Intel (INTC) just finished 2012 $53.3 billion in revenue, $49.4 billion of which was semiconductors. About $4.4 billion was of lower technology and probably outsourced. The $45 billion left over was spread over about 350 million x86 processors (client and server) for about a $128 ASP. The average die size was about that of an i5 processor, or about 160 square mm. That average chip probably yields about 140 good die per wafer. The average wafer, therefore, generates about $18,000 in revenue. That means that 2.5 million wafers were run to generate $45 billion in x86 revenue. The company claimed that it was running below 50% utilization rate in the fourth quarter (to burn off some inventory), but the quarter before, 50% utilization was mentioned.
So, Intel went out of 2012 with capacity of about 5 million wafers of leading edge technology. That means that Intel could produce $90 billion in revenue at full factory loading. 100% factory loading will never happen, so crank that back to $80 billion. The 2013 CapEx will be $11 billion ex the 450mm spending. The $11 billion will produce at least another million wafers per year or $18 billion worth of capacity that could be on-line by the middle of 2015. (Note: not all the $11 billion will go toward new capacity. About $5 billion is for maintenance of existing capacity, so about $6 billion will go to new capacity. Using the same reasoning, only $21 billion of the 36 billion total CapEx spending from 2011 through 2013 will have gone to new capacity, which is capable of producing about $63 billion in new output. Obviously the company had some capacity, that is still being used, before the $36 billion total new CapEx orgy began in 2011.
Anyway you cut it; Intel has about $90 billion of total capacity today, with another $18 billion coming on in about 18 months.
What's going on?
Intel is noted for its ability to trim output quickly in slow times. Because of this the company didn't get hit hard in the Great Recession. The company is also noted for carrying some extra capacity, so that they never run short of the market on Intel products, but well over 2X the required production capacity is more than a little guard band for the upside.
Is Intel management just dumb? I've never known that to be the case.
What products need this type of technology? The chips involved need to be very fast, very low power, and physically large.
PCs and servers need it, of course, that is the present $45 billion in business.
Smartphone and tablet Application Processors and Baseband processors need the technology. That market, from low end to high end, might average $40 per unit with line of sight to 1 billion units worldwide by 2015-16. So there is $40 billion.
High end networking chips can use the technology on a foundry basis. The rumor is that the Cisco deal is $1 billion per year. Maybe there is another 1 or 2 billion dollars in that market.
FPGAs need the best technology available. Judging from the Altera and Xilinx recent cost of goods, that market is about $1.4 billion total.
That's about it. If Intel got every penny of the expected high performance business it would explain the need for the $45 billion of new capacity that is sitting unutilized today. That still leaves $18 billion worth of future new capacity unaccounted for.
What would it take for Intel to get every bit of the high performance business available in the relatively near future?
The semiconductor industry is all about moving decimal points around. The number of transistors on a CPU chip has grown from 1000s in the beginning to billions today. A million times growth. The CPU clock rate has gone from a few megahertz to a few gigahertz in that same time. The word length has gone from 16 bits to 64 bits. A 64 bit word can carry something like 281 trillion times the number combinations of a 16 bit word! Memory chip size, in my career, has gone from 1Kb to 4Gb, a million times growth, six decimal points.
The network business and the FPGA business combined do not move the needle much. A deal in hand from Apple for 200 million "A" something chips, even at $30 each, is only $6 billion, against $45 billion in new capacity? Not even close.
In order to render all other technologies obsolete in the mobile market, something miraculous would have to happen. Something like Intel would have to reduce the power of mobile chips by a factor of ten; move the power decimal one place to the left. They would also have to move the speed performance one decimal point to the right, or ten times as fast as the best current technology. It wouldn't hurt if the price on such a device was the same as or less than today's price.
The above conditions would be enough to move all of the mobile business to Intel in the short term. Maybe only factors of five would be good enough to make the move.
I have no idea whether the immense amount of new capacity is aimed at a run for every dollar's worth of mobile business or not, but $45-63 billion worth of new capacity is a nugget of information that, if I were an analyst, I would ask Intel management about again and again until my voice was hoarse.
That huge chunk of capacity needs to be used in the short future; it is a wasting asset.
There is something unprecedented about to happen, and it has to happen very quickly in order to use that capacity. By the time whatever this is is known, it will be too late to take a position in Intel to make the really heavy sugar.
I'm in Intel up to my ears.