Yesterday morning DigiTimes published an article, the content of which was attributed to DRAMeXchange, suggesting that mobile DRAM prices had declined 5-10% in January. My first comment is that the credibility of DigiTimes has been spotty, but given that, there is no reason to dispute that mobile DRAM could be down in price in the weakest month of the weakest quarter.
Naw, I am going to go into the credibility issue:
In the article:
"In addition, production at SK Hynix' DRAM fab in Wuxi, China, which was hit by a fire in September 2013, is recovering gradually. Earlier supply concerns have been eased, according to DRAMeXchange."
A Bernstein analyst comment after a three day road show with the Hynix CFO and IR team:
"..Wuxi fab is back to full output and normalized yields as of January and yet DRAM pricing has not seen any impact due to extraordinarily low inventory at Hynix, other DRAM makers, and PC OEMs. Despite current seasonal weakness, it will take time for DRAM inventories to normalize and by then we expect demand to pick up again for the smartphone, tablet and PC build cycle."
Who do you believe?
Now back to the price theme.
We can easily look at one type of DRAM price on DRAMeXchange and that is PC DRAM. Those prices are in module form and 1, 2, and 4Gb chips. The configurations are DDR, DDR2, DDR3, and soon DDR4. In mobile DRAM there is LPDDR, LPDDR2, AND LPDDR3, in three densities each.
In the old days we had one type of DRAM and maybe two densities to worry about. Now we have at least 12 types and densities/speeds for PC DRAM. We have the same number of mobile DRAM parts, and then we have about four configurations of graphic DRAMs, known as GDDR2, 3, etc. Then there are a multitude of module and package combinations and MCP (Multi Chip Packages). All of these possibilities are in flux all the time. For example 2Gb chips are moving to 4Gb chips, soon 4Gb chips will be moving to 8Gb chips DDR3 will be moving to DDR4. The point of all this is that there is no way that even DRAMeXchange can discern with certainty whether the business is moving up or down on their tiny bit of a very complex business. What DRAMeXchange was good for was tracking the remarkable price increase of spot PC DRAM devices from December 2012 to the present. It will continue to provide some very rough approximation of general DRAM prices.
Now, about contracts. Kipp Bedard, VP of Investor Relations at Micron (NASDAQ:MU) gave us a little insight yesterday into the world of "contracts" in the memory world. Here's how it goes: Graphics and mobile contract for six to nine months, network customers about six months, data center customers about three months and finally PC customers who want to negotiate prices every time they change socks. Even the PC guys now have to commit for a month.
As we know, Micron had targeted the higher end networking and data center business and had reduced their exposure to the PC segment. Elpida could build the high quality, low power DRAM used in mobile and graphics. So both Micron and Elpida actually had the longer term contracts when prices began to rise. This is why it has seemed that the price increases took longer to materialize at Micron…because they did.
Imagine one of those long-term contracts (nine months) being re-negotiated today. The new prices could perhaps be slightly lower than some December contracts, but compared to the old contract prices, there would be substantial increases. These are three month contracts that have been re-negotiated 3-4 times since December of 2012, each at much higher prices that the previous contract.
This is why I have said in previous articles that prices at Micron could continue to increase long after peak market prices have been achieved and those price will stay high longer if there is a price decline.
Tracking NAND prices through DRAMeXchange is truly a fool's errand. In the case of Micron, 40% of the NAND parts are destined for the SSD business (solid state drive), with much of that being retained by Micron's Crucial division to become SSDs. Generally speaking these parts will be the largest (128Gb) available, in order to make SSDs with the fewest number of chips. The same can be said for the other three NAND manufacturers since they all participant in the SSD business. DRAMeXchange doesn't even post a price on a 128Gb chip.
Now that doesn't mean that we can't take a shot at what Micron's NAND chips are being sold for. We can look down the list at the DRAMeXchange prices for SSDs. For example a Crucial 120GB SSD (actually 128GB) has an average price for the period covered of about $99. That SSD is made up of eight 128Gb NAND chips, a controller chip at a couple bucks, and another $3-4 in passive components and assembly. Let's assume the seller made 10% on that SSD. That would be about $89 coming out of the Crucial (Micron) factory. If we subtract about $6 from the $89, we would have the value of the eight 128Gb NAND chips. $83/8 = $10 and change for the NAND chips. The cost on those chips is about $5, so Micron is getting 50% gross margin for big NAND chips configured as SSDs. BTW, the Crucial price is a very low price for a 120GB SSD, $120 is more realistic. At that price the NAND chips are worth about $12.50 or 60% gross margin.
It is just a matter of time before about 70% of the total NAND dollar output will be in the form of SSDs made by the NAND manufacturers. The rest will go into mobile devices and memory cards, thumb drives, etc. The bottom line is that DRAMeXchange and other guessers tells us virtually nothing about the real trend in NAND pricing, we have to look at SSD pricing for that. Contract or spot pricing for below 128Gb chips is pretty irrelevant. Those parts go into products that will be dwarfed by SSDs and mobile.
The change in the memory business in a decade is remarkable.
In 2004, there were two, maybe four types and sizes of DRAM and two market segments, PC and Networking, and NAND was small and in iPods and thumb drives. The supplier base was 8-10 manufacturers fighting tooth and nail for every point in market share.
In 2014, there are a couple dozen types and densities of DRAM and seven definable market segments with different requirements. NAND is huge in mobile and SSDs with all the other trinkets still there. NAND now exceeds DRAM in total sales with a $50 billion SSD opportunity on top of that right around the corner. This is all being supplied by a base four memory companies and only three of them are DRAM manufacturers.
I could be wrong, but I think if there is any pricing contention in the memory business from now on it will be caused by accidental and short-term collisions in certain market segment or a particular device type
By the way, here's the latest DRAM assessment by DRAMeXchange. Trying to reconcile the two articles just makes your head spin.
The bottom line is that Micron is worth at least 2X the current price and the analysts are so conflicted that there is no consensus on earnings estimates or target price.
That, my fellow retail investor, is good news for us!