How many MOCVD reactors is too many?

We took a look recently at the bubble in sales of MOCVD reactors for LED production. We set out to sort out who wins and who loses. It turned out more complicated than we first thought.

There have never been so many orders in the history of MOCVD. Just to give you an idea of the scale, in January there was a rumor that Golden Concord Holdings in Hong Kong wants to purchase 500 reactors as part of a new $2.5 billion investment in LEDs. Several companies have orders to buy over 100 reactors each. Aixtron and Veeco are working like crazy to deliver them.

But it’s too many. The figure below compares what we think the world needs to meet near-term LED production to what the world seems to be asking for. Since then, Barclays Capital raised its estimate of MOCVD shipments for 2011 to 900, from 800. While we may disagree about what actually may be delivered, much less what is actually brought into production, this much is sure: there is a big mismatch. Our question is: who wins and who loses?



I suggest that you read the full article in LEDs Magazine for the whole story. In the meantime, suffice it to say that there will be a lot of winners: end-users of LEDs, China, and the MOCVD reactor vendors, to name a few. Lower-tier LED suppliers may feel the most pressure. And yes, a few investors may get stuck with expensive paperweights.

Paying for the solar market

I don’t usually venture into solar energy discussions, even though it is also an opto technology. For one thing, it depends a lot on policy decisions and I've been there, done that once before. And there's already plenty written elsewhere. But it's worth pointing out Vinod Khosla’s recent posting on the requirements of investing in solar.

Khosla’s piece is long with detail, but he basically says that startups have to “be competitive with silicon cells at thin film costs or be competitive with III-V cells (well over 20 percent) at silicon costs. Then you have a 50/50 chance of making it. But a billion dollars of capital and billion dollars of debt will be hard to pay off.” A lot of is just basic market sense, and that's exactly the point.
Khosla echoes more or less what we have seen in solar for years. Strategies Unlimited followed the solar industry for decades while it had steady 25+% compound annual growth. (Don’t believe me? Check out the figure below.) Now that solar is finally in the public imagination, overinvestment has become a increasing concern.



Source: Strategies Unlimited and Paula Mints (Navigant Consulting).

I wrote already about the market for lasers needed for making thin-film cells (first here and the sequel here). My point then was that the cycle is amplified because it’s the “second derivative."

I worked on solar cells myself, back in 1978 at Texas Instruments. It's great to see it finally make the big time, and if oil prices go up, it will be even bigger. I'm hoping so. A lot of smart people are working on it. Great things are still to come.

Meanwhile, if you're following solar, read Khosla's piece, and read the comments, too. It makes interesting reading.

InterSolar, PW, and gadgets are in; big Semicon West tools are out

SEMI's still newish InterSolar show outshined it's collocated Semicon West show this week, and that may be a sign of the future. InterSolar now has 3 floors of Moscone West, has all layers of the supply chain represented, and was simply more exuberant. The building has big windows. It feels more open. The solar industry this year is down, too, and there are still fewer vendors than at Semicon, but it just seemed brighter at the InterSolar show.

In contrast, Semicon is underground at Moscone North and South. It was cheerful enough, considering the downturn and the lack of windows, but the party was at the solar show. And judging from the signup map for next year, the North and South Halls will be even emptier. In fact, Photonics West is now clearly bigger than Semicon West--at least in exhibitors and floor space, and maybe attendees too. (Rumors that Semicon will be squeezed into the South Hall turned out to be false. They are putting all the wafer processing booths in the South Hall. The North Hall will have everything else.)

To be fair, no one was expecting Semicon West to be much of a party this year. After all, SEMI just announced that tool sales will drop 50% this year to the unspeakable low of $14 billion. (It was $43 billion in 2007.) But it's notable who is showing there nowadays, or rather who isn't. Semicon is a tool show, but Applied Materials, KLA Tencor, LAM Research, ASML, and many other major toolmakers don't have conventional booths anymore. For example, Applied and KLA only had meeting rooms at Semicon, while they showed their solar tools across the street at InterSolar.

Semicon West is now really about the gadgets that the major vendors attach to their systems, the materials they use, and R&D lab equipment. This means that there are suppliers for everything from microscopes and instruments to encoders and bearings. It's a good show for this kind of product development and lab stuff, and some of the specialty tools that are used in North America.

But this year, even that was down. In lasers, Cymer, Gigaphoton, Coherent, and JDS Uniphase--each one catering to the semiconductor industry--all didn't show. The laser companies that I saw were Deep Photonics, Eolite, DPSS Lasers, Innolas, Jenoptik, Quantronix, and Rofin-Baasel. IPG and Newport showed their lasers at InterSolar.

SEMICON is still the biggest fab tool show of the year for North America, and a very important one. One instrument vendor told me that even in a year like this they expected to get some sales from leads at SEMICON. New tool development still goes on. But it's a shadow of its glory days, about 15 to 20 years ago. The SEMICON shows in Asia are much weightier on the big tool side. And that makes sense, since as much as 75% of the world market for semi tools is in Asia, according to SEMI.

And, the industry has grown so big that there are other, more specialized shows to choose from. Like, for examplek, the lithography people like the SPIE Advanced Lithography meeting held in San Jose. There are meetings like this for every sort of nuance you can think of.

Photonics West is also a gadget show aimed at product developers and lab workers, like Semicon, but it spans more industries, particularly healthier ones, like biomedical and security. And it has a cross-cutting technical conference too. (And don't forget the Laser Focus World Marketplace Seminar, collocated with Photonics West every year!)

I think it's a sign of the times. Solar is in. Photonic gadgets are in. Semi tool gadgets are still in. But the days of the big tool show in North America may be over.

Substrates for GaN devices: growing in more ways than one

Our new market on substrates for GaN-based devices is now available. Our take on that market? That's one market segment that's growing, and we don't mean growing just crystals. We forecast merchant substrate sales growing from $280 million in 2008 to $470 million in 2013. But, native GaN has a ways to go, and 6-inch substrates are a long way off.

The new report from our SU analysts Bob Steele and Hank Rodeen updates our earlier reports, addressing both the devices and the substrates used for shorter-wavelength LEDs, blue lasers, and widebandgap electronics. The substrates are mainly sapphire or SiC, with the GaN grown on top, although there are a number of variations and alternatives to get the GaN layer. The substrates have been mainly 2-in in diameter, but are quickly moving to 3- and 4-inch diameters.

The rapid migration stops there, however, in part because of the difficulty to work with GaN but also the manufacturing costs involved in the specific devices fabricated in GaN. In widebandgap electronics, for example, Cree points out that the substrate technology will be ready before the volumes will be. And the electronic devices are relatively large, but not large and pricey enough to merit the jump to larger substrates.

And as for bulk GaN substrates--well--they are simply too difficult to grow and expensive to use for the time being. Inexpensive bulk GaN would be ideal for devices based on GaN, but nature provides alternatives today that are good enough at a lot less cost, and still getting cheaper. Without that volume, bulk GaN remains perpetually behind, or at least far enough behind to be unimportant to our 5-year forecast.

A disruptive application for widebandgap electronics

It's not often that I call a new development "disruptive," but there is one application of widebandgap semiconductors that has that potential. Here, I'm referring to the definition of disruptive as a technology that offers lower performance initially, but fills a market niche that the conventional technology is not filling. The market for the disruptive technology eventually exceeds the conventional one, sometimes completely displacing it.

The application I have in mind is for SiC or GaN devices for power management devices in hybrid and electric vehicles. Their use in these vehicles is not news, but it's easy to overlook the transformation that this technology can help bring to the auto industry.

Widebandgap devices can run hotter, switch faster, and are more efficient that equivalent devices based on silicon. They are more expensive too, but efficiency is very important in the power circuits in hybrid vehicles, and anyway, the price is coming down as volume increases. SiC devices are currently the market favorite, but GaN may be able to play here too.

What is really exciting is how the electric vehicle can truly transform our definition of a car. Once you eliminate the engine, transmission, and drive train, you can do many things very differently. (See, for example, this interview in Tech-On! with Yukitsugu Hirota, from the R&D Center of Calsonic Kansei.

It reminds me of the history of the electric motor, in the 1800s. The first thing that innovators did was to replace water wheels with large motors, driving everything off of a single, long shaft. This is what they knew how to do, since they were used to water wheels, not motors. Then they began to break up the workload into smaller pieces, each driven by its own motor. This process took time--about 30 years, in fact.

It's not difficult to see the same thing happening to the auto industry. Internal combustion engines won't go away, and electric golf carts are nothing new. But the auto industry will transform over the next few decades to a different type of vehicle, and it will need devices based on SiC and GaN.