Micron to buy Qimonda? Why, and then what? Micron Technology, who seems always to be in the market to buy distressed DRAM companies (and sometimes does) appears to be back shopping and negotiating for Qimonda, who for the past 12-18 months has been perhaps the most fragile DRAM company, piling up tons of red ink and with only enough cash to sustain itself (at its current burn rate) for a few more quarters, according to analysts.

The headlines read, “Micron could buy Qimonda at 'fire sale' price, analyst says."

But why? For what purpose and to what end would Micron want to do this? Why would Qimonda shareholders (incl. 70% stakeholder, ex-parent Infineon) agree to let loose of its only begotten DRAM child, which once drove 50% of Infineon’s revenues…and sell it at a price that would allow Micron to financially capitalize on Qimonda’s stressed condition…when they could just liquidate it themselves and gain the surplus value for themselves?

Why would Micron want to buy Qimonda? What do they plan to do with it…the technology, the physical assets, the people, the IP, and existing cross-licensing agreements? If they expect that one fewer DRAM players in the business would benefit them by reducing competition, this would only help if Micron was willing to idle some of Qimonda’s capacity…Richmond (VA), Inotera in Taiwan, or Dresden…which idling would further hurt Micron’s gross margins (the depreciation charge stays even though the ownership changes…only now it shows up in Micron’s COGS). What would Micron do about the “recently-signed technology agreement that Qimonda has with Micron rival DRAM player, Elpida, who is arguably second only to Samsung in technology strength and financial stability? What has Micron done with its other DRAM acquisitions of Texas Instruments’ DRAM assets (1998) and Toshiba (2002)? Has Micron benefited from those acquisitions? For certain, the DRAM market did not become less competitive as a result of either the TI or Toshiba purchases...the 2001-03 DRAM downturn, and the present one starting late 2007, rival anything that went before in terms of financial damage…and Micron’s losses during the 2001-03 period were their worst ever.

Rambus License? Though Qimonda has recently begun production ramp up of 512M XDR DRAMs, and signed a revised and extended licensing agreement with Rambus, there is NO WAY that this license is transferable to Micron, whose pique at Rambus apparently knows no limits. Would Micron management bury the hatchet and come to an agreement with Rambus (and vice versa)? I will not hold my breath.

Add Additional Capacity?: Both the NAND and DRAM markets are awash with excess production capacity, and will be so for a while longer (maybe a long while), which is the proximate cause of today’s prices being way below costs.

When everyone is losing money, higher market share only means more losses, and Micron, in its recent analyst meeting in Sun Valley, swore off market share goals in lieu of better profitability. In that same meeting, we were reminded that Micron, for a few quarters in 2000, was the Number One DRAM maker…bigger than Samsung! But that was long ago, when their share price was in the $90 range vs. their recent 15-year low of $3.61 (closing at $4.65 on Friday, 9/26)

Indeed, it was not until Intel and Micron launched their quite successful IMFT NAND Flash venture in 2005-06, that Micron FINALLY was able to build-out unused capacity it had been carrying on its books at Lehi (since 1995!) and Manassas VA (since 2002). From their TI and Toshiba acquisitions, they still have all the fabs they acquired in play, running Micron products, except an empty fab shell fab in Texas which came with the TI acquisition that they sold immediately to Atmel as soon as they acquired TI’s assets. It would appear that Micron is philosophically averse to selling its fabs...and will find a way to fill them up, at whatever the cost, but maybe without a carefully thought out and financially viable plan. (They have TI’s TECH fab in Singapore, TI’s Avezzano fab in Italy, the former TI-Kawasaki Steel JV in Japan (KMT), and Toshiba’s Manassas fab in Virginia.) Micron's dispersed fab situation is entirely contrary that of the other DRAM and NAND flash players, who, with the exception on Samsung's DRAM/NAND fab in Austin, are highly concentrated in one place, one time zone...often sharing one plot of land.

Today, Micron (IMFT, or, specifically, IMF Singapore) also has a large (60K 300mm wafers/mo capacity) NAND Flash fab in Singapore, awaiting equipment, which facilitization they have had to postpone twice due to weak NAND prices and perhaps cash conservation.

Redundancies? When IBM Micro closed off its DRAM factories in Europe in the late 1990s, as a prelude to a full exit from the DRAM market, they had to pay huge penalties for the layoffs; something like $200-250K per laid-off employee. This should give pause to anyone who takes control of a manufacturing business in Europe, with the intent of shutting it down…it will cost a lot, and someone will have to pay.

Assimilation, again: Buying Qimonda would bring with it its own raft of ‘assimilation’ issues, which were substantial in both the cases of the TI acquisition, and even more so with Toshiba’s Manassas plants (see the Mummy’s Curse, page 4 of that DMR issue): process and roadmap reconciliation takes new CapEx, tools and time. If Micron is just going to unload the Qimonda fabs as-is, why would Qimonda shareholders let Micron gain any profit from such a sale, instead of doing it themselves? If Micron is NOT going to get the fabs, what’s the purpose in the acquisition? Certainly the status and future of the Inotera JV between Nanya and Qimonda needs to be clarified, since Nanya has moved on to a Micron technology development partnership, with its own JV fab being built (MeiYa). But resolving the Inotera issue does not in any way require Micron to buy all of Qimonda, only its stake in Inotera.

We can see how buying Qimonda on the cheap has some appeal; my closet is full of nice shirts, some still in their original packaging, that I could not pass up because they were on sale. But it is hard to see why Qimonda shareholders would not sell it off themselves in pieces, if necessary, at a maximum revenue basis, and it is equally hard to see what benefit Micron would expect to get from buying it. It would not raise DRAM prices, it would pose another assimilation problem for Micron, it would further drain their cash resources (and increasingly in DRAMs, Cash is King), or further dilute Micron stock (BTW, TI took a large fraction of payment for their DRAM Asset sale to Micron in Micron stock, which was also depressed in value at the time, then sold it in 1999-2001 at a huge profit…eventually more than doubling the apparent sale price at the time)

To Qimonda shareholders I would ask, “Why sell this asset at a price under book value and under what you might achieve if you liquidated it and then walked away?”

To Micron I would ask, “Can you honestly say that the TI and Toshiba acquisitions paid off in terms of profits, market standing…and Micron’s shareholder value?”… and, “Why bring another stack of issues to your plate, as if DRAM and NAND prices under water were not enough?” Add to this Samsung’s apparent “drive them into the sea” investment and competitive strategy. I would say Micron, and about everyone else who participates in the NAND flash or DRAM market these days, has a plate that is full-to-bursting, with survival and ‘profit restoration’ issues being paramount. Maybe the best thing for everyone is to let Qimonda find its own way to the grave; or, to better share the pain, since all DRAM makers are brothers-in-misery, have a DRAM-makers consortium collectively buy Qimonda, and ‘dispose of properly’, taking its fab capacity with it.

Author’s Confession: After a long stretch at Dataquest and as an independent memory market analyst, I worked at IBM Micro from 1995 until 1999 in memory marketing, and during that time, my opinion of Siemens-to-become-Infineon-to-eventually become Qimonda changed markedly for the better. Before I joined IBM, I saw from a relative distance a company which had some good skills, but could never make it work, and never make a profit. Siemens Semiconductor Group, of which about 35-50% was DRAMs, showed only two years of profit in the decade from 1983 to 1993. Unable to make their own 1Mb DRAM design work, they bought a technology license from then-DRAM market leader Toshiba and ran with it; at the 4M level, the much heralded Euro consortium of Philips (SRAM), STMicro (NVM) and Siemens (DRAM) fell apart (ST did OK, others went back to the drawing board or gave in), and Siemens DRAM supply line was filled with 4Mb die purchased from Mitsubishi and rebranded. Then at 16M, they aligned with IBM, shifted to trench cell DRAM, which I viewed as yet another dependency-due-to-internal failure. It was, but from then on, ‘Siemens’ made significant technical contributions to the DRAM design and roadmap, improving it from generation to generation, and ran with it until earlier this year, when they moved to an impressive Buried Word Line (BWL) DRAM architecture…so Team Infineon/Qimonda did their own 64M (0.25um…or 0.24um), 128M and 256M (0.2um, eventually down to 0.12um), 512M and 1Gb (down to 58nm = 0.058um), each generation requiring significant improvements in the cell to maintain low SER, high performance and high yield.

Siemens = Infineon = Qimonda survived in DRAMs far longer than either of the other two Triad partners, IBM exiting at a cost of about $2.5B in 1999, and Toshiba following a few years later with the sale of their DRAM operations and fabs to Micron in 2002.

If Qimonda disappears in the next year, it will not have been taken to the graveyard of DRAM makers for lack of technical skills, so much as business strategy mistakes, slowness to respond to market changes, and a mercilessly competitive DRAM marketplace. Of course, there in their DRAM grave, they would be in some of the best company in the history of the semiconductor (DRAM) industry, too: Intel, TI, and all Japanese, exc. now Elpida (was Hitachi + NEC, each was a DRAM market leader from the 1980s), Mostek and Inmos from the 1970s and 1980s, and a host of players-around-the-edges who dabbled and are now mostly forgotten.