Micron Technology (MU) is in the final stages of an acquisition of Elpida Memories of Japan. The price for Elpida started out at $2.5 billion. Since the price is actually denominated in yen, the price paid in dollars declines with the decline in the value of the yen. Since the deal was approved, the yen has declined vs. the dollar by about 20%, thus reducing the amount to be paid by Micron by $500 million. A group of American unsecured debt holders appealed the court decision at the last minute on March 31.
Knowing that the acquisition will ultimately close and given the stimulus actions taken by the Bank of Japan, after the appeal, one could come to the conclusion that Micron is in no hurry to close the deal.
Product Of The Year Award
While this is going on, the news from Micron and Elpida just keeps getting better.
The interesting thing about this is that the HMC is a joint development of Micron and Intel (INTC). There is no mention of this in the award. In fact, you have to look hard for any indication of Intel's participation in this groundbreaking development anywhere on the Micron website. There is, however, a Hybrid Memory Cube Consortium made up of Micron, Samsung (OTC:SSNLF), Hynix (OTC:HXSCF), IBM (IBM) Xilinx (XLNX), Altera (ALTR) and other interested parties, but NOT Intel. One can only surmise that Intel's absence is a result of an entanglement with the patent troll, Rambus (RMBS), which ultimately resulted in 10 years of legal hassle for all DRAM manufacturers. For those of you suffering from insomnia, the Rambus short story is here. Intel will never again want to be viewed as endorsing ANY kind of memory technology.
The HMC crashes though the "Memory Wall" with a product that has 15 times the bandwidth of traditional memory modules, while consuming only 30% of the per bit power, and does this in a footprint of 10% of the other memory modules. This will be a big hit with the high performance computing and high performance server crowd, and eventually find its way into PCs.
This HMC technology involves TSVs (Through Silicon Vias), in order to stack the memory chips on top of each other. TSVs are a difficult technology that has apparently been mastered by Micron (Intel?).
DRAM contract memory prices continue to move toward convergence with spot prices, with the latest numbers for a 2Gb chip as of the first half of April being $1.44 contract vs. $$1.70 spot. The lows back in ~September were about $.60 and $.70, respectively. The low prices represent about a negative 50% gross margin. The current contract price represents about 38% positive gross margins. The contract prices will probably continue to increase toward that $1.70 contract price, which would produce 47% gross margins. The other good news is that the conversion from 2Gb chips to 4Gb chips is moving right along, since the normal 4GB for a PC can be accomplished with only eight of the 4Gb chips. The 4Gb chips sell for twice the price of the 2Gb chips, but only cost about 50% more to make, so the gross margin on 4Gb chips is already over 50%.
Mobile DRAM is a little harder to pin down from a price standpoint, but back in September of 2012, iSuppli published a costed bill of material for the iPhone 5. The mobile DRAM was estimated at $10.45 for 1GB of mobile DRAM. That DRAM was probably made up of two 4Gb chips. That would make those chips cost about $5.20 each. The best information that I can find is that mobile DRAM cost about 25% more to make than PC DRAM, but sell for 100%+ premium. Back in September, a mobile 4Gb DRAM might have cost $1.50-2.00 to make. So, worst case, Mobile DRAM was generating 50-60% gross margin, even in September. The problem is that Micron doesn't sell much Mobile DRAM, but Elpida did and does! The Apple (AAPL) orders for mobile DRAM were given to Elpida in May of 2012. Using 16 weeks for initial delivery, Elpida has only been shipping memory to Apple since September of 2012. The most recent financial information that we have for Elpida is in the Micron 8K filed back on February 5, 2013, and it included almost no post September 2012 information for Elpida. It is entirely possible that Elpida has been shipping 600,000 memory wafers per quarter since September at a revenue rate of $1.8 billion/quarter and 50% gross margin per quarter. That kind of business should generate about 25% net profit or $450 million per quarter, that we have no way of seeing until the deal closes. Given three quarters of zero visibility, Elpida could generate $1.35 billion dollars by the time the deal closes. It would be deliciously ironic if Micron uses Elpida profits to pay the first payment of ~$700 million to purchase Elpida.
NAND prices are up about 20-30% from the bottom back in September of 2012, and the same thing is happening here as in DRAM. The prime market is moving toward the 128Gb chip since it only takes eight of these chips to make a 128GB solid state drive. The 128Gb chip sells for about twice the 64Gb chip, but only costs 50% more to make. So Micron NAND, in general, is about a 40% gross margin business from maybe a 20% gross margin business back in September. NAND could be making $250 million per quarter for Micron at this point.
An interesting point here is that about 25% of the Micron NAND memory is used in house for SSDs. The price for the SSDs is tracking the price for individual chips at about $.80 per GB. That makes the SSD business about a 40% gross margin business as well.
Paul McWilliams of NextInning is calling for possible allocation of NAND memory this summer, which would blow any memory price estimates I make here out of the water.
It appears that Micron and Elpida are hitting on all cylinders. The combined businesses should be between $16-20 billion in revenue and earn between 10% and 20% net profit, or $1.60 to $4.00/share within 1-3 quarters following the close of the acquisition.
The closing could be dragged out until the May-June timeframe, or we could wake up to a settlement resolution of the appeal any day with an accompanying pop in the share price.
I'm very long MU call options of various strikes for January of 2014.