have used Samsung before...maybe with middling results but if push came to shove, Samsung's fabs have been the backup I think.
You could've said the same about Intel at one point, but here we are years later.
2011 article, so GP's timeline fits.
It's not a 2011 article. The alleged statement by Tim was made in 2011.
What is Apple going to do? Not use TSMC?
Clearly the next step after building your own CPU and SOC is to start Apple Foundry and become totally vertically integrated?
Just think where Intel would be today if they had truly leaned in and made the necessary changes to supply the iPhone chips.
Tim Cook would have negotiated away their profit margins?
And how is that different than what is happening now?
https://www.intc.com/news-events/press-releases/detail/1716/...
> Third-quarter GAAP earnings (loss) per share (EPS) attributable to Intel was $(3.88); non-GAAP EPS attributable to Intel was $(0.46).
Yeah, like they've done to TSMC over the past decade. (/s)
Wouldn't they still be behind on the chip technology curve? Seems like this would hurt Apple, not dramatically help Intel.
If Intel Foundry starting in 2011 had a non-trivial external customer base, Intel management would not have been so easily able to ignore the signs of trouble with their 10nm process and remain in denial. Their wake-up call would have been Apple jumping ship after a year or so, instead of Intel wasting several years re-releasing Skylake without even considering porting their newer CPU architecture to be built on the process that actually worked.
The first 10nm chip Intel actually got out the door (in tiny quantities) had the entire GPU disabled because their high-density transistor library was broken. That's exactly what Apple would have been using for most or all of an Intel-fabbed iPad chip, so Intel would have not been able to maintain the fiction that their 10nm process was mostly or even somewhat healthy.
Well, he's not wrong
Subdivisions that only work for one customer grow around all their idiosyncrasies and it's hard to adapt later
Actually putting the design work into something you can work with is hard work
I’ve seen that with startups. I’ve seen that with established very successful companies that try to sell to new customers.
You try to do something different, and every business process and management member is there to stop you because of what they learned previously … when working with a different customer or market.
The Steve Jobs story about resistance to developing a mouse in house is a good example.
I've seen it kill more than one company. The company is so bent out of shape for that one special customer, that as soon as the relationship is over or isn't as profitable as they'd hoped, they aren't able to actually support any other customers.
What's the Steve Jobs mouse story?
This is the interview I'm thinking of:
https://www.youtube.com/watch?v=s4Cz49MLh4o
The general topic applies to what I mentioned but even just the specific mouse story is at this point in the video:
https://youtu.be/s4Cz49MLh4o?t=90
Folks who operate in a certain context / process or such and just can't imagine doing any differently and will stop you.
And just a note, Intel is losing money and still as of the third quarter of last year paying a dividend while complaining about cash flow
Intel suspended the dividend.
Thanks to neoliberal economic theory, the public CEOs must deliver shareholder value at all costs! In this lens, they are doing “great”
But actual company itself and products are bleeding internally from decades of mismanagement.
Particularly relevant with the 100% tariffs that Trump is going to try to ram through:
https://www.tomshardware.com/tech-industry/semiconductors/ta...
They’re already running a plant in Arizona, and I heard it’s going really well. It won’t matter.
Especially now that Taiwan has removed restrictions on producing TSMC's most advanced nodes on foreign soil.
https://www.techpowerup.com/331104/tsmc-granted-government-p...
I wouldn't count on that. It's going to be some time before the Arizona plant is fully operational, and it's not a substitute for the production capacity in Taiwan. It's also very possible that TSMC will pause or halt production in the US if they're targeted by trade sanctions.
The TSMC Arizona plant is already operational they started production of TSMC 4nm process node this month, and will be ramping up volume over the next few months.
https://www.reuters.com/technology/tsmc-begins-producing-4-n...
TSMC is also looking to add a chip packaging plant in Arizona.
https://pr.tsmc.com/english/news/3174
Amkor will also be near
Even so, the ramp-up to full capacity is going to take a while - and even at full capacity it's not going to be a replacement for the facilities in Taiwan.
They don't have to replace all of the Taiwanese manufacturing, just the chips that are US bound. There are plenty of customers in other countries that need TSMC chips.
Besides the fact that the AZ plant does not produce TSMC's bleeding edge process (3nm) but the older 4nm, the product from that fab can't be delivered directly to a customer. Every wafer at that plant has to be shipped back to Taiwan for packaging and can only then be delivered to a customer.
So a blanket tariff would definitely apply even to the production of that plant. Which is why lots of CEO's (including Tim Cook) are trying to suck-up to Trump right now, to get an exception made for their imports.
TSMC can work with partners for packaging (TI and IBM come to mind)
Packaging is not trivial. I don't think either of those companies has been near the frontier for packaging in ... more than a decade? Two decades? This would be something where it would take a lot of time and throwing away badly packaged wafers to create the experience necessary. TSMC is probably better off doing it themselves in a new build versus working with partners who are so far from their level.
That is in an international economic zone and would be subject to tariffs as well as for supplies.
The point of the foreign trade zone is no tariffs, avoid customs. And the point of the tariff is to get plants built in the US, so tariffing them would make no sense.
The plant in Arizona is deliberately not equipped to produce current-gen chips.
Did you forget the sarcasm flag?
Cook knows how to kiss up to Trump. There is a reason Apple wasn’t affected by China tariffs the first time .
> "When the customer asks a lot of things, we have learned to respond to every request," Chang said. "Some of them were crazy, some of them were irrational, [but] we respond to each request courteously. […] Intel has never done that, I knew a lot of customers of Intel's here in Taiwan and all [of them] wished that there were another supplier."
Companies just don't get it, that customer service is almost always the most important aspect of the company. The customer will put up with a lot of bullshit if communication is rock solid.
Handing off your customer service to agents that don't have reading comprehension, who don't have any authority, or who are completely non-understanding is going to hurt business.
And I'm not talking about customer service that can write flully emails thanking and apologizing and butt kissing, I'm talking about good customer service is when the agent understands the request, takes it seriously, runs it up the flag pole if needed, and can act on it.
I have to manage my personal urge to reject or overreact with customers all the time.
Customers do come up with a bonkers ideas / UI that's horrendous or just unworkable internal logic. And yet if I sit with them and we talk about "what can we do to solve the problem / enhance the product / get your goal done" I find they often come up with some good ideas and are happy to discard their initial madness.
That's "Beans and Noses": https://archive.uie.com/brainsparks/2011/07/08/beans-and-nos...
That was great, thanks.
[flagged]
You only need to look at the history of Apple with and without Jobs, with and without Cook, to see how replaceable Jobs was and how irreplaceable Cook has been.
Apple’s original crash happened while Jobs was in charge, which was why they fired him in the first place. Jobs’ return brought Cook, and since the second Jobs exit the company has done better than ever.
Cooks biggest blind spot is not playing nice with regulators and not being willing to give an inch. I think Jobs would have been worse. Could you actually see Jobs ever being a trophy for Trump and biting his tongue like Cook does?
Steve reportedly told Tim to be himself and not try to be Steve.
Sure they do. They just don't have any capacity for producing the chip density that Apple demands. It's like showing up to a motorcycle factory and complaining that they don't know how to manufacture cars.
If America wanted EULV fabrication, it had to be organized and funded by the state. American fabs already made their business decision.
The article is talking about Intel's capabilities almost 15 years ago. EUVL is not part of the conversation. This was when Intel was bringing up their first FinFET process.
It's pretty widely known and documented that Intel at that time was in a horrible position to be a foundry for outside designers, especially ones that wanted to be able to design for more than one foundry.
> If America wanted EULV fabrication, it had to be organized and funded by the state.
The DoE funded initial research in to EUV via the national labs and EUV LLC back in the 90s. The IP was licenced to ASML, whilst Canon and Nikon (the leaders in lithography at the time) were blocked.
People in government in the 90s made a good call, and we haven't done anything since then to ensure it actually played out, until the post WW2 world order was already extremely unstable.
We have had multiple rounds of "why are we paying for any of this?" In our federal government since then.
> They just don't have any capacity for producing the chip density that Apple demands.
TFA says that the "Intel just does not know how to be a foundry" quote is from 2011 (14 years ago), and referred to Intel's lack of a "customer-centric mindset" rather than specific manufacturing capabilities.