Is Intel Going Back Into The Memory Business?

Mar.18.13 | About: Intel Corporation (INTC)

There is an 800 lb gorilla in the computer industry room and everyone is pretending that it is not there.

The 800 lb gorilla is SSDs, Solid State Drives. SSDs improve the experience of personal computing by providing much higher performance, 10 second cold boot times, thinner, lighter, more reliable, longer battery life devices. Who wouldn't want a solid state drive in their next PC? Well, they are expensive about $1 per GB, $128 for 128 GB and $256 for 256GB. Even given the expense, all areas of the computing world will move slowly or quickly to SSDs to replace HDDs

Here's the problem:

SSDs are made from NAND Flash memory and a sophisticated controller chip, a few other components and that is about it. Intel, in defining the Ultrabook first insisted on SSDs and then relented to allow hybrid drives. A hybrid drive is a tiny (20GB) SSD bolted up to a large HDD. We get the fast boot times, but nothing else.

The SSD genie will not be put back in the bottle. The problem is that we need a whole lot more worldwide NAND production capacity to begin the mass conversion to SSDs in PCs.

Consider: If the average capacity of a SSD is between 128 and 256GB or 192GB, the average number of 128Gb chips required to build only a third of PC with SSDs would be 100 million units times 12 of those 128Gb NAND chips. That chip in MLC (basically 2 bit per cell) 20nm technology is about 200 square millimeters in size. 1.2 billion of them would be required. At about 250 memory chips per 300mm wafer, 4.8 million new wafers would be required. Since a giant wafer fab can process about 1.2 million wafers per year, the new capacity required will be about four $5 billion fabrication facilities. For a full conversion to SSDs, for 340 million PCs, the number of fabs required would have to be over three times, or 12 $5 billion fabs. At this time no new fabs are being planned that we know of. Who has the money to build 12 fabs, or even 4 fabs?

Without enough NAND memory to provide the SSDs required or desired in the new Ultrabooks or convertible computers, Intel's CPU chip business will have another headwind to contend with.

From here on is my speculation. Let me repeat that; speculation, as in conjecture.

Intel (NASDAQ:INTC) has spent about $35 billion in CapEx for new facilities over the past three years. These facilities are now running at less than 50% capacity with more factories scheduled to come on line this year. The total Intel capacity can build all the world's need for CPU chips and all the world's need for mobile chips, with about 50% left over.

Intel builds SSDs today.

So, NAND flash memory is important to Intel to build SSDs to keep the PC user experience moving up and to the right.

Intel has made a multi-billion dollar investment in ASML (NASDAQ:ASML) to hurry along 450mm technology. There are two parts to 450mm technology; 450mm handling equipment and EUV (Extreme Ultraviolet Lithography.) The thought is that the two technologies will come at the same time. Maybe not, maybe the handling equipment is available sooner, maybe even now.

Intel is scheduled to open a factory this year in Chandler, Arizona that is scheduled to produce 14nm process on 450mm wafers.

Now my speculation is: Intel has developed a 14nm NAND cell that would shrink the 128Gb chip size to 100 square millimeters. A 450mm fab would produce over twice the square inches of silicon that a 300mm facility would. Thus a 1.2 million wafer 450mm plant would produce about 1.4 billion 128Gb 14nm NAND chips. That would be one $10 billion factory with the output of four $5 billion 300mm plants.

Can Intel make money in the NAND business? Let's just figure a $7000 cost for those 450mm wafers. The 128Gb chips would have to sell for about $14 to Intel itself or the world at large in order to deliver a SSD at $1 per GB. Those big wafers could produce about 1200 100 sq mm chips. At $14 each, the wafer is worth about $16,800 against a $7000 cost is 58% gross margin. When the depreciation is out of that factory, the gross margins would reach into the high 60% range.

There are a lot of numbers to be digested here, but the bottom line is that Intel can make a lot of money on the NAND memory that they are going to need for those SSDs in Ultrabooks. And, no small thing, Intel would be first to 450mm wafer production.

Just another reason to own Intel.

Disclosure: I am long INTC. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.