We need a 'Giga-fab' for Gallium Nitride, somewhere

But it is still in its infancy, costly, and needs much development. It could take a decade or more to make an impact. As with any young technology, the industry around it today is fragmented, comprised mostly of small players trying to gain a foothold.

The question: Should we let the market develop at its own fragmented pace, or should we make a bold, concerted move at the national level  -  through industry and government partnerships  -  to accelerate its adoption for the common good?

The technology in question is the Field Effect Transistor (FET) based on the compound Gallium Nitride (GaN). By all accounts it is set to displace the venerable silicon MOSFET. The advantages of GaN over silicon stem from its basic crystal structure  -  not only do electrons flow faster in GaN, the material can also withstand greater voltages before breaking down into a conductive state. These properties enable greater energy efficiency, higher power density and smaller device size.

Gallium nitride, as a semiconductor material, has been around for a while. LEDs have used it. But now, the GaN FET built on a silicon substrate  -  in contrast to GaN on a sapphire substrate, or bulk GaN that is favored for LEDs  -  is on a trajectory that could make these devices producible in volume and thus affordable. The semiconductor industry, recognising that the power switch of choice, in the not-so-distant future, is likely to be based on gallium nitride, is beginning to react.

A number of companies are supplying 'epi wafers'  -  standard silicon wafers topped with a thin 'epitaxial' gallium nitride crystal layer  -  on which FETs and other devices can be built, but with no real economy of scale. Gallium nitride devices for power converters are produced by the thousands each week; silicon devices are made by the billions.

From half a century’s experience with successive generations of semiconductors we know that, notwithstanding the fact that it can do more, a new technology only succeeds at displacing the previous one when it costs the same, or less, and is proven to be just as reliable. But it’s difficult for any single company to advance the state of manufacturing art sufficiently to compete with silicon, which has benefitted from decades of research on how to purify, dope, and grow it.

Let’s look to Tesla for inspiration. Tesla’s 'Gigafactory' for manufacturing Li-ion batteries has a simple goal  -  accelerate the transition to sustainable transportation. This requires cheaper batteries. As with any manufactured good, the cost declines with volume. Gigafactory will make more batteries than anyone has ever imagined. It’s a 'Hail Mary' pass.

Now, let’s take a page out of Tesla’s strategy handbook. If GaN is so obviously disruptive, we need a 'Giga-fab' to make GaN-on-Si wafers  -  in the highest possible volumes and of the highest possible size (12" diametre), to supply to the device makers.

As always, history serves as a good guide. Tesla is relying on the well-proven concept of the Experience Curve, postulated by the Boston Consulting Group, nearly 50 years ago: Unit production costs fall by a predictable amount — typically 20 to 30% in real terms  -  for each doubling of 'experience', or cumulative production volume. In other words, since you can predict your future costs, you can price for competitive advantage today, stimulating adoption, generating even higher production volumes and a speedier slide down the Experience Curve. The entrenched competition for GaN power devices today is silicon. Aggressive pricing on GaN, based on predictable costs will speed the displacement of silicon.

Giga-fab may smack of industrial policy to economic purists who balk at the thought of the government picking winners and losers.

But here, too, the Gigfactory example is illustrative. Tesla and its major partner in Gigafactory, Panasonic, are industrial companies, but they’re benefitting handsomely from government largesse: Over the next 20 years,Tesla could take in nearly $1.3bn in tax benefits from the state of Nevada, site of the Gigafactory.

US taxpayers may think they have done their bit already:

PowerAmerica, a public-private partnership of industry, the US Department Of Energy (DOE) and academia, is attempting to accelerate the adoption of advanced semiconductor components made with silicon carbide and gallium nitride into a wide range of products and systems. DOE has committed $70m to the effort over five years.

DOE’s GIGA project (GaN Initiative for Grid Applications) was initiated in late 2009 at a time when GaN‐on‐Si development efforts in the United States were still in their infancy for power management applications. The project has yielded a number of important advances and valuable insights in the development of GaN‐on‐Si power devices and provided critical proofs‐of‐concept to demonstrate what is possible with GaN‐on‐Si technology.

Proof of concept is one thing. High volume manufacturing is another. The roadmap for semiconductor technology development is long and winding. History shows that it can take as much as 20 years to bring a new technology from theoretical proposal to commercial production. And that requires collaboration, just as much as it does capital. Collaborative efforts on reliability testing procedures and packaging designs are as important as establishing high volume production processes.

But, most important, it requires a vision: The GaN Giga-fab would be a collaboration of industry and government on a scale designed to make an impact on the economy in 5 years, not decades. A rising tide does lift all boats.