TSMC to make advanced 2-nm chips in US sooner to meet AI demand

[samwilde]

You can't compare units directly, because the cost of a mobile chip is quite a bit less than for a CPU. An A-18 chip is reported to sell for $45. An A-20 chip is reported to sell for $280. An Intel I9 14900KS lists for $730. From what I have heard that was why Otellini walked away from the iPhone deal.

Ottellini was just dumb. Minicomputers were also cheaper per unit than mainframes, and microcomputers were cheaper than minicomputers.

As the price and form factor goes down, the volume goes up. A lot. I think the smartphone market moves like 4x more units than the PC market. And that is if you ignore tablets.

As time passes the smaller solution also usually grows upwards and threatens at least the bottom end of the bigger solution market. So you now have ARM on laptops and beyond.

Lip-Bu Tan seems to think that inference and agentic AI are going to be much bigger than the training market that Nvidia is currently dominating and is taking steps to get into that market. You'll have to forgive me for assuming he knows what he is doing. His track record is hard to argue with.

He is right than inference workloads will dominate training. But I suspect the ideal inference architecture will definitively not be a GPU.

 

[siliconbruh999]

He is right than inference workloads will dominate training. But I suspect the ideal inference architecture will definitively not be a GPU.

It may not be but if you want to sell your hardware to public it needs to be programmable easily which Gaudi wasn't

 

[Xebec]

If we could turn back the clock and make Intel’s 10nm process deliver on time, assuming everything else remained the same, would Intel have great products, fast growing revenue, and strong profits today?

My guess is no. It might have only delayed the troubles Intel faces today by two or three years, but fundamentally, Intel was already on the wrong track.

I agree, though I think it pushes back the clock more like 7-10 years (3-4 full nodes).

The 10nm debacle allowed AMD to take nearly half of the x86 server market share, and also weakened the x86 market overall vs competitors in the high margin server space. Intel also was rumored to have had a large customer at one point lined up for Intel 10nm, and the rumor was the customer lost $Bs in sales because of what happened. i.e. 10nm effectively killed Intel's foundry ambitions, too..

Though the real shot in the foot that Intel is dealing with right now goes back to the mobile market (2007?). If Intel had captured the iPhone, TSMC would be slower on nodes post 16nm, and maybe crawling around 7/6nm until the AI market we're seeing today.

It's interesting that Intel evolved INTO the PC business (from memory), but has yet to be successful evolving past the PC business, as they've been trying for decades.

 

[Artificer60]

Ottellini was just dumb.

It would seem I failed to make my point about the economics of the iPhone deal. Intel's process was not cost effective for building a low margin part. Everyone I know who has knowledge about the deal says Intel would have lost money. How does that build a stronger Intel?

Now you can make an argument that seeing there was an emerging market that Intel couldn't afford to play in should have been a wake up call. Andy Grove may have seen the winds changing and seized the opportunity, but how many Andy Grove's are out there? If that is your standard to not be "dumb" then very few CEO's aren't idiots.

People always point to the iPhone miss when they talk about Otellini. They never remember Centrino, which made modern laptops what they are today. I don't think the laptop market ever would have become what it is today (and fueled Intel's profits for over a decade) without Centrino. And Centrino was Otellini's baby, so I think generically dismissing the man as dumb is a bit much.

 

[siliconbruh999]

It would seem I failed to make my point about the economics of the iPhone deal. Intel's process was not cost effective for building a low margin part. Everyone I know who has knowledge about the deal says Intel would have lost money. How does that build a stronger Intel?

Now you can make an argument that seeing there was an emerging market that Intel couldn't afford to play in should have been a wake up call. Andy Grove may have seen the winds changing and seized the opportunity, but how many Andy Grove's are out there? If that is your standard to not be "dumb" then very few CEO's aren't idiots.

People always point to the iPhone miss when they talk about Otellini. They never remember Centrino, which made modern laptops what they are today. I don't think the laptop market ever would have become what it is today (and fueled Intel's profits for over a decade) without Centrino. And Centrino was Otellini's baby, so I think generically dismissing the man as dumb is a bit much.

Otelnni later admitted that he was wrong

 

[MKWVentures]

It would seem I failed to make my point about the economics of the iPhone deal. Intel's process was not cost effective for building a low margin part. Everyone I know who has knowledge about the deal says Intel would have lost money. How does that build a stronger Intel?

Now you can make an argument that seeing there was an emerging market that Intel couldn't afford to play in should have been a wake up call. Andy Grove may have seen the winds changing and seized the opportunity, but how many Andy Grove's are out there? If that is your standard to not be "dumb" then very few CEO's aren't idiots.

People always point to the iPhone miss when they talk about Otellini. They never remember Centrino, which made modern laptops what they are today. I don't think the laptop market ever would have become what it is today (and fueled Intel's profits for over a decade) without Centrino. And Centrino was Otellini's baby, so I think generically dismissing the man as dumb is a bit much.

I have said this before. Intel was correct to walk away from Iphone deal. The numbers were horrible. Trust me, I was there.

Intel had the leading phone AP at one time and pretty high volume. they lost billions. Intel cannot do Mobile chips for a variety of reasons.... some of which I was responsible for LOL

 

[siliconbruh999]

Intel had the leading phone AP at one time and pretty high volume. they lost billions. Intel cannot do Mobile chips for a variety of reasons.... some of which I was responsible for LOL

do you think there was 0 market for Intel in Mobile?

 

[MKWVentures]

do you think there was 0 market for Intel in Mobile?

Intel cannot make money on high volume phone chips. Its a different market that is opposite of how Intel does business and opposite of how Intel wafer costs are.

 

[siliconbruh999]

Intel cannot make money on high volume phone chips. Its a different market that is opposite of how Intel does business and opposite of how Intel wafer costs are.

So they had to change their entire cost structure across product and foundry to have a foothold in Mobile?
That's a tall order.

 

[MKWVentures]

So they had to change their entire cost structure across product and foundry to have a foothold in Mobile?
That's a tall order.

Lets say THEORETICALLY that Intel managers calculated a unit cost that was ~1.5x the price Apple wanted to pay (best case). Then those managers were told to create a business model and theoretical number set to get the cost below the price. Most people thought it was impossible.

After working with Micron, we realized that Micron could do it easily. Their entire operational model is different and it leads to dramatically lower cost. Obviously TSMC is the same. There are some basic reasons why (its not volume in this case)

Otellini's reason for change of heart was that he felt that it would have driven Intel to be a cost effective fab company..... Not that it would have made money. But that would have been like teaching Shaq to play Baseball. its possible..... not likely to lead to success.

 

[hist78]

You can't compare units directly, because the cost of a mobile chip is quite a bit less than for a CPU. An A-18 chip is reported to sell for $45. An A-20 chip is reported to sell for $280. An Intel I9 14900KS lists for $730. From what I have heard that was why Otellini walked away from the iPhone deal. Intel's process was optimized for processor speed and producing chips on the process was far more expensive than the corresponding foundry process. At the time Intel drove the industry and cost was not an object.

Intel offers both high priced processors and low-end, inexpensive ones. However, we can’t accurately analyze Intel’s true pricing structure because the company doesn’t disclose its processor ASP (average selling price) and shipment volume. Still, by looking at the sharply contrasting volume trends, explosive growth in smartphones versus a declining PC market, we can see that the PC business has been weak for years. Unfortunately, it remains Intel’s most important revenue source.

Meanwhile, companies like Qualcomm, MediaTek, Apple, TSMC, Samsung, and other smaller players have been making strong profits from smartphone related products. In contrast, Intel’s OEM partners, HP, Dell, Lenovo, ASUS, and Acer, have struggled for years with net profit margins stuck in the low single digits. It’s also worth noting that many smartphones today are priced higher than many PCs available in stores or online. Consumers upgrade their smartphones far more frequently than their PCs, underscoring that the PC market isn’t a great place for Intel to pursue major growth opportunities.

Recently, I spoke with a group of five 5th-grade students. At school, they all have access to PCs provided by their school. But at home, only one of them has both a PC and a smartphone; the other four rely solely on smartphones and tablets. They told me they don’t even have a PC at home.

Intel certainly has manufacturing challenges, but in my opinion, its biggest problem is that it doesn’t have the right products to sell.

 

[hist78]

Otellini's reason for change of heart was that he felt that it would have driven Intel to be a cost effective fab company..... Not that it would have made money. But that would have been like teaching Shaq to play Baseball. its possible..... not likely to lead to success.

But Paul Otellini and many Intel leaders should have realized that if Intel didn’t change, its partners and customers eventually would. They would pursue new markets and new sources of profit with or without Intel.

 

[floppydisk]

Jensen claims that inference is going to 1,000,000,000x for whatever that's worth.

Couple things LBT said in the earnings call:

LBT: Finally, on our AI accelerator strategy, I continue to believe that we can play a meaningful role in developing compute platforms for emerging inference workloads driven by agentic AI and physical AI. This will be a far larger market than that for AI training workloads. We will work to position Intel as a compute platform of choice for AI inference, and we look to partner with a range of incumbents, as well as emerging companies that are defining this new compute paradigm. This is a multi-year initiative, and we will strike a partnership when we can deliver true differentiation and market-leading products. In the near term, we will continue delivering AI capabilities through Xeon AI PCs, Arc GPUs, and our open software stack. Looking ahead, we plan to launch successive generations of inference-optimized GPUs on the annual cadence that features enhanced memory and bandwidth to meet enterprise needs.

LBT: In terms of the AI strategy, we clearly are defining the, you know, Crescent Island that we talked about. We also have a new product line in the lineup. They're addressing the agentic and the physical AI and more in the inference side. I think stay tuned. We will update that.

 

[hist78]

I have said this before. Intel was correct to walk away from Iphone deal. The numbers were horrible. Trust me, I was there.

Intel had the leading phone AP at one time and pretty high volume. they lost billions. Intel cannot do Mobile chips for a variety of reasons.... some of which I was responsible for LOL

Because you were there, do you know why Intel didn’t cut its costs to make them feasible or equally profitable?