In 2010, Zel Diel, managing director of US-based power electronics analyst business, Venture-Q, started investigating how and when GaN-on-silicon devices for power conversion applications would be commercialised. Within months, he had concluded that by 2016 industry would see high voltage field effect transistors in low quantities so designers could use them to build demonstration systems. These structures would be deposited on 200mm silicon wafers to achieve an acceptable cost parity with silicon alternatives.

Nearly four years on and having just unveiled his latest report on commercialization strategies for GaN-on-silicon power devices, Diel's time-frame has shifted. "We're not so far off now. By 2018 system designers will be able to say, yes, I can buy [HEMTs] for $5 or $6 and at a system level cost parity," he says.

Still, the road to true commercialisation is going to be rocky. As Diel asserts, time and time again, market research forecasts, largely relying on vendors' unrealistic product availability timetables, have underestimated the challenges GaN-on-silicon developers face and proven optimistic.

While Efficient Power Conversion (EPC), US, introduced the first commercially available low voltage GaN-on-silicon HEMT nearly four years ago, this company and every other vendor has since struggled to deliver commercially-viable high voltage - 600V -  devices. EPC has repeatedly pushed back delivery of high voltage devices and the only generic 600V GaN-on-silicon HEMT device publicly introduced by Transphorm, has been offered to select partners under non-disclosure agreements.

So why the slow progress? For starters, myriad industry infrastructure constraints exist. "We are dealing with a completely different technology in a system level design that power system designers have never seen, never had experience with and never been trained with," he says. "We have no reliability systems in industry, only JEDEC, but this is a silicon standard so qualifying for JEDEC doesn't mean a thing; you can't even qualify a part and send it to be integrated into a system as there is no standard on paper."

What's more, both GaN-on-silicon developers and oscilloscope manufacturers have also highlighted the test and measurement challenges posed by high voltage, high bandwidth GaN devices. "Tektronix has said it had no oscilloscope fast enough to measure switching in these devices," highlights Diel.

And then there's the technology issues. While the lattice mismatch between the silicon substrate and GaN layers has hampered development, so-called current collapse is proving to be the primary stumbling block stymieing progress.

The phenomenon is an inherent attribute of GaN-on-silicon, as well as GaN-on-SiC HEMT device structures, which manifests as an increase in the device's on-resistance during high voltage operation, degrading performance. Crucially, current collapse, especially in GaN-on-silicon devices, is exacerbated at 400V operation and higher and is now the the primary focus of research and development efforts in industry and academia.

"Current collapse is the kiss of death and we cannot find a cure," says Diel. "Since 1999 at least 1000 PhD theses have been written on this and still people are struggling with the problem. Vendors are tackling this challenge but no-one has a solution."

So don't these technology challenges twinned with industry infrastructure constraints make a 2018 commercialisation timeline ambitious? Diel thinks not, highlighting recent industry developments that signal real progress.

In October 2013, Powerdec, a developer of GaN-on-silicon HEMTs, demonstrated a 600V, 10A HEMT it had produced alongside the University of Sheffield, UK. The device was fabricated using the organisation's proprietary polarisation super-junction process, which Powerdec claims suppresses current collapse.

And then in November 2013, Japan-based ICT giant, Fujitsu, and University of California Santa Barbara spin-off, Transphorm, signed a deal to integrate the GaN power device businesses of each. The move will boost the market competitiveness of Transphorm's cutting edge GaN technology by capitalising on Fujitsu's manufacturing prowess. In short, industry should see high volume, high performance and cheap GaN devices sooner rather than later.

"Due to long-time research at UCSB and with Cree, Transphorm's technology is very advanced... but moving into manufacturing brings many complex issues," says Diel. "Transphorm has realised this, and this where Fujitsu, a manufacturing powerhouse, comes in."

And we can expect more of the same. Diel predicts we will soon see EPC, in reality the only vendor currently offering GaN-on-silicon HEMTs on the open market, teaming up with a systems manufacturer.

"More and more vendors are now recognising that what they are dealing with is not as simple as research and development presented at conferences," he says. "I speculate that this Fujitsu-Transphorm deal will be repeated quite frequently as we go forward, and those that don't do this will become laggards."