For an industry that has grown so rapidly over the past few years, the top tier of high-brightness LED (HB-LED) manufacturers has proved remarkably resilient to new entrants.
Nichia still heads a list of the six leading chip manufacturers by some distance, with Epistar, Osram Opto Semiconductors, Cree, Philips and Toyoda Gosei fighting it out at the top of the table.
And so it ever was – save for a few spin-offs and acquisitions. But now a challenge to the status quo appears to be emerging, with major corporations in Japan, Korea, Taiwan and China apparently eyeing the HB-LED sector enviously.
As the leading supplier of the MOCVD equipment that is essential for chip production, Aixtron probably gets to hear about these moves before anybody else. In the German firm's August investor call, Paul Hyland came clean: "Companies that have traditionally operated in the parallel silicon sector have expressed interest in investing in LED equipment," said the CEO, "as well as liquid-crystal display manufacturers, who are operating further down the food chain."
It's an open secret that those LCD panel makers include the Taiwanese duo Chi Mei Optoelectronics and AU Optronics (AUO), both of whom rank among the world's leading suppliers to TV and PC monitor producers. A third Taiwanese LCD company, HannStar Display, is a subsidiary of the huge Walsin Lihwa Corporation, which recently purchased a production MOCVD tool. And the company from the silicon business that Hyland alluded to is widely felt to be the giant Taiwan chip foundry TSMC.
Add to these three firms the trio of electronics powerhouses that have existing LED experience – Samsung, Panasonic and Toshiba – plus the recent acquisition of several Aixtron reactors by the government-backed Chinese upstart Century Epitech, and suddenly you have the makings of a very different looking top tier of LED producers.
Although all of these companies are keeping their cards very close to their chests, one thing is abundantly clear: if they wanted to, any of them could change the landscape of the LED industry at a stroke by means of an acquisition.
However, that hasn't happened – yet. Instead, it seems that they each want to produce LED die internally. Long-time LED industry commentator Bob Steele, an analyst at Strategies Unlimited, reckons that Chi Mei has been working on LEDs for at least the past couple of years. Rumors would even suggest that the flat-panel specialist is already shipping some low-end die in China.
Chi Mei's rival AUO is only just starting to buy reactors, and is at least a couple of years behind, but it shares this desire to become vertically integrated from the epitaxy stage onwards. The suggestions are that TSMC will do the same and, given the resources at its disposal, this is not altogether surprising: "Compared with the resources required to build a silicon semiconductor fab, or a flat-panel display fab, this is pocket change," said Steele of the investment in MOCVD reactors so far. But the strategy of AUO and Chi Mei seems more confusing.
"If you look in Taiwan, where there is so much capacity to supply LEDs, you wonder why they would go to the bother of making an investment to build up their own internal supply," Steele told Compound Semiconductor.
When you consider that neither AUO nor Chi Mei currently makes the cold-cathode fluorescent lamps that are used to backlight their LCD screens, this strategy seems even more curious. Perhaps these two companies anticipate a lucrative sideline as merchant suppliers of LED die?
"The natural thing to do – once the capacity is there – is to sell into the merchant market as well," added Steele. "But given the capacity that's out there, especially in Taiwan…well, do we really need another merchant chip supplier?"
For these flat-panel makers, control over the supply of LED die appears to be a key strategic decision. But for a foundry company like TSMC, that argument is irrelevant. So what could be TSMC's big plan?
Asif Anwar, director of the GaAs service at Strategy Analytics, has also been monitoring developments in the LED industry closely. "When you start talking about a company like TSMC and compare them with even the biggest LED manufacturers in Taiwan, such as Epistar, then they are absolute giants," he commented. "So, if they put their minds to it, then they will change the landscape."
However, any market entry for TSMC would only make sense if demand for LEDs with specific wavelengths and emission characteristics becomes huge, reckons Anwar. He believes that this could ultimately be the case in general lighting, where three or four individual chips might go into every solid-state lighting "bulb" for future domestic applications. "TSMC would want to establish a DRAM-style production regime," said Anwar. "And even though it has grown quickly, the LED industry is not yet truly 'volume' in the same way that silicon is."
Two major hurdles
With such an elevated level of demand years away, TSMC has time to develop its own processes. If that's the strategy that it wishes to pursue, a lot of time will be needed. As Steele points out, any new player trying to make a significant impact in the LEDs arena is going to have two major hurdles to overcome – technological development and intellectual property.
"Buying a reactor is one thing, but being able to design and produce high-quality chips is another," Steele said. Part of the key to that is to acquire experienced technologists, something that Chi Mei is believed to have done successfully when the Taiwanese LED industry went through a protracted period of consolidation a couple of years ago.
However, even with the pool of MOCVD talent that is available in places like Taiwan and Japan, it will take these companies years to catch up with the industry pioneers, and TSMC is likely to be looking at more than five years before it can realize any home-grown LED ambitions. Even with good technology, there is the minefield of intellectual property to negotiate – although the size of the companies seemingly entering the LED business would suggest that they are more than capable of matching the firepower of lawyers working for the likes of Nichia and Seoul Semiconductor.
The business strategy that would appear to make a lot more sense than independently reinventing GaN LED technology is outright acquisition of one of today's major players. Financially, that ought to be no problem for the likes of TSMC, Panasonic and their ilk, but who could be a realistic target?
Nichia, the world's top LED chip maker by far, will no doubt have been subject to these enquiries in the past. However, as a private company with a fiercely independent culture, it has resisted any bids made up until now, and would appear likely to do so in the future. Of the remaining major players, two – Osram Opto Semiconductors and Lumileds – are already subsidiaries of huge multinationals and are critical elements of each of these companies' future generations of lighting technology.
That leaves Cree, Epistar and Toyoda Gosei. As public companies free from parent owners, Cree and Epistar are more likely to be viewed as potential acquisition targets. Rumors about acquisitions of US-based Cree by a major silicon player are pretty common fare in financial circles – although rarely backed up with any hard evidence. For TSMC at least, a bid for Cree might be less likely than one for its near neighbor Epistar.
Aside from the obvious geographical proximity, TSMC also has strong links with California start-up Bridgelux – a company that is in the middle of a patent dispute with Cree over GaN chip technology, and whose foundry partner is Epistar.
Logistically, Epistar and TSMC looks to be the most likely hook-up. But, if the rumors are true, why has TSMC invested in MOCVD reactors rather than build up a share in the LED maker? Perhaps, as Steele muses, the foundry giant is "dabbling" in LEDs, familiarizing itself with the technological processes required and figuring out its next move.
"One thing that you have to look at is the investment that has gone on at all of the major players to reach current performance levels," commented Steele. "It's enormous." That said, the LEDs required by Chi Mei or AUO may not necessarily have to be on the leading edge of emitter technology.
Perhaps a steady investment in process development will end up bearing fruit for TSMC and co, or perhaps they will eventually come to the conclusion that the only way to crack the LED business is via a high-profile acquisition. Either way, it will be some time before today's leading players are displaced from their rankings. Steele and Anwar both suspect that the same five or six companies will lead the way for as much as another five years. And beyond that, it's anybody's guess.
View pdf of article