As I write this I sit heavyhearted in the EVA executive lounge returning from my 69th trip to Taiwan. I go every month or so, you do the math. This trip was very disappointing as I can now confirm that just about everything you have read about TSMC 28nm yield is absolutely MANURE........
> Nenni, what about 28nm production stoppage at TSMC? Was the rumor true? Are the production lines up now?
Not true at all. TSMC is much more transparent about wafer and yield data. Try and get wafer info out of Intel or Samsung.
Q1 numbers will be announced next Thursday. Lets see if 28nm ramps as I predict.
A better question for him.... Whose payroll are you on?
Is TSMC possibly one of your paying customers? I would image they would welcome just the sort of 'fair and balanced reporting' that you keep putting out there.
Your blog reminded me of the old saying inside the bathroom door at the Dew Drop Inn: "Here I sit so heavy hearted...." Except the writing on the bathroom door made more sense, and was funnier.
If I average your guesses, I come up with 6% of wafer REVENUE for 2012. Revenue being the important point here. The characteristic of the TSMC 28nm business is small volume and many customers at this point. That all means very high wafer prices.
I've estimated previously that the 28nm prices could be 5 times the average price of wafers from TSMC. Converted to 300mm equivilent wafers and to make the arithmetic easier, let's assume 5 million wafers for $14 billion USD revenue. Average TSMC price is $2800 per wafer.
28nm could be $14,000 per wafer. 6% of revenue is $840 million. Divided by $14,000 is 60,000 wafers for all of 2012 and all customers. That is real mice nuggies. Use any price you wish (obviously it will be much higher than the average price), and the number of 28nm wafers out of TSMC is very small.
I think DAN has just verified what we know and that is that 28/32nm is extremely tight. It's going to be a tough time for many fabless semiconductor guys for the next 2-3 years and maybe forever.
TSMC has underinvested in 28nm and, if DAN is right, they represent 100% of foundry capacity. this can't be good for QCOM, NVDA, TI, etc.
Apple alone would need 850,000 wafers per year....2/3 that if they could find 32nm capacity.
Based on what DAN just told us, does anyone doubt that Apple has a deal with Intel?
> If I average your guesses, I come up with 6% of wafer REVENUE for 2012. Revenue being the important point here. The characteristic of the TSMC 28nm business is small volume and many customers at this point. That all means very high wafer prices......
A very valid argument, thank you for that line of reasoning.
I would not expect 28nm shortages to last past 2012.
UMC has 28nm ramping as does GFI and Samsung. The customers that Intel have taken thus far are very small but shortages will certainly help the Intel foundry business.
About that time 20nm will start ramping so we start this YIELD nonsense all over again.
The question I have is when will Intel release 14nm foundry business? What a great race! Thank you Intel.
"Apple alone would need 850,000 wafers per year....2/3 that if they could find 32nm capacity. Based on what DAN just told us, does anyone doubt that Apple has a deal with Intel?"
If Apple leaves itself at the mercy of TSMC and Samsung's declining fabrication capabilities, they deserve what they get. I can't believe that they are so lame as to not be able to grasp where all this going. I'm thinking Apple is talking seriously about a 14nm move to Intel...
I think your write up is a riot.
What I got out of it is that your 69 trips to Taiwan suggests that you are either financially independent and a frustrated Dead Head who missed out on touring with the band, or that you work for an insanely stupid employer.
How else could anyone explain your travels and the unmitigated dreck you spew.
What a laugh!
Regarding just 28nm. From your article, I think you have one data point where you got the 2% from TSMC public documents. You then guessed (word you used in article) that the TSMC supplied 2% number would go to 4% in April, 8% in July and 12% by October.
It just gives me the impression that you have drawn a straight line based on only one data point. The problem that I have is that the slope of the line is based on your "guess".
28nm wafers are all on 300mm. One 300mm wafer contains 2.25 times the silicon as the 8 inch wafer. TSMC will charge a company at least 2.25 times as much for the 28nm 300mm wafer as the older technology 8-inch and maybe a multiple of that for the more attractive 28nm process.
It appears like what you are saying is ...
If my guess is right, then I will be right.
.... and therefore the opposite ....
If my guess is wrong, then I will be wrong.
If that is what you are saying, then I agree with you.
> Regarding just 28nm. From your article, I think you have one data point where you got the 2% from TSMC public documents. You then guessed (word you used in article) that the TSMC supplied 2% number would go to 4% in April, 8% in July and 12% by October. <
Q1 wafer numbers will be released next Thursday. I wanted my predictions on record to see how they track. I have over/under lunch bets so I was being conservative.
I actually feel dumber for having read that collection of barely coherent random thoughts. You actually put your guesses in writing for 28nm yields without proof of anything, then point to a yield chart which shows 45nm % as it was ramping like this somehow proves anything.
Full disclosure D.A.N. how much have you invested long TSMC and short INTC?
> Full disclosure D.A.N. how much have you invested long TSMC and short INTC?
I'm happy to answer questions, personal attacks aside.
I do not own TSMC or Intel stocks. If I write about a stock I own I have to make a disclosure on that article.
If I did I would be long on both.