|Micron's NAND comments from the CC:|
So first of all, Micron's partnership with Intel on NAND Flash is, of course, a very important relationship for us as well as a significant piece of our business. As many of you are aware, we reached definitive agreements this quarter to restructure or rejuvenate that partnership. And as part of that process, we agreed to purchase Intel's share of the output from the jointly-owned NAND operations in both Manassas, Virginia, and IMFS Singapore.
The purchase capacity was roughly 30,000 wafers per month, and Micron will sell a portion of those wafers back to Intel under a for-profit take-or-pay supply agreement. The incremental profit, along with -- the incremental profit capacity, along with the completion of the initial ramp of the new fab in Singapore, will lead to structural margin improvement going forward.
The life of the remaining joint venture facility, which is in Lehigh, Utah, was extended to 2024. And the scope of our technology relationship was expanded to include certain memory technologies beyond the floating gate and NAND products we had historically been working on together. So I think in summary, both partners consider the relationship to date highly successful and were both very excited to be carrying this relationship on in the future.
On the technology development front, we had a strong quarter. We were excited to complete the construction of the new addition to our R&D cleanroom in Boise, Idaho, and begin installing tools there to support our leading-edge NAND, DRAM, NOR and phase-change memory nodes as well as a number of interesting, new emerging memory technologies. We made good progress on our next 2 NAND technology nodes in the development fab, and our -- as well as our 20-nanometer DRAM node. We also made progress scaling up to 300-millimeter substrates on our 45-nanometer NOR process and in developing a new 300-millimeter phase-change memory process. Moving forward, we look to deploy both of these new 300-millimeter NOR and phase-change processes in the manufacturing fabs over the next year.
In manufacturing, we had a nice quarter relative to production ramp and yield execution. I'm sure Ron will comment in more detail shortly, but a couple of highlights to note are the completion of the ramp to roughly 70,000 wafers per month of IMFS in Singaporeand the 30-nanometer DRAM node introduction and successful early ramp at Inotera. We also had good early yield learning -- sorry, we also had good early 30-nanometer yield out of the DRAM fab in Singapore. So in aggregate, we now have successful transfer of that 30-nanometer node into all Micron DRAM fabs. Additionally, it's worth noting that we can now ship 20-nanometer NAND out of all 3 Micron NAND fabs, the 2 wholly owned ones as well as the joint fab in Lehigh.Overall, we had strong bit growth and cost reduction in both NAND in DRAM during the quarter.
Switching now briefly to markets and our products. The second quarter was obviously a weaker environment for pricing than we would have liked. Mark Adams, I think, will likely comment more in Q&A, but seasonal factors, combined with supply chain disruptions due to the Thailand flooding, as well as, I think, general slower economic activity all had an impact.
Recently, we're seeing improvements in the DRAM market. And while we don't predict what's going to happen going forward, depending on application, I think concerns over supply seem to be having a positive or at least stabilizing effect on OEM pricing.
For NAND, the mix of high-density 2- and 3-bit-per-cell as well as seasonal demand weakness drove our ASPs lower during the quarter. While we were able to offset most or all of this with cost reductions, we did suffer some degradation of NAND margins due to sell-through of Mnemonics legacy NAND products, which had been purchased in the market at market prices. Overall, we see generally healthy supply and demand outlook for NAND moving forward as measured by -- as measured industry fab ramps are offset by anticipated demand growth in a number of key markets like smartphones, tablets and SSDs.
Wireless NOR, on the other hand, volumes continue to decline in the wireless space with the transition from feature to smartphones. Embedded NOR, however, remained strong, and we see long-term attractive margins and share gain opportunities in that space.
In the product area, highlights for the quarter include continued success with Hybrid Memory Cube enablement and design and activities across high-performance computing, including high-end server applications as well as across networking; good success with all 3 design wins across a number of different networking companies; strong design and activity of embedded NOR in both automotive and amusement markets. Additionally, relative to NOR, we're now shipping the industry's broadest portfolio of SPI NOR products.
We saw good growth in mobile DRAM bit shipments, in roughly 20% range, and we're seeing early signs of design and acceleration given some of the challenges some of our competitors are facing.
Finally, in NAND during the quarter, we shipped [indiscernible] samples of our 3-bit-per-cell, 20-nanometer, 128-gigabit NAND Flash to enabling controller vendors; we shipped over 0.5 million solid-state drives; and finally, we were prominently endorsed by both EMC as a preferred partner for PCI SSDs and our VFCache product, as well as by Dell for their PowerEdge servers with the Express Flash PCIe SSDs.
Trade NAND bit sales to Micron customers grew 36% in the second quarter, primarily as a result of a higher level of production from IM Flash Singapore quarter-over-quarter as that operation performed at its targeted level for substantially all of the second quarter. Our volume mix of MLC versus SLC NAND in the second quarter was consistent with the first quarter. Production costs per bit for trade NAND products decreased 18% in the second quarter compared to the prior quarter, primarily due to the higher production volumes on advanced technology nodes in the period.
Margins on trade NAND products decreased slightly in the second quarter. However, selling prices declined more than cost -- the cost reductions. Quarter to date for the third quarter, selling prices for tradeNAND products are down mid-20s compared to the average for the second quarter.
DRAM revenues have been fairly flat for the last 2 quarters as higher bit sales have been offset by decreases in selling prices across both periods. Specifically, in the second quarter, per-bit DRAM selling prices decreased 16% compared to the first quarter. This decrease was impacted by both market price declines and a somewhat lower mix of specialty products sold in the premium markets in the quarter. DRAM bit production and bit costs in the third quarter are expected to be down a couple of percent as we shift mix to maximize margins. Quarter to date for the third quarter, selling prices are relatively flat compared to the second quarter average.
Sales of SSDs, including NAND components sold to fabless SSD manufacturers, grew about 15% quarter-to-quarter as these devices continue to gain acceptance in the marketplace.
David M. Wong - Wells Fargo Securities, LLC, Research Division
On your supply agreement with Intel, can you tell us how long the supply agreement lasts for? And is it for fixed wafer prices? Or do the prices move with market NAND prices?
Ronald C. Foster
This is Ron, David. The supply agreement with Intel is set up so that it's a fixed amount of volume. And we have a profit margin built into the agreement, and it's set up to extend for a number of years into the total agreement period.
James Schneider - Goldman Sachs Group Inc., Research Division
That's helpful. And then looking out into next year, there seems to be a little controversy in terms of the NAND market and what the expectations for NAND bits -- bit growth in the industry will be, some people calling for as high as 70%, others calling for more like 50%. I was wondering if you could maybe weigh in on where you see that bit growth coming in for next year and what factors are going to drive that either higher or lower.
D. Mark Durcan
Sure. Our internal modeling is in the sort of 65% to 70% range.But there's -- obviously, there’s some guardrails around that. We don't have complete accuracy or precision relative to our information. What can drive that up and down, obviously, is mix of 3-level cell versus 2-level cell versus single-level cell. So product mix going to the market can have a significant impact on bit growth without necessarily having a significant impact on profitability or cost per wafer. We have seen some level of new capacity come -- certainly planning to come into the market, and we've seen that be what I would characterize as relatively measured and, in some cases, muted relative to initial plans. And so I think what we anticipate is something in the 65% to 70% range. And if the market's stronger, maybe it'll be a little bit higher. And if the market's weaker, maybe it'll be a little bit lower.