Monday, February 28, 2011

WaferNEWS Watch: FORM shaping up; an AMAT mea culpa

One market analyst says probe card vendor FormFactor is "back on track" to retake market share in DRAM test. And another analyst wants a do-over for his previous bearish stance on AMAT.

FORM shaping up

FormFactor (FORM) has had its fits and starts in recent quarters, and some have even called for it to get busy trimming, and/or get busy selling. (One of our readers, meanwhile, believing this a far too simplistic synopsis, points out that not only did the company recognize the need to trim manufacturing, but its fab ops issues are more complex, concerning improper utilization and burn rates.)

But where there's a trough, there's a coming upswing, and chance to rebalance the playing field. Citing FORM's talk at his outfit's recent tech conference, Patrick Ho of Stifel Nicolaus feels the company is "back on track" with its operating model, as long as it can win back share in DRAM, and longer-term could push into NAND flash and system-on-chip designs. Initial quarters under new CEO (and former AMAT exec) Thomas St. Dennis have had "some hiccups," he notes, but 1H11 could spell the bottom for the company.

AMAT mulligan?

On the heels of strong Applied Materials' (AMAT) fiscal 1Q11 numbers, particularly in solar, we've got a Mea Culpa sighting: Piper Jaffray's Gus Richard says he shouldn't have downgraded AMAT back in July when it was deep-sixing its thin-film solar biz. In fact, estimates for 2011 and 2012 are too low, he now says, with backlogs well in excess of $3B and even approaching $4B by year's end.

Key to overall performance, though, is the company's bread-and-butter biz of tools for advanced semiconductor manufacturing, and here the shift to 28nm manufacturing will be the proving ground. "We believe yields at 28nm are going to be challenging as older gate stacks do not offer sufficient process latitude, high-k/metal gate processes are not mature at 28nm, and foundries move to double patterning is likely to be challenging as well," he writes. A rocky 28nm node transition will mean adding more wafer capacity, which actually is good news for suppliers -- and AMAT in particular, as the 800lb gorilla.

Dr. Burn Lin of TSMC compares EUV and MEB

Dr. Burn Lin, VP of TSMC, was the keynote speaker at KLA-Tencor's annual Litho Users' Group forum. Among the topics he covered was a comparison between multi-ebeam and EUV lithography for future nodes. Lin first noted that economics will drive the decision at the 16nm node - whether to use EUVL or MEB ML2. However, at 8nm, EUVL has better resolution, MEEF, DOF, and overlay margin, while MEB has the potential for better CDU and OL accuracy.

EUV needs more source power to compensate for throughput loss per node, which breaks down as follows: a 2x loss due to shot noise, and a 2x loss due to the need for more mirrors at higher NA lithography. MEB needs either more parallelism or source brightness to compensate for its throughput losses per advancing node; MEB's throughput losses per node break down like this: MEB's problems are due to a 2x loss because of the huge volume of data generated by the process (e.g., the data rate is >7.5Gbps/beam - in the MAPPER tool, for example, a single electron source is split into 13,000 Gaussian beams), and there is a 2x loss due to shot noise.

The last two points are probably obvious: EUVL has a high cost and MEB is less developed as a technology. (DV)

Friday, February 25, 2011

Editorial coverage at SPIE Advanced Lithography

I'll (Debra Vogler) be doing podcast interviews at the SPIE Advanced Lithography conference next week. You'll see them all in the Daily Pulse!

Podcast coverage at Strategies in Light

Coverage at the Strategies in Light conference includes podcast interviews I (Debra Vogler, senior technical editor) did with Chris Moore/Semilab AMS, Jeff Desroches/ATMI, Mike Plisinski/Rudolph, Ravi Kanjolia/SAFC Hitech, and Thomas Uhrmann/EVG. Some have already been posted - you can hear Jeff Desroches' interview at Additional podcasts are at

Thursday, February 24, 2011

WaferNEWS Watch: 450mm is coming, but after EUV and TSVs

Barclays analyst CJ Muse updates his views on a 450mm wafer-size transition, following TSMC's stated switchover plans. His take: 450mm will happen by or before 2018, but other key chip manufacturing transitions will have to come first -- namely through-silicon vias (TSV) and EUV lithography.

After an initial push by the self-appointed 450mm Big 3 (Intel, Samsung, TSMC) asking for a 2012 pilot line, things quieted down (at least publicly) due in part to vociferous reception from suppliers, and then the global and industry downturns. Now, with recent news of Intel's forthcoming 450mm-capable D1X facility in Oregon and its planned new AZ fab (also 450mm capable), and TSMC's stated 450mm schedule (pilot line in 2015-2016, ramping production in 2015-2016), industry chatter about 450mm has "begun to percolate" again, he notes. Equipment makers, long resistant to the idea of 450mm, have started commenting publicly to the eventuality of a transition. (The 300mm wafer-size upgrade increased processing surface area by 2.25x, but they only could get roughly 40% more for 300mm tools, Muse points out.)

But capital intensity for leading-edge semiconductor manufacturing is on the rise, and there are more pressing transitions that will need to be addressed first, he says. First, expand use of through-silicon vias (TSV). Then, incorporate EUV lithography for finer features where immersion (and its associated tweaks) can no longer go, presumably sometime soon after 20nm. And being third-fiddle to TSV and EUV will actually help 450mm's cause, because 450mm will help chipmakers reduce their rising manufacturing costs, and rising capital intensity translates into more business for tool suppliers so they'll have a better economic foothold for the R&D.

"We will likely see a chicken and egg game, but we do expect chipmakers to help support the tool development efforts with equipment companies, at the same time, sharing some of the higher dollars received in the current golden era of capital intensity," Muse writes. His back-of-the-envelope calculations (a $40B equipment market, 15% R&D spend on it, and a seven-year transition period) suggest tool makers will spend roughly $6B on 450mm, which they "can rather easily come up with," he says, as long as chipmakers add their own funding "in the order of billions" to help the transition.

Muse also lays odds on who will benefit most, and least, from the 450mm future. Adding more wafer surface area means that beam-tool process steps (i.e. litho, implant, and metrology) will need to be improved to maintain their throughput, so suppliers in these fields should see a boost in capital spending. Areas that won't see much help from 450mm would be in vacuum-based process equipment.

"The ROI argument still exists for 450mm, as has existed for EUV," Muse sums up. "But if [chipmakers] all want it in no uncertain terms, and are willing put money to make it happen, it will happen."

Tuesday, February 22, 2011

LED manufacturing workshop at Strategies in Light

At a pre-conference workshop (Strategies in Light; 2/22-24/11; Santa Clara, CA), industry experts discussed manufacturing issues and strategies for LEDs. In the coming days, watch for my podcast interviews with workshop presenters Chris Moore/Semilab AMS, Mike Plisinski/Rudolph Technologies, Thomas Uhrmann/EV Group, and Ravi Kanjolia/SAFC Hitech. They'll be posted to and featured in the Daily Pulse and WaferNEWS.

Monday, February 21, 2011

Strategies in Light conference coverage this week

Just a reminder that Electroiq (i.e., Solid State Technology, Photovoltaics World, Small Times, and Advanced Packaging) editors will be covering the Strategies in Light conference this week (Santa Clara Convention Center).

Next week - we'll be covering the SPIE Advanced Lithography Conference.


Friday, February 18, 2011

WaferNEWS Watch: War of the ASPs at Mobile World Congress

Eyeing the Mobile World Congress through a semiconductor industry lens, Deutsche Bank analyst Ross Seymore sees the battleground of tablets and operating systems shifting from basebands to application processors, and ultimately boiling down to integration and prices.

Nokia: MeeGo to Windows. Android's growing popularity in smartphones and tablets was in the spotlight at the recent Mobile World Congress, but what created buzz among semiconductor vendors was Nokia's announced shift to Windows from the Intel-backed Linux-based MeeGo platform. "Most chip vendors at MWC were trying to put on a positive spin on this change, but we found a pervasive sense of uncertainty as to the chip procurement implications," writes Seymore in a research note. Perceived winners in this shift: Qualcomm (incumbent in the Windows ecosystem) and possibly Texas Instruments (incumbent at Nokia in basebands). Intel and Broadcom would be potential losers amid the uncertainty surrounding the OS shift, ramp timing, and opex requirements, he speculates.

War of the app processors. The emergence and user embrace of smart phones and tablets (netbooks, we hardly knew ye!) is shifting the battle of silicon from basebands to application processors, and there's a "core war" brewing among numerous (at least 10) vendors trying to differentiate on how many ARM cores (single, dual, quad-core) and frequencies (1GHz-2.5GHz) they can offer.

It's all about price and integration. Ultimately, end-users generally don't care about the guts and components of their smartphone or tablet unless it impacts usability. So, inevitably, the key advantage for mobile chip components will be about price, and favor those who can leverage cost-cutting benefits of integration (baseband, connectivity, etc.)

Room enough for everyone? Bottom line, the markets for smartphones and tablets are probably big enough that there's significant growth potential for chip vendors, but competition will be fierce and even intensify, from basebands to application processors to connectivity. Once handset vendors whittle down their OS choices, look for technical differentiation to give way to pricing pressures, and the winning OEMS will be those who can offer various wireless silicon solutions that can be integrated to lower silicon costs. (Seymore's looking at you, QCOM, BRCM, and MXIM.)

Thursday, February 17, 2011

Strategies in Light conference coverage

Those of you interested in LED manufacturing issues and strategies will want to attend the Strategies in Light conference next week (2/22-2/24/11, Santa Clara Convention Center; I'll be covering the event and our editor-in-chief, Pete Singer, will be moderating a workshop on Tuesday, 2/22, 8:00AM-noon. I have limited time available in my schedule, but if anyone is interested in doing a podcast (audio-only) interview, please contact me at to see if arrangements can be made. (Debra Vogler, Sr. Techical Editor)

Monday, February 14, 2011

WaferNEWS Watch: Advice for MENT and EDA: Fight and pray

Late last summer Carl Icahn raised his ownership stake in Mentor Graphics achingly close to the 15% threshold trigger of the company's newly enacted "poison pill" amendment, set down just two months earlier in what was viewed as a preemptory response to just such a shareholder maneuver.

And now we see why: Icahn now says MENT should put itself up for sale, and he's gathering forces for a potential proxy battle, ostensibly at least in part due to the company's sudden acceleration of its shareholders meeting date. "At the very least [Mentor Graphics] should be put up for sale and see what the shareholders want to do with it," he told CNBC. "The company is substantially undervalued versus its peers," added Donald Drapkin of Casablanca Capital, a 5.48% stakeholder that also is suggesting a new slate of board members. "Management's done nothing to promote shareholder value, they've just been just sitting on their hands [...] It's just a sleepy company run like a country club."

We asked EDA market watcher Gary Smith of Gary Smith EDA what he thought of the situation. Antitrust risks would probably keep EDA rivals Synopsys and Cadence out of the running for M&A, as the fragmented MENT assets would be "far less valuable than the whole," he told SST. A more logical fit would be with a mechanical vendor (e.g. Dassault or PTI), but a true inflection point of system design automation is still years away -- it "probably won't happen until the next decade so the acquisition wouldn't reap benefits for quite a while." From an engineer/user's perspective, acquisition by any other type of firm would mean a breakup of MENT, loss of leadership, perhaps "a slowdown in DFM R&D," and even a sectorwide EDA breakup "as we are seeing in the embedded software design market," Smith thinks.

His recommendations: "I think Mentor should remain independent so they can carry out their present market strategy without interference." And his advice for all the EDA companies: for MENT, "fight;" for SNPS, "pray"; and CDNS, "pray harder."

Friday, February 11, 2011

NanoArt 2011 - open to all artists and scientists (18 yrs or older)

The NanoArt 2011 international online competition was announced today. It's open to all artists and scientists 18 years of age and older. Basically, contestants are given 3 hi-res monochromatic electron scans of nanosculptures. Participants have to alter the provided image(s) in any artistic way to finish the artistic-scientific process and create NanoArt works. Artists and scientists may also use their own images as long as these visualize micro or nanostructures. Go to for more details. (DV)

Monday, February 7, 2011

WaferNEWS Watch: Popping the MOCVD bubble for LEDs

Market analysts at our sister organization Strategies Unlimited recently took a look at the rapidly rising sales of MOCVD reactors for LED production, whether it's a true market bubble, and what it means to both sides of the supply chain -- who wins and who loses. Turns out it's a bit more complicated than at first glance.

"There have never been so many orders in the history of MOCVD," asserts Tom Hausken, director of components practice. For an idea of just how wild & wooly this sector has become, consider:
  • A rumor from last month, that Golden Concord Holdings (Hong Kong) sought to purchase 500 reactors as part of a new $2.5 billion investment in LEDs;
  • Several companies have orders for >100 reactors;
  • There are two primary supplier beneficiaries -- Aixtron and Veeco -- who are "working like crazy" to deliver tools, he says.
Seems easy to conclude that there's way too many tools in the pipeline. Hausken and S-U peers drew up the following chart to calculate what they believe the world needs to meet near-term LED production (not counting normal excess or competition), vs. what the devicemakers seem to be asking for. (Note that this doesn't include assumptions about what might actually get delivered, accepted, and put into production...)

Estimated LED MOCVD reactor shipments. (Source: Strategies Unlimited)

Clearly there is a big mismatch, but the question is -- who stands to win, and who will lose? Hausken reveals all in an article for SST's sister magazine LEDs Magazine, but suffice to say that there are a lot of winners, from LED end-users (excess capacity means lower prices) to MOCVD reactor vendors (cash windfall). And even China, which is pushing this entire envelope with subsidies, generally comes out ahead due to technology and investment infusions, much like what happened with the nation's solar PV push.

On the other hand, in the event of an LED glut lower-tier LED suppliers probably will suffer, Hausken notes. And investors "may get stuck with some expensive paperweights."