TSMC Threat to Intel in Everything but Memory

[Arthur Hanson]

TSMC with its combination of deep experience with ARM processors and InFO packaging it is going to become a serious threat to Intel's core processor business. With Microsoft going to supporting ARM processors and HPC already using ARM processors in the data center, TSMC is now in the position to go after Intel's core business. Since they have recently passed Intel in market cap, they are now in a position to be the dominant player in high end chips for both fixed and mobile applications. Combine this with their moves into MEMS combined with other technologies where InFO packaging gives them an additional edge and this puts TSMC in the position to even more dominate the foundry market and even pass Samsung in chip production. It will be interesting to see how this plays out over the next year or two. It looks like the Wintel alliance has seen better days. It looks like Intel may again revert to its original strength in memory with 3DXpoint, which it shares with Micron. Full disclosure, I hold interests in MU and TSM.

 

[Fred Chen]

Intel's advantage in memory is probably more well-grounded in the 3D NAND rather than 3D XPoint. The 3D XPoint has been confirmed to be phase change memory with an ovonic threshold selector: Intel 3D XPoint Memory Die Removed from Intel Optane™ PCM (Phase Change Memory). These materials are not that easy to work with.

 

[Daniel Nenni]

Personally I think Intel is Intel's biggest threat. They keep missing markets and spending billions of dollars playing catch-up only to give up. Mobile, Wearables, FPGAs, Foundry, etc... Fortunately for Intel, AMD or ARM or anyone else has not made progress in the CPU business but that day will come, absolutely, especially since the foundries now have the process advantage. Just my opinion of course but I would not want to be an Intel executive hoping to end their career on a high note.

 

[Daniel Payne]

Arthur,

Intel and TSMC both manufacture ICs, so yes, they are comparable in one sense, however TSMC will fabricate any customer design. In fact, TSMC even fabricates chips for Intel. TSMC does not design their own chips, or sell their own silicon as a standard part on the open market.

Intel is an Integrated Device Manufacturer (IDM) and own their own fab, which may seem to be a subtle distinction. Intel sells its chips to customers that then build systems. Yes, Intel also has a foundry business that competes with TSMC's foundry business.

TSMC wants to fill up their fab with customer-designed parts.

Intel wants to fill up their fab with Intel-designed parts.

TSMC is not in any sense of the word a product competitor to Intel, because TSMC doesn't design or sell an IC as a product.

TSMC doesn't really care which customers bring a core to them for fabrication, they are architecture-agnostic, they simply want to fabricate parts for their customers. TSMC is not aiming at Intel's data center, they are just fabricating ICs that their customers bring to them. If an Intel competitor chooses to fabricate at TSMC, then that is not equivalent to saying that TSMC is competing with Intel. TSMC is serving their customer, and that customer is competing with Intel.

 

[Arthur Hanson]

Dan P. I think you missed the point, TSM enables the competition to Intel by giving them access to a world beating fab. Yes, technically TSM is not beating Intel, they are the enabler that allows EVVERYONE ELSE to attack Intel's various markets. In the process of doing this, TSM will concentrate with laser focus on the process. TSM is the threat, for if they didn't exist, it would be much harder to attack Intel's market. This process is already well under way and will continue to turn up the heat on Intel. Instead of one competitor, TSM enables many, many competitors on every front. I doubt they will become a fab for memory, but you never know.

 

[smeyer0028]

I worked at Zilog (wrote PL/Z machine code generator). Intel started as a memory
manufacturer. At the height of Z80 popularity, Intel was considering going back to
memory. Maybe now they really will become a memory manufacturer again.

 

[Daniel Payne]

smeyer,
Intel never liked the boom-bust cycles that define the volatile memory market, however if they acquired Micron then it could make sense for Intel to re-enter the memory market in a bigger way.

 

[hist78]

TSMC may not want to directly go into commodity memory manufacturing business. But they might be forming a partnership with Hon Hai/Foxconn and Apple to take over Toshiba's memory division. Hon Hai can take care the operations, Apple provides steady orders and financial backing as needed, and TSMC provides advance memory and manufacturing technologies. This is purely my own speculation.

The benefit is Apple can secure their long term memory supply at the same time Hon Hai and TSMC secure their long term relationship with Apple. Furthermore, they are making their common enemy Samsung's life a little bit harder.

 

[Fred Chen]

TSMC may not want to directly go into commodity memory manufacturing business. But they might be forming a partnership with Hon Hai/Foxconn and Apple to take over Toshiba's memory division. Hon Hai can take care the operations, Apple provides steady orders and financial backing as needed, and TSMC provides advance memory and manufacturing technologies. This is purely my own speculation.

The benefit is Apple can secure their long term memory supply at the same time Hon Hai and TSMC secure their long term relationship with Apple. Furthermore, they are making their common enemy Samsung's life a little bit harder.

TSMC dropped out of the bidding contest for Toshiba's NAND Flash. Customers probably find it cheaper to get high-density memory from current suppliers rather than ask TSMC to make it.

 

[Fred Chen]

The following should be considered: Poor old Intel. No flash profits for you • The Register

Intel NVM Solutions Group operating losses;

• • First quarter 2017 - $129m
  • Fourth quarter 2016 -$91m
  • First quarter 2016 - $95m

"Losses were driven by the company’s ongoing investments in 3D NAND and its 3D XPoint development." "However, the company expects to run at breakeven for the core business in 2H2017 and be profitable for the entire non-volatile memory business near the end of 2018."

 

[hist78]

TSMC dropped out of the bidding contest for Toshiba's NAND Flash. Customers probably find it cheaper to get high-density memory from current suppliers rather than ask TSMC to make it.

TSMC dropped out the bidding competition doesn't mean they dropped out the game. Be a bidder in the first round of process can provide them a lot of Toshiba's confidential information that later can be very useful for them in deciding some kind of partnership with Hon Hai, Apple, Amazon (from news today), and Western Digital.

 

[Arthur Hanson]

Fred, anytime a company is working a new breakthrough technology such as 3dXpoint, they are going to loose money and sometimes a lot of it. It won't be until production is ramped up and the bugs worked out that they will make serious money and sometimes if things don't quite work out, loose serious money. From what I have read and little has been released, I think they are in the process of getting yields up.

 

[benb]

TSMC, Intel and Samsung are now so big their growth is not much different than global growth. This is one headwind TSMC faces--they can just kill it at about everything, and still face the prospects of slow global growth. I think it makes it hard to analyze the situation objectively, because they are certainly killing it, today, but could still be in real trouble soon, as the economic cycle ages and eventually turns down. They are so tied to broader economic factors.

This is true of all manufacturing--very leveraged on the upside, very leveraged on the downside. Things are good today, so I won't fault TSMC. In 5 years, these are some structural problems that will undoubtedly tug their balloon back to earth:

1) No one but Apple and Samsung makes money in smartphones. Eventually, this is going to cause a serious crash.
2) It is getting more and more unreasonable to manufacture logic devices, in the quantities that customers need, in the timeframes needed, with the yields needed. Today, those 3 things can be mostly all accommodated, but in 5 years, I think customers will be forced to pick one or two of the 3.
3) How do you avoid a peak-smartphone plateau, and fall-off? What fills the void?

 

[hist78]

TSMC, Intel and Samsung are now so big their growth is not much different than global growth. This is one headwind TSMC faces--they can just kill it at about everything, and still face the prospects of slow global growth. I think it makes it hard to analyze the situation objectively, because they are certainly killing it, today, but could still be in real trouble soon, as the economic cycle ages and eventually turns down. They are so tied to broader economic factors.

This is true of all manufacturing--very leveraged on the upside, very leveraged on the downside. Things are good today, so I won't fault TSMC. In 5 years, these are some structural problems that will undoubtedly tug their balloon back to earth:

1) No one but Apple and Samsung makes money in smartphones. Eventually, this is going to cause a serious crash.
2) It is getting more and more unreasonable to manufacture logic devices, in the quantities that customers need, in the timeframes needed, with the yields needed. Today, those 3 things can be mostly all accommodated, but in 5 years, I think customers will be forced to pick one or two of the 3.
3) How do you avoid a peak-smartphone plateau, and fall-off? What fills the void?

There are some additional considerations I'd like to suggest:

1. Smartphones need to be replaced every several years, depending on the quality and usage of the phone, country, income, technologies advancement, and carriers' requirement. It's not like a flat screen TV can sit there and being used for ten years. We can have a group of people sharing a TV or a PC system. But we can't have a group of people share a smartphone. So the annual smartphone growth will slow down but it might not have significant drop of total volume.

2. Each smartphone will continuously built with more semiconductor content due to the technology improvement and new features. It will create more demand on the semi industry.

3. Beyond smartphones, what will be the driving force on the semi industry for the manufacturing capacity and technology? AI, automobiles, IoT, High Performance Computing, automation, or healthcare and medical? If one or two of them create new demand on the semi, the volume and value will be big, or even bigger than the smartphones have achieved.

 

[count]

3) How do you avoid a peak-smartphone plateau, and fall-off? What fills the void?

The void will be filled by compute required to support tomorrows future blockchain applications. There will be blockchains for almost every type of transaction imaginable, and they will require a significant amount of compute.

 

[Arthur Hanson]

benb, TSM, Samsung and others are in a market that's not only expanding, but rapidly getting broader as these are the companies that have the best hold on nanotechnology, which is just in it's infancy. Transportation, MEMs, Solar, Display, Permeable actively controlled barriers (think active filters, desalinization, chemical separation, batteries, super capacitors, etc.. These companies have a very long and broad expansion path.

 

[Arthur Hanson]

Fred, you may be right for it seems INTEL may be having serious development problems with its 3dXpoint technology as indicated by being behind stated delivery dates. If this become a serious problem, Intel my be in real trouble.

 

[Arthur Hanson]

Now it looks like even the memory part of Intel might have problems with 3dXpoint. This is something Intel/Micron could easily clear up, whether they have hit just a speed bump or a major stumbling block.