MarketWatch Commentary

At Smith we buy and sell memory every day. We handle DRAM, SRAM, Graphic RAM, Nand and NOR Flash. As the commodity manager for memory I can see supply and demand across components and across our global offices, which gives me a pretty complete world view of the memory market. Here's what I see for 2010:

• 2010 should see constant tightness and long lead times on most types of memory all year. The main two reasons are 1.) Due to the credit crunch, chip manufacturers reinvested less in 2009 to grow production capacity than in previous years. Although 2010 will see a return to high re-investment in this area, (a predicted $8-10B worldwide) it will not have an effect on production capacity until at least 4Q10. 2.) The wide array of products now available from chip makers makes it harder for them to forecast accurately, which is exacerbated by the continued introduction and successes of new products and fads or unforeseen failure or obsolescence of technology in the retail market; which increase or decrease expected demand accordingly.

• Over the last year almost every chip manufacturer has gone from a commodity "build-it-and-they-will-come" plan - where they build a bit of everything and hope the customer will come calling - to a "build to forecast, but forecast accurately" plan. That means fewer parts in excess on the end user shelves, longer lead times and stricter ASPs. Industrial and extended temps parts, odd cas latencies, slower speeds, and odd packages (BGA for DDR1 and SDRAM, since TSOP is the norm) will be the casualties in this plan, and become more and more scarce as we move further into the year.

I see these factors affecting the market of the various memory types in 2010 differently.

• Personal computer demand around the world continues to climb, especially in China and developing countries. Though the demand for desktop PCs is growing at a lesser rate than it did during 2009, the higher rate of laptop demand more than makes up for it. That, of course, means demand for DDR -- a mix of DDR2 and DDR3.

• At the end of 2009, DDR3 was used in about 25% of the new computers - PCs, laptops and servers combined - coming out. By the end of 2010, it should be up to 75-80%. While DDR3 will not have many other usages outside of this area, it will be the mainstream technology for most new PCs; PC3-10500( DDR3-1333) being the principal speed.

• Look for server demand to continue to pick up in 2010; many companies who have put off upgrading or re-invigorating infrastructure for the last couple of years really need to do so now, as software demands upon a network are at an all time high. New servers developed in late 2009 and 2010 will use DDR3 memory. However, DDR2 production cutbacks at the chip manufacturer level should make DDR2 server modules and components harder to come by. We've already seen open market demand increase and pricing up 5% in through January end.

• Already, DDR3 delivery is behind demand and that should continue during the first half of 2010. The estimate is DDR3 overall will be 2-4% short worldwide for the entire year.

• The introduction of memory hungry Windows 7 will also help increase the GB/unit in the PC arena.

• As I noted earlier, the rise of DDR3 will mean the demise of DDR2. Over 2010 it will drop from being in about 75% of the new computers to about 20-25%. But again, production cutbacks on the chip manufacturer level should make DDR2 availability tight.

• SDRAM PC133 demand should decrease some in 2010 vs. 2009, but with wafer start decreases and EOL-ing of highly popular configurations or accepted generation revisions, availability will be limited. Lead times and pricing will likely increase throughout the year from all MNFs. Already Samsung and Micron have plans to EOL their very popular 32x16 SDRAM, which should start causing some problems come March. So, though demand will be down in 2010, pricing and should stay high and lead times long all year.

• The Graphic RAM market is still in shortage due to the demise of Qimonda, but it is beginning to right itself. There are still shortages and long lead times and we expect to see that continue until at least mid-2010. Most MNFs however, have EOL-ed their multiple technology offerings to concentrate on gDDR3, which pairs best with current motherboard/PC needs. So builders must be highly aware of lifetime estimates in this area if creating new products.

• Nand Flash pricing has steadied due to production cutbacks and rising demand, a trend which should continue throughout 2010.

• Lower density NAND products, 1Gb and below, will all soon be going EOL one by one - starting at the smallest - and we could even see 2Gb & 4Gb go that route by the end of the year if chip production capacity does not increase. The continued introduction of new products using NAND as the main data storage also contributes to tightness on mid-range densities (8Gb-64Gb) and longer lead times. Fewer wafer starts on SLC product is causing longer lead times and tightness for NAND, and possible trouble in the industrial, aerospace, military and automotive fields.

• For NOR Flash, usages are continually shrinking, but it will continue to be hard to find in the open market. Continued monetary uncertainty at Spansion, talk of Samsung stopping production, and rumors of a Micron/Numonyx merger make the NOR market unpredictable.

Thus, I expect in 2010, memory chip makers will likely: Promote DDR3 for computing; cutback on general DDR1 and SDRAM wafer starts; level off DDR2; and make only enough NAND Flash to fulfill forecasted orders, promoting higher densities over smaller densities.

Comment (0)

Last week was the SEMI 33rd annual Industry Strategy Symposium (ISS); SEMI is a leading global semiconductor industry association.  At this conference industry analysts and economists alike presented research and forecasts pointing to the significant recovery that the semiconductor industry is expected to face in 2010.  However, with these rosey forecasts came a number of cautionary words that the industry supply chain does not appear to be poised to enjoy all of the benefits of this positive projection.

The semiconductor industry has been faring better than many industries based on 4Q09 reports.  The 1H10 forecasts continue to show the tendency for pent up demand to be strong from consumer and corporate sales alike.  However, with the loss of almost two years of CAPEX investments to support fabs, the growth that we're engaging now may prove to be a tough challenge for the industry because of these "missed investments," as underscored by Bill McClean, President of IC Insights.

According to McClean, IC Insights research and data show that despite the "worst recession in 63 years (since 1946) in 2009, flat PC unit shipments were quite an amazing accomplishment. Moreover, cellphone unit shipments were down only 5%.  Now consider that both PC and cellphone unit volume shipments are forecast to increase at double-digit rates in 2010."  This begs the following questions from the recent McClean Report: "What effect will this have on the IC industry? Are we on the cusp of an all-out IC market boom for 2010?"

IC CAGRs are forecasted in the 9% range; ASPs and revenue are set to rise across the board for memory, even for DRAM;  increased demand is expected for PCs; and emerging economies' consumers are hungry for electronics.  Underscoring this positive reality is the recently released, strong, 9%  increase in North American-based semiconductor equipment manufacturers book-to-bill ratios, now at 1.03 for December, holding onto a 6-month rise. 

The 2010 forecasts are no longer just New Year's dreams.  While exciting and welcomed news, these forecasts pose interesting challenges for the industry that must now grapple with being able to meet what TSMC calls "urgent recent increases in customer demand."  TSMC has begun both new construction and capacity expansions at their fabs in Taiwan, increased R&D spending for 2010 by 25%, and plan to hire 3,000 new staff, "primarily engineers." 

Those companies who remain from the last round of survival of the fittest now have a new, positive battle ahead.  Economic indicators are up 1.1% for December 2009 and point to economic growth this Spring.  Are we really ready?!

Comment (0)

November was another active month in memory trading, with prices on most types of products ending the month higher than they began.

DDR2 pricing increased the first week of November, flattened out the second and third  weeks, and decreased slightly the last week. 667 materials  decreased the most, due to the falling demand of the slower speed, while 800 products held their value much better. The slight degradation of pricing can be traced to a couple of factors. The first is simple end-of-the-year softening, when OEMs are reluctant to hold inventory over from calendar year to calendar year and most of the products are bought for the holiday builds. The second is the re-entering into the market in greater amounts of the off-spec/non-major brands, like the Taiwanese manufacturers. Also in this mix is the production lost when Qimonda sold Inotera; those buyers are now shipping product from that source into the eTT/uTT and upgrade markets. The greater availability of the off brand products have given upgrade consumers a cheaper solution than the 5 majors they were basically forced to buy during the last 6 months.

DDR3 pricing saw an increase in pricing over the last month, and has flattened on most products during the US holidays. 1333 speeds are the high runners and have been lingering at a 15% premium over their 1066 counterparts. Module demand continues to increase on the PC side, as DDR3 continues to slowly replace DDR2 as the standard memory for new computers and servers.  More discrete chip inventory has found its way into the open market, as production increases at the chip manufacturers enable chips to get into the hands of the generic module makers.

DDR1 and SDRAM/PC133 continue to be harder to get as the lower densities and odd configurations head towards EOL and the manufacturing output capacity changes to more generic and newer technologies with broader ranges of use. Already we've seen popular parts configurations like 32x16 PC133 phased out without replacement from some of the major chip makers, and industrial or extended temperature tested chips are at the made-to-order stage, causing long lead times from chip makers.

Since Qimonda dissolved, Graphic DDR has been very tight, having only two viable options as vendors left, both of whom were already at full capacity and still running short of demand. However, with the normal slowing down demand of the year's end and the introduction into the market of more capacity from 3^rd parties, much of the delivery issues are working themselves out. There are still some common configurations that are tight and getting premium pricing, but that also should work itself out heading into the New Year.

NAND Flash has continued to see price increases and lead time extension on most MLC and SLC large block items. November saw a 10% price increase on several sizes 8Gb-64Gb, and industrial temperature items are garnering long lead times from the manufacturers. Demand also continues to rise as NAND finds it way into new technologies and hybrid ASICs and controllers, but with capacity running as high as it can, further price hikes and longer lead times are likely to occur for the next qtr.

Overall, December is shaping up to be as active as November.  That is, at least for the next couple of weeks - until holiday scheduling has its predictable calming effect on memory trading.

Comment (0)

DRAM pricing has been rising quickly over the past weeks due to the combined influence of a number of factors: 

• an uptick in demand from larger than expected need in China;
• the introduction of Windows 7; 
• recent and upcoming Asian holidays;
• new forecasting methodologies at the chip manufacturing level; and
• the expansion of DDR3 production at the expense of DDR2 production.

This has resulted in higher contract prices all around and a large rise in open market pricing - especially on DDR2-800 (PC2-6400) items. For DDR2-800, contract prices have increased about 20-25% and open market pricing approximately 30% since mid-August, with most of that increase occurring since September 1.

Riding on the heels of DDR2, DDR3 prices, after softening in late August, have rebounded by about 3-5% since September 1.  Upward pressure on both DDR2 and DDR3 pricing is expected to continue in the coming weeks. 

Comment (0)

Of course we all know the important news out of Japan for 2Q09, the first GDP uptick in five quarters.  That's tremendous and portends serious positives for the APAC region as well as global economic recovery momentum.  But there's more to the news out of Japan: as part of the government's stimulus efforts, increases in public spending and use of public funds to bolster those private companies worst effected by the global recession have made cash available.  With an eye to increased consumer spending in Japan as well as improved export numbers, up 6.5% for 2Q09 QoQ (Credit Suisse Japan Macro Flash 17 Aug 2009), the time to invest in local companies is well aligned and is inline to support the positive trend for recovery.

At the center of upstream semiconductor news is Japan with the now-back-on-again merger of Renesas and NEC Electronic into "the world's third-largest semiconductor maker after Intel Corp [...] and Samsung Electronics Co [...]," according to Reuters recent updates.  One of the hurdles though is the amount of debt refinancing necessary and the issue of government support to the tune of roughly 200 billion yen.  This on top of Japan's recent pledge to help Elpida Memory Inc, also severely hurt by the global economic crisis. 

While MarketWatch Commentary tends to not focus on individual companies, the merger, now seen as likely to be completed in early 2010, signals important changes to the supply chain and the competitive landscape, particularly for MCU production.  As for the impact on the industry, the increase in competition is important and will be interesting to watch, particularly as chip makers were hit hard by the global crisis.  There will be an uphill period for the new merged company, likely to fall under Renesas's name, as a result of redundancies in product, design, and then the restructuring post-merger (see HERE for a pro-con summary from EETimes).

While the impact on the global supply chain is likely to be delayed until the restructuring is completed, the opportunities for Japan to restake it's claim to semiconductor manufacturing are ripe.  Along with other trends diversifying into solar PV and the help to Elpida, Japan will be a country to watch as a new, global economy awakens.

Watch for a review of the new geo-economic landscape for the semi industry in the upcoming  MarketWatch Quarterly.

Comment (0)

It was the fabs' dramatic CAPEX reductions that truly ignited my cause for concern many quarters ago (as discussed in this MarketWatch Quarterly article) and made even more worrisome when utilization levels went below 50%.  It is now with a sigh of relief that the situation is truly coming full circle (see this MarketWatch Commentary post also).  Recently, fabs have been posting forward looking schedules to begin the ramp up of 40nm process technology at the end of 2009 - see this early MarketWatch Quarterly article for details on this technology.  This puts 40nm on track for 2010 release and for 2011 as a main process technology.

While overall global capacity even at the 40nm-scale, at this point, is not expected to increase given the continued muted, although rebounding, consumer demand, the cost savings for the fabs by moving to 50nm and then on to 40nm are significant.  DigiTimes 8/17/09 reported Micron's estimates of lowered production costs by 50% for 50nm and an additional 30% for further migration to 40nm.

Meanwhile, TSMC and UMC continue to run at 80% utilization levels again - more good news given the extreme lows just a few months ago.  Both offer that order visibility is good and holding in positive territories.  UMC, as reported by DigiTimes 8/13/09, has also released its move to 40nm process nodes during the end of 2009, adding that CAPEX levels are up and new equipment to further support the next generation technology is purchased.

Others in the Taiwan DRAM production arena are actively seeking funding to increase their production levels.  DigiTimes 7/20/09 speculated that this is in anticipation of DRAM price rebounds during 2H09 - let's hope they're right...

Comment (0)

As we become well entrenched in the halfway mark for 2009, the promises of 2H09 showing relief and recovery are happily bearing fruit. 

News from the wire and in the markets is showing not only truly healthy inventory levels for memory, but even more importantly, strengthening demand to the point that ASP increases are holding for both DRAM and NAND.  In DRAM, DDR2 pricing has been rising as supply is tightening with back to school orders and shorts helping to support 10% rises in DRAM pricing over at least the immediate-short term (see this Circuits Assembly article for publically available details).  Also, on 5 Aug. 2009, DigiTimes cited DRAMeXchange as stating that "DRAM revenues grew 27.1% sequentially to US$4.04 billion in the second quarter of 2009, when the average price at the contract market went up by 23% [...] which attributed the price rally to suppliers' production cutbacks and PC OEMs replenishing inventories."

NAND is also doing well, particularly flash for cell phones.  We've been watching the cell phone market for a while, see also the recent MarketWatch Quarterly here and here, so while this doesn't come as a surprise to regular readers, it's certainly welcome news.  The best part of this news is iSuppli's 2008-2013 forecast of a sixfold increase in global sales.  Wow.  Even more astounding is their prediction of a CAGR of 41.4% for the 2008-2013 period, as stated here).  All this based on the nearly fivefold increase in NAND storage amounts in the average cell phone from 1Gb to almost 5.8Gb for the same period.

But wait, there's more!  It's not just memory that's having a good run in the markets, some TFT LCDs are also showing strengthening revenue numbers, according to Circuits Assembly, "large-area (10" and larger) TFT LCD shipments reached 130 million units in the second quarter, up 10% year-over-year and 42% sequentially, says DisplaySearch."  Furthermore, "'We expect that strong demand will push panel shipments to another record high in the third quarter. However, strong panel shipments in some cases are due to downstream players piling up inventory, so we believe it will be necessary to watch the supply chain inventory situation closely at the end of the third quarter,' said David Hsieh, vice president of DisplaySearch."

All interesting and good news with some legs that ought to carry us forward.

Comment (0)

Wednesday iSuppli presented an interesting sector briefing  on mobile memory based on their tear downs of handsets performed over the past couple of years. 

Mark DeVoss, Senior Analyst for Mobile and Emerging Memory Technology, reported that mobile devices are certainly an important growth area, showing a compound annual growth rate (CAGR) forecast from 2008 to 2013 of 4.6%.  Given the recent year, and the rebound still to come, that is a good number.

In concert with the broader mobile sector outlook, of particular focus and interest are the data that inform the trajectory of the components that comprise the mobile handsets.  While the mobile sector itself is certainly a driver, the component architecture will tell us more about the direct impact and forecast for subsectors within the semiconductor supply chain.

To summarize the iSuppli findings, while NOR memory dominated the 2008 mobile handset component market, the trend is obvious and steep in the movement away from NOR and pSRAM and towards NAND and DRAM solution sets.  According to DeVoss, data supporting this trajectory are found in the declining ASPs and flat megabyte shipment levels for NOR and pSRAM, while there is an increase in NAND and DRAM megabytes shipped over the same period.  Furthermore, as DeVoss presented, the NOR versus NAND based mobile handset units from 2008 to 2009 show a striking trajectory:

	2008 Total Percentage	2009 Total Percentage
NAND	47%	81%
NOR	53%	19%

A final point, though not the only other one raised, is the impact of memory on the bill of materials (BOM) for mobile handsets.  As DeVoss notes, memory is having a lessening impact on BOM for mobile handsets as prices have decreased per megabyte, equating to a lower memory ASP overall.  Meanwhile, other features of the handset, such as display, graphics, and similar types of "touchy-feely" differentiators, are picked up by consumers as being increasingly important and are increasingly costly compared to memory.

iSuppli finds that NOR and xRAM based solutions are loosing out to NAND + mobile DRAM (mDRAM), multi-chip package (MCP), and package on package (PoP) solutions for the mobile handset sector. 

Interesting findings and an interesting report, as always, from iSuppli.  We'll be watching how these forecasts continue to play out in memory, the mobile sector, and as possible trends across the industry.

Comment (0)

Mobile internet devices (MIDs), smartphones and netbooks are proving to be true drivers in today's tough market.  That says a lot about these types of devices' appeal and future.  It also speaks volumes to both the components' and the end-products' effects on the electronics supply chain. 

In the middle of the month, DigiTimes 4/15/09, released some good news about 3Q09 foundry sector growth to come because of orders from netbook and 3G handset makers.  Despite Nokia and Sony-Ericsson's slumping 1Q09 earnings calls, drilling down into the MID, smartphone and 3G subsector, we come to a likely source of the mobile market driver and the better news.

A great example of such a source is the SanDisk earnings call transcript from 4/21/09.  Herein we find some important information underscoring the positive impact that MIDs and their product cousins are having along the supply chain (check out this ZDNet blog for more commentary).  According to Dr. Eli Harari, Chairman and CEO of SanDisk:

In the second half of this year, we [at SanDisk] expect demand for NAND to continue to grow particularly for mobile and portable computing platforms and this should hopefully absorb the industry supply growth projected for second half. [...]

As for demand creation, I believe that the handset business is being transformed on a scale similar to that which the web experienced in its early days and this has far-reaching implications for our mobile storage business.

That's great for the component specific side of the coin, but what about the rest of the mobile device? The mobile OEMs are also transforming, now favoring ‘insourcing' as bell weathers like Nokia reverse course to reduce cost and improve profitability.  According to a report by iSuppli's senior analyst, Jeffrey Wu, "Nokia in 2008 decreased the percentage of its outsourced manufacture volume to 17.1 percent, down from 21.5 percent in 2007.  This reflects a larger trend in the mobile-handset supply chain."

iSuppli's Wu has a new white paper that looks more closely at "the fatal pitfalls in wireless handset outsourcing."  This is an important read because the relationship between OEMs and CMs is certainly changing and the supply chains are changed by the present market forces.  These types of supply chain changes are also explored in more detail here.  Certainly, it's not business as usual - as if we didn't already know that.  But knowing why is different than knowing what.

It'll be interesting to see what the decrease in outsourcing will mean to the ODM-EMS turf wars as we continue to walk through the new territory of 2009.

Comment (0)

Memory is back in the news yesterday and today as analysts of all flavors and chip manufacturers alike are jittery, again. Yes, you guessed it, it's that old inventory problem.

But what of all that good news around how low utilization rates kept supply healthier, sudden ASP rises along with big buys just the other week, and news of increased visibility and some ramp ups in production? Oh yeah, right, as financial traders say, ‘even a dead cat will bounce if it falls from a great height.'  Granted, this means that I'm offering that there is not a reversal as we'd all hoped, and that I'll join the growing crowd in sticking my neck out and say that inventory issues are still severe and not likely to be overcome in the end (yes, not just the short-run, but in the end!).

As iSuppli analysts offered yesterday, "Due to long-lasting glut of DRAM, the imbalance between supply and demand is too great for this market to recover to profitability any time soon."

Furthermore, and critically for those companies reporting and forecasting for 2H09, iSuppli is warranted in their caution that: "Unless prices increase by more than 200 percent, cash losses will persist for these [DRAM] Taiwanese suppliers. [...] The industry needs a dramatic price recovery of a few hundred percentage points to make any kind of impact."

In case news from analysts alone isn't bad enough, Samsung and Hynix news hit the deck early this morning on their websites. EETimes reports today that Samsung "posted a 72 percent drop in quarterly profit and said it was too early to call a recovery in demand or prices."

Well, you might retort, Samsung is a large company and surely their profit problems don't stem from memory chips alone. True, true, but I never like it when the executives say things like Robert Yi, head of Samsung's investor relations, do. According to EETimes, Yi stated that "a sharp recovery in the memory chip or LCD markets was unlikely, [...] the company forecast prices of its mainstay DRAM chips would rise in the low single digits this quarter, but [...] that oversupply would stifle big price rises." This just doesn't sound like it's going to get near the 200 point correction iSuppli and similar analysts say are needed for profitability to return.

Oh, and don't even look to Hynix or other chip manufacturing brethren for better news. No reason to beat the ‘dead cat' on more bad DRAM news, but even Hynix's NAND numbers are down 4% QoQ even with an ASP increase of 10%, as reported in EETimes. Sure, NAND is better off than DRAM, but with demand sparse economies, recovery is still pegged to consumer movement.

So, Samsung's release WAS the good news, after all, since they got the Apple NAND order their numbers could have been worse... TGIF!

Comment (0)