Intel is flogging some its flash foundry capacity in its IMTF joint venture to Micron, its partner in their manufacturing love-in.
Chipzilla is selling its stake in two IMTF wafer plants to Micron for $300m cash and a $300m amount held by Micron to pay for future NAND supplies from the plants or to be refunded as cash. Intel and Micron have also extended the scope of their development agreement to include "emerging memory technologies".
The two plants are IM Flash Singapore (IMFS) and the IM Flash Technologies (IMFT) in Manassas, Va. A third plant in Lehi, Utah, continues to operate with minimal changes and is churning out 20nm NAND.
Both Intel and Micron provided canned quotes extolling the joint venture and this development, featuring phrases such as "a model of innovation, productivity and effectiveness" and "industry-leading NAND Flash memory technology". An Intel executive said the JV change "gives Intel better flexibility to meet growing demand for SSDs and other products" - but, if there is growing demand, why is Intel giving up capacity? It's like a coded announcement. What does it mean?
Decoding the emerging memory technologies bit is fairly simple: back in October 2009 Numonyx, 45 per cent owned by Intel, was trying to develop phase change memory (PCM) as a NAND-follow-on with DRAM-like speed and addressability. It was facing a scaling problem; it couldn't increase the capacity of PCM dies without killing access speed.
Step forward Intel which found a way to help Numonyx build vertical stacks of PCM without a performance reduction. A few months later Micron bought Numonyx, paying Intel and the other owners in Micron shares and not cash.
In effect Intel is getting a stake in post-NAND technology development. More on this in a minute.
Decoding the wafer plant interest sales is harder. Is Intel getting out of the flash business? Such a move wouldn't fit with its aggressive introduction of new SSDs and its roadmap.
Jim Handy of Objective Analysis reckons the explanation is straightforward. Intel does not having the channel to sell all the chips it receives from the joint venture that are in excess of its own SSD requirements and the chips it sells to Hitachi GST. That firm is likely to be acquired by Western Digital this month and we have no word on WD's thoughts about the longevity and usefulness of the Hitachi GST-Intel SSD supply and development agreement. WD might well decide its SSD future lies in vertical integration, as is the pattern with its hard disk drive business.
Handy states: "Intel does not compete in the NAND chip business against the four other NAND chip suppliers: Samsung, Toshiba, Hynix, and Micron." He says Intel is avoiding the situation SanDisk found itself in when it had excess chips in a few years ago:
At that time SanDisk had too much capacity for its needs, and had no existing means of selling excess raw NAND flash chips (chips sold without controllers) on the open market, that is, it had no raw NAND chip sales channels, no existing customers for raw NAND flash chips, and no internal support system for raw NAND chip sales. All of these would have been very costly to develop in order to solve a short-term problem and would establish the company in a business in which it had no interest.
SanDisk solved this problem by selling a significant share of its interest in joint venture fabs to partner Toshiba, with an agreement to purchase chips from the JV in the future. Intel is taking the same step.
... Both Intel and SanDisk built NAND manufacturing capacity to satisfy overblown growth expectations for the SSD market. The SSD market has grown more slowly than these companies expected, and they subsequently require less NAND than they planned, so they are taking steps to match their capacity to their needs.
It's no big deal really. Micron is playing ball with Intel because it has the channels and market focus to sell more NAND chips and so compensate for a stagnating DRAM market. Also, by teaming up with Intel for its post-NAND research and development expertise Micron increases its strength in that area.
Handy says the flash players all have post-NAND bets: "SanDisk and Toshiba seem to be going in different directions, with SanDisk promising ReRAM and Toshiba MRAM. Although Hynix and Samsung have been less vocal, Samsung's acquisition of Grandis in August 2011 indicates that this [Spin-Transfer Torque RAM] is the company's intended direction."
We know Hynix and Toshiba have signed a joint MRAM development agreement and Hynix also has a deal with HP concerning Memristor production.
This aspect of the IMTF re-jig represents Intel and Micron's stake in the post-NAND casino, although Micron has multiple bets.
The transaction should close by the mid-point of the year. ®