Evolving Cree hits a crossroads
Cree is a quintessential start-up success story. Founded by a handful of researchers from the State University of North Carolina at Raleigh less than 20 years ago, the company has grown to more than 1000 employees, made annual sales of $400 million and become the world leader in SiC technology as well as one of the primary manufacturers of GaN-based LEDs.
Now, Cree faces a new challenge. Its one-time paradigm-breaking technology – blue-green emission from GaN – is being implemented by dozens of rival manufacturers and penetration of LED die has reached saturation point in its most important market: that of lighting up cell phone keypads and displays.
As John Palmour, one of Cree s original founders and now its vice-president of advanced devices, explains: "If cell phone growth slowed [in the past], we were able to power through it because we had not fully penetrated the market. So even if cell phone growth was flat, we were able to increase revenues. We have now penetrated the cell phone market and are riding with [its own] fluctuations." With continued falls in average selling prices, sales increases are required just to maintain flat revenues. Cree has faced this problem for the last four quarters.
Entering a transition period, Cree needs to penetrate new markets in order to grow. When it comes to LEDs and new applications, Mark McClear, solid state lighting director, is quick to recite a three-pronged mantra when determining whether Cree s GaN-based LEDs are positioned to enter a new market. "Brightness, efficacy (the optical output for a given electrical input) and the economic value proposition must all be met," he said. When GaN can compete with existing technologies on each of those fronts, McClear considers it a market ready for Cree to tackle. In his sights is what he characterizes as "a trillion dollar market" – McClear s estimate of the total installed base for white light.
If the name Cree has become synonymous with blue LEDs, then to navigate this crossroads successfully, it needs to become synonymous with white light. The markets that Cree is eyeing within this arena are the backlighting of large LCD screens (primarily in notebook PCs) and general white-light illumination.
Before Cree could even consider attempting to sell LEDs for the large panel LCD market, it needed to at least match the performance of today s dominant backlighting technology – cold cathode fluorescent lamps (CCFLs), which run at an efficacy of 50–60 lm/W. "There is a real expertise in designing these systems and you have to know the chips to make the design. A customer who has been using CCFL is not suddenly going to be a world expert in LEDs, so it is very natural for us to take that next integration step," McClear said.
This is exactly what Cree has done and its display now delivers 11% more lumens per watt than the CCFLs. By the end of 2006 McClear predicts that Colorwave systems will be shipping and generating revenue. Low-hanging fruit General illumination has long been dominated by two standard sources: the Edison socket filament bulb and the fluorescent bulb. Edison bulbs deliver 10–15 lm/W, while fluorescents have an efficacy in the 70–100 lm/W range. Cree s latest commercial XLamp packaged die produces 70 lm/W.
From this perspective it would appear that the Edison socket market has the potential to be penetrated by XLamps. However, because of the relatively high cost of LEDs and the amount of time it would take for customers to feel the benefit, McClear is less than enthusiastic: "I am not a big fan of the Edison socket for LED applications – there is lower-hanging fruit to be picked." The lower-hanging fruit includes streetlights, torches and large high-intensity discharge (HID) warehouse lighting. "Torchlights are dominated by LED bulbs and this has only happened in the last year," said McClear.
The reason that these applications can now succeed is that there can be more to the economic value of a bulb than just its upfront cost and the cost of powering it. "It costs about $150 to buy a new streetlight and another $150 to replace it when you send a crew out with a $200,000 truck," McClear explained. "So if a bulb lasts only 1–2 years, you end up paying for it five times [in 10 years]. An LED lamp can have equivalent light output and costs about twice the amount from an original purchase point of view, but it lasts at least 10 years."
For municipal authorities with thousands of streetlights, this represents a significant saving and the LED advantage is much easier for a large-scale consumer to appreciate than an individual. When the additional costs associated with replacement of bulbs are included, streetlights, HID lights and torchlights based on LEDs start to make economic sense.
Historically, there is roughly a one-year time lag between developments in Cree s research labs and turning that breakthrough into a commercial product. Recently, the company produced laboratory LEDs with an efficacy of 131 lm/W. "If historical trends continue, next year should see a commercial XLamp device with efficacy above the 100 lm/W range," said McClear.
At that point, a potentially massive market becomes accessible – the replacement of fluorescent bulbs by white LEDs – a key element in McClear s vision of the "trillion dollar" white light market. While many other GaN LED manufacturers will tread the path to LCD backlights and general illumination, Cree s experience in SiC technology gives it more options. Besides the obvious advantages of an internal source of substrates, SiC and GaN-on-SiC offer a wide range of electronic applications and moving these devices into the commercial market is another key aspect of Cree s current transition.
Nowhere is this more apparent than in the recent opening of its new facility in Raleigh s Research Triangle Park. The 275,000 sq ft fab and its 150 employees are solely devoted to the design, manufacture and test of SiC and GaN-based electronics, while the site also houses Cree s SiC MMIC foundry line. Currently running 3 inch SiC substrates, it is already configured to make the anticipated jump to 4 inch material, a critical yield consideration when fabricating large MMIC die.
Despite the rather unique materials it processes, this fab looks like, and is run as, a fairly conventional GaAs or silicon facility, with bays of I-line steppers for photolithography, dielectric and metal deposition, dry etching and inspection equipment. The only portion of the process remaining in the older Cree facilities is that of SiC boule growth, and SiC and GaN epitaxy using equipment that serves double-duty for LED production.
The other important link between Cree s LED and electronic offerings is the increased energy-efficiency that both types of device deliver. SiC Schottky power diodes are a prime example of improvement in device performance reaching a critical threshold that has enabled commercial penetration.
Cree Schottky diodes operating at reverse voltages of 300, 600 and 1200 V, and a variety of currents, are now available. What has made these diodes a commercial reality is not so much the fact that the wide bandgap of SiC can withstand such large reverse biases, but the reduction in micropipe densities that are fatal to SiC Schottky diodes. Under reverse bias, breakdown will take place along the length of these voids, which run through the wafer. Micropipes have now been sufficiently reduced that the large-area Schottky diodes can be manufactured with a high yield.
These devices can be inserted into many systems that currently use silicon PIN diodes. Unlike a PIN, a unipolar Schottky diode does not require charge to be stored or dissipated before it can switch on or off. This property makes it a faster switching device, in turn improving system efficiency. "When you take a switched mode power supply that is [already] 85% efficient and improve that 3–4% by switching out the silicon PINs with SiC Schottkys then your total losses actually go down by 33%," explained Palmour.
This improvement also means that a system using SiC Schottkys dissipates less heat, which allows for a more compact system footprint. The higher efficiency and faster switching of SiC Schottkys is already driving their penetration of the silicon PIN market for motor-control and electrical inverter applications.
Add to that the potential for both SiC and GaN in high-power, high-efficiency base station amplifiers for wireless communications and Cree appears to be well on its way to making its transition from a "simple" blue LED die manufacturer. The end result of that change should be a much broader solution-oriented company that is focused on producing a range of energy-efficient components. "We are moving into LED backlighting, general illumination, and SiC and GaN devices for power and wireless applications," summarized McClear. "Any one of those market segments would make a great business, but we have all four."
