Will Intel have to change its business model to compete in the netbook market? After all Intel likes fat margins on its CPUs, while netbook CPUs are going to be low-margin, and Intel builds its fabs for high performance, while netbook CPUs require low-power CPUs - or they will when ARM-based netbook CPUs hit the market later this year.
A good many of these ARM-based netbook chip-sets will find their way to TSMC to get fabbed. Qualcomm and Freescale are likely to be the first out of the starting gate. TSMC is in pole position to deliver the high-performance/low-power processes needed to deliver netbook chip-sets which can combine a good computing experience with solid communications capability and an acceptable battery life.
"We believe in the convergence of computing and communications and our technology is ready for the convergence. We can offer high performance computing processes together with low power communications processes. We believe we have a lot to offer," Maria Marced, President of TSMC Europe told me yesterday. TSMC makes PC chip-sets for Intel.
Intel can do high performance but, as sure as eggs are eggs, it can't do low power. And even if it develops a low-power process would it want to spend the money to build a fab to run it?
Then there's the margin issue. Intel likes nice big fat margins on its Pentiums and Celerons and, with Atom based netbooks currently retailing at $400+ it can still get a reasonable margin on an Atom, but when the likes of Freescale and Qualcomm hit the market enabling ARM-based netbooks retailing at $200, what does Intel do to maintain its margins?
Intel claims it can out-run the opposition on process technology and deliver smaller, cheaper die with lower power than the opposition, simply by going faster down the geometry trail but, in terms of volume production, Intel seems to be on a par with TSMC with both companies currently ramping 40nm/45nm processes.
So faced with both a process issue and a margin issue what does Intel do?
TSMC's Marced reckons: "Intel will have to change their business model to compete in the netbook market."
Intel seems to be on a par with TSMC with both companies currently ramping 40nm/45nm processes.
No true, 45nm has been in Volume production for over a year know, while the appear to be ramping 32nm.
Thanks Anonymous, on the 40/45nm comparision inn the absence of figures for wafer throughput it's so hard to judge. When you say Intel is ramping 32nm, do you mean it's selling 32nm chips?
First of all, Intel will ramp their high-performance 32nm process in 4Q09 as far as I can tell, while TSMC will ramp their 40nm process in late 1Q09/early 2Q09 for GPUs. TSMC's density on 40nm is excellent, and quite a bit better than Intel's; although they still can't match Intel's 32nm it would seem, not very surprising there.
Secondly I agree with most of what you said David, but you've got one tiny fact wrong: Intel does have a low power process, and it is really really good on 45nm. TSMC is just spinning on that point or (pretending to be?) ignorant; there's one paragraph on it in this article: http://www.semiconductor.net/article/CA6623854.html
However their low power process will always lag behind their high performance one timeframe-wise, while that is not the case at TSMC, so Intel is similarly spinning here: their traditional time-to-market advantage for process technology does not apply in this market. The same was true for XScale which they kept using very old fabs for when it might have benefited a lot from leading-edge ones.
I suspect the current Atom platform ironically does not use their low power process yet (except for the 130nm chipset; ugh!), while Moorestown will. This, along with a higher degree of integration, is probably where most of their claimed power efficiency gains will come from rather than circuit design or architectural innovation...
Finally, while I understand TSMC's point of view when they say Intel's business model needs to change, I think that's missing the point. Intel is currently still benefiting from extremely high ASPs for chips that offer less and less differentiation.
If ASPs go down dramatically, nobody cares if they manage to ship 25% more units. They're still toast, and shareholder value will still go down the toilet. ARM netbooks are, in my view, just one factor that will force them to be more aggressive in pricing, and risks pushing them into a downward spiral.
Arun, thanks for putting me right on the lntel-TSMC process debate, and again on Intel having a low-power process. This is certainly news to me. It will be interesting to see what effect the low-power has on Intel's 45nm Atoms.