You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
It's clear that the fabless model is not broken, but it's also clear that Intel is able to leverage their process technology to give them an advantage (a technology lead) over what their (fabless) competitors can deliver. I think Bohr was mainly targeting their chief competitor, AMD, with his remarks. Let's not forget that microprocessors are still Intel's core business and he knows that whatever he claims will be something that AMD has to deal with in their next investor/analyst call. It's just good business to hit your competition hard when you have an advantage. I'm not arguing that AMD doesn't have some advantages of their own, but so far Intel has been winning the race since AMD went fabless- and it looks like Intel is accelerating. From a process perspective I don't think that even TSMC can keep up- and that's really a financial argument, not based on the ability of the company or their people. Intel can use their microprocessor business, which is very high margin- and can tolerate a high cost for new processes- to fund their process development at every new node. They have more to spend and can do it better and faster than anybody. Let's not forget that anything within 5 years of release is no longer non-competitive, so at that point it isn't an eco-system, but individual cost to develop a new technology for manufacturing. Sure the common platform shares development, but TSMC isn't involved in that, they have to go it alone- Intel can spend more (than CP or TSMC or anybody else really), so they can afford to be the first with every new technology- and it looks like that is their game. TSMC may have plenty of revenue, but even Samsung, which processes more Si than Intel, doesn't have the cash to match Intel in development (look at Intel revenue vs. SS, and then remind yourself SS is actually processing more Si- memory is lower margin). Intel has a lead and for the foreseeable future it looks like they can keep it. But I really think they are just trying to stick it to AMD with their comments- it's a pretty clear ploy from that perspective- and TSMC doesn't really need to worry about it- they know their model works. AMD has to answer all the questions now- "Why'd you go fabless?" "Aren't you just falling farther behind?" " How can you catch up?" etc. I'm sure Bohr will somehow get a listen to that audio and have a good laugh.
Thank you to everyone that has replied... I think there is general consensus that it's largely hot air w.r.t the fabless model (as expected) however it's also clear that if you are dominant in a market that can stomach high margins then owning your own state-of-the-art fab is a competitive advantage. Intel have done a phenomenal job at maintaining their margins and the number 1 slot since the early 90's. However, we can see the end of Moore's Law now (there is a finite size transistors can shrink to - atoms do not get smaller) hence a finite time frame for manufacturing technology prowess based on the 2D Moore's Law approach. Equally, in many markets the existing processor capability outstrips the application requirements at which point other constraints become more dominant (price, power etc.) so it is inevitable that Intel will meet an increasing need to figure out their leadership position on another dimension at some point even if it is not immediate. This is not beyond Intel of course however there is a big eco-system to compete with and given their past behaviour of market dominance ("creosote tree") that eco-system is will be ever cautious on letting them in.
Hi, in Asian. Most of foundry do not post positive profit as being the best foundry require latest tools which normally cost a bomb. And why are people still keen in setting up foundry..... I believe the only foundry that is striving is TSMC. Can you name others? Posted by Richard Teo
I agree, the main thing Intel will have to learn to succeed as a foundry is - put in black and white terms - evolve from an arrogant leader to an obeying slave.
This is not a kind of subject that I know too much.
For sure, there are some (internal and hidden) reasons for Intel affirming it worldwide. From my point of view and for what I learnt so far, working for semiconductors companies, I won´t bet on the collapsing of this model, at all.
We have many examples of successful semiconductors and high tech companies that started using this model, several years ago and continues growing and gaining market segment share, they are not even slightly concerned or dissatisfied with this, quite the contrary. Posted by Osnir Gracia
Another IDM who was ahead of TSMC in 32nm is Samsung. Although in Common Platform club it seems to be ahead in advanced processing of GF - probably thanks to Apple-ASIC and Memory volumes
The landscape is fast morphing -- In 10nm Intel, Samsung and TSMC will be in 450mm wafers - Intel in 2016/17.
One item that will further differentiate Samsung could be 3D packaging of Memory and AP via TSVias. An IDM like Samsung has both ICs internally - no finger pointing for yield.
All,
I understand that this topic is communciation toward market to put pressure on competition. However, I would like to know if someone have seen hypothesis, and calculation that support this assumption.
And how the transition toward 2.5D and 3IC will change the value change and reset the game.
