The Old Order Changeth: Samsung Takes the Crown

The most famous line of Tennyson's poem Morte D'Arthur is "The old order passeth, yielding place to new." It is often quoted when a new king, real or metaphorical, is crowned. Last week, Intel and Samsung announced their financial results. Intel grew 6% to $63B for 2017. Samsung's semiconductor business grew to $69B. For over 25 years, since 1992, Intel has been the largest semiconductor manufacturer. Now, Samsung is.

There are two main reasons for this: mobile and memory. Samsung is big in both, and Intel is in neither. However, a lot of the headline numbers for last year's semiconductor market growth is not increases in wafer volume, but strengthening DRAM prices (good for Samsung, no direct effect on Intel).

Intel

When Intel started it was a memory company. There were not enough transistors on a chip to build a processor. The microprocessor was invented by Intel as part of a custom project for the Japanese company Busicom. You can see the product that resulted if you visit the Intel Museum in Santa Clara (see my post The Intel Museum for more details). The first merchant microprocessor was the Intel 4004. Given the specifications for a modern microprocessor, you can have a laugh at the numbers for the 4004: manufactured in 10um (10,000nm the way we talk today), it had 2300 transistors and a clock rate of 0.75MHz.

When the Japanese decided to get seriously into semiconductor manufacturing, they focused on memory. It is a good choice since it has very high volume for relatively few products, so doesn't stress design infrastructure and allows a lot of process learning to happen quickly. They learned so quickly that they pretty much drove Intel out of the memory business and they wisely focused on the promising growth area of microprocessors. I consider it one of the most amazing business transitions ever, since usually a major transition like that requires the existing management team to be fired, and the new team to change the strategy. Andy Grove recounted the story in his book Only the Paranoid Survive. This is quoted from memory since I can't find my copy of the book:

Andy Grove and Gordon Moore are in Gordon's office. The business is not going well since they are losing a lot of money in memories. There is a board meeting the following day.
"What will happen if the board fires us tomorrow?" said Andy.
"They'll find a CEO who will get us out of memories."
The two men looked at each other. "Why can't we walk out of that door, come back in, and do that ourselves?"

They did. Intel has more recently got back into memories. It is still possible that 3D Xpoint (that Intel now calls Optane) will be a big success, but it always seems to be a year or two away from hitting it big. They are also in flash, too. However, Intel is dominant in microprocessors for PCs and servers, with only AMD occasionally snapping at their heels.

A later transition came along that Intel didn't manage so well, namely mobile. Paul Otellini, who won Apple's business for microprocessors in the Macs, said when he retired that Intel passed on the opportunity to supply Apple because the economics did not make sense:

The thing you have to remember is that this was before the iPhone was introduced and no one knew what the iPhone would do... At the end of the day, there was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more and that price was below our forecasted cost. I couldn't see it. It wasn't one of these things you can make up on volume. And in hindsight, the forecasted cost was wrong and the volume was 100X what anyone thought.

The result was that Intel "missed mobile." They had several programs internally, but Arm became the standard for mobile (see my post The Design That Made Arm).

Samsung

When Korea decided to get seriously into semiconductors, they started with memory and pretty much drove the Japanese out of the market, in the same way that the Japanse had done to the Americans. China has several memory projects and clearly are hoping lightning strikes a fourth time in the same place.

Samsung is, of course, a major memory supplier in both DRAM and flash. A little over 10 years ago, they entered the merchant foundry business. Early last year, they reorganized and made foundry a completely independent business (for more details, see my post Samsung Foundry Forum: Roadmaps). From a customer's point of view, the ideal foundry company is pure-play, with no product lines of its own. Next best, is that the foundry business really does run independently of other businesses so that they can both deliver foundry and also have products in the same space. A couple of years ago, Apple (#2 in the handset market) was using Samsung as a foundry, at the same time as they competed head-on with the high end of Samsung's Galaxy line. 

Samsung is the #1 supplier in the world of handsets, with about 30% market share. So they are big in mobile in that sense, and they consume a lot of their own chips. But they are also a major manufacturer for Qualcomm, who supply their Snapdragon chips to a lot of the handset market, especially at the high end. It must be a delicate balancing act, but for someone like Qualcomm, there have really only been two foundries that have proved that they can deliver the most advanced process nodes on a ridiculously steep ramp.

DRAM

Most (and according to some observers, all) of the growth in the semiconductor market last year is due to the strengthening of prices in DRAM. DRAM is a commodity market and the demand is rarely in step with the supply. The result is a feast or famine market, where prices can strengthen or weaken. The spot market (buying DRAMs without a long-term contract) fluctuates a lot. But large consumers of DRAM, such as smartphone manufacturers and PC makers, have long-term supply contracts. Since these are confidential, the spot market is only the roughest of guides to what is going on in the overall DRAM market, although it is pretty clear whether prices are going up or down. Those long-term contracts eventually expire.

However, this can cut both ways. Samsung at $69B last year is ahead of Intel at $63B. But if DRAM prices soften, that is not an insurmountable lead, and Intel (who is unaffected by DRAM prices) could be back on top. Another possible sign that may or may not be meaningful is that IDC is reporting that iPhone has passed Samsung in smartphones. That doesn't directly affect the DRAM market (both brands contain DRAM), although it may affect Samsung's semiconductor revenue since I'm assuming they build some of the chips in their own phones (although they also use Qualcomm). More worrying for the semiconductor market in general is that there are signs that the mobile market has peaked. It is certainly growing slowly at best, far from the 30% per year of a few years ago.

And in a weird graphical coincidence, both companies have logos that are in the form of a blue ellipse. #3 SK Hynix with a red and orange butterfly is clearly not going to get there. But #4, Micron, has a blue ellipse in their logo. They are in with a chance.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.