Mark Liu, AI chips to Displace server chips

[Daniel Nenni]

I could write a 200 page book on things Intel does that TSMC doesn't. OK, maybe not that big.

If you're a programmer, what compilers do you use? Hmmmm, Intel always perform the best, don't they? Even for AMD processors they normally do. How about operating systems? Who makes the fastest Linux? Yeah, Intel. Intel designs firmware. TSMC? I doubt it. And I'm just touching on software, there are so many other things Intel does that TSMC does not in software.

Hardware, does TSMC have the first idea how to design a high performance processor, or GPU? How about car driving software and hardware? I could go on and on, but on balance, I'd have to say Intel technology is among the best in the world, whereas TSMC is very good at a market segment, but simply doesn't have the breadth of technology companies like Intel have.

Does that make them worse? Of course not. I just object to overreaching statements like they have better technology. It's such a strange and difficult comparison, I don't know how you could possibly judge it. Based on finances? Please. Especially when Intel typically makes much more money than TSMC. It's a mistake to take a snapshot in time and expand it to become an almost universal or broad truth.

I'm not taking a slap at TSMC, just broad comparisons that are really difficult to make because they are so different.

There is a very big difference between an IDM (Intel) and a Pure-play foundry (TSMC) . A book on it would be more than 200 pages. Same can be said with the difference between a fabless company and an IDM or a fabless company and a pure-lay foundry. Even if you combined a fabless company and TSMC there would still be differences. Can you even compare Intel and Samsung?

That is why it is a little funny when people pit Intel against TSMC or TSMC against Samsung. It's like comparing apples to avocados.

 

[count]

There is a very bit difference between an IDM (Intel) and a Pure-play foundry (TSMC) . A book on it would be more than 200 pages. Same can be said with the difference between a fabless company and an IDM or a fabless company and a pure-lay foundry. Even if you combined a fabless company and TSMC there would still be differences. Can you even compare Intel and Samsung?

That is why it is a little funny when people pit Intel against TSMC or TSMC against Samsung. It's like comparing apples to avocados.

How many pages long was Fabless? Excellent book on the subject

TSMC isn't and doesn't try to be a software company. Those optimizations are left to partners who do a pretty good job (Apple, NVidia). Intel has better performance transistors because they optimize more for transistor performance. That doesn't mean TSMC couldn't have as good performing transistors, just not what TSMC optimizes for, since they are serving a broad range of markets.

 

[Tanj]

.. whereas TSMC is very good at a market segment, but simply doesn't have the breadth of technology companies like Intel have.

The concern is that Intel monetizes little of this outside the CPUs. Yes, I have seen how engineers trust and depend upon the depth of Intel support for board design, thermals, power, PCIe, LP/DDR4/5, security, and are raising up CXL, PCIe. But these are burdens which AMD and even Ampere and ARM can surf on with little cost. I wonder how much Intel will restructure to stop doing that, and what will replace it.

Yes Intel has much stronger margins than TSMC. If you split Intel into two, with a separate IFS, then their IFS capex and opex would make it a struggle to match TSMC, while the CPU side as a fabless design house would in principle be more like Microsoft Office than TSMC fabs. If they can get to a point where they could divest a healthy IFS then their total valuation could be very high, each business tuned to do what it needs to.

 

[nghanayem]

The concern is that Intel monetizes little of this outside the CPUs. Yes, I have seen how engineers trust and depend upon the depth of Intel support for board design, thermals, power, PCIe, LP/DDR4/5, security, and are raising up CXL, PCIe. But these are burdens which AMD and even Ampere and ARM can surf on with little cost. I wonder how much Intel will restructure to stop doing that, and what will replace it.

I hope they don't axe that stuff, as it is super cool. But I do agree they probably do need to find someway to monetize this activity beyond this just acting as a MSS moat, as there is no way this sort of activity doesn't negatively impact their books.

Yes Intel has much stronger margins than TSMC. If you split Intel into two, with a separate IFS, then their IFS capex and opex would make it a struggle to match TSMC, while the CPU side as a fabless design house would in principle be more like Microsoft Office than TSMC fabs. If they can get to a point where they could divest a healthy IFS then their total valuation could be very high, each business tuned to do what it needs to.

If IFS was in a healthy place, why would you ever divest it. That would be like if Amazon divested AWS. If a subunit of the business does well there is no reason to get rid of it. If intel design eventually could get to an AMD like 45% GM and intel manufacturing only gets to a 30% GM. Divesting intel manufacturing would cause intel corp margins to drop from 59% to 45% on said IDM products, decrease their total revenue, and lower intel's TAM substantially. Additionally, they will then start losing 100% of any MSS lose to other design houses, rather than having the opportunity to capture some of that value. As I see it, the only situation where it would make sense to divest the fabs would be if intel was about to go bankrupt like AMD was when they went fabless.

 

[Tanj]

If IFS was in a healthy place, why would you ever divest it. That would be like if Amazon divested AWS. If a subunit of the business does well there is no reason to get rid of it.

As I have said before, they are very different business structures. Use of capital is different. Customer bases barely overlap (indeed, they overlap as an almost 100% conflict of interest, and nothing else). The fab needs to listen to other voices, equally. Both to keep the customer happy and to make good choices on future technology, capacity, and location. Meanwhile, the IP side needs to be able to shop freely for technology.

Actually, why is AWS still part of Amazon? So Amazon can have a cash cow to prop up other businesses? But that is a handicap on AWS using its cash for its own growth, its own compete. Amazon should issue bonds or shares if it wants to finance internet pharmacies and warehouses.

Vertical integration works in immature industries where the infrastructure is being invented. It can also, if you ignore legal issues, be very profitable in monopoly or oligopoly situations. Arguably Intel has been an IDM for both reasons and both phases are now in the past.