As of 2019 the MOCVD technology became wide spread across the globe. The major quantity of Compound Semiconductor (CS) based products are made with it. The CS market growth in some areas riches two digits numbers. The area of possible applications is growing. But there are some limitations as well. Few but important areas like microprocessor technology, commodity electronics and at some extent power and RF electronics are still not accessible by CS-based products or has very limited penetration depth in special applications, where other solutions simply don’t work.
The situation is beginning to change with introduction of GaN-based devices. GaN technology has obtained its momentum with the rapid growth of so called white LEDs around 2009. The market volume is keeping growing at present time as well. Moreover due to the LED efficiency growth and further technology development 10 years later the phase transition towards micro-LED is observing. It is not excluding, that during coming 10 years time period yet another phase change will occur. The broad penetration of white LEDs stimulated the development and expansion not only of MOCVD equipment, but the whole infrastructure as well. The gas supply, metal-organics became more or less commodity goods specially in some regions of our planet. Never the less the time before present can be characterized with the one important mark, namely the CS-technology were used where nothing else will provide the same functionality. It has been chosen based on physical property and not because of economical assessment. The rule - everything what can be done in Silicon will be done in Silicon - remained unbroken. The Si-based technology is by tens years older, than CS counterpart. It is more mature, has much higher market volume and most importantly cheaper in production per unit base. Considering electron-based devices it is not a surprise, but we all know photon-based devices specially in PV applications are almost without an alternative if comparing by W/$ measure. Nothing else even comes closer. It’s a pity situation for CS materials as only some of them posses a direct band gap, which improves photon-electron conversion by a huge factor.
In fact the situation can be turned other way around. Everything what at the moment Silicon is used for can be done with CS-type of materials plus all those photonic devices, where CS used anyway. What prevents that to happen? It is solely the production cost as we firmly believe. The big semi-industry companies are watching around how to jump off the Si needle. They require higher operation frequency for their CPUs, lower power consumption and all those features CS has to offer, but the potential product end price keeps them away from that move. Over 40 years time period customers get used to enjoy all the same prices while getting double of performance every time they are ready for the next upgrade. The necessity to pay more even if performance increased and/or power consumption went low will stop them from the next upgrade move. The motivation to buy a faster PC/laptop has declined a lot over the last 10-20 years time. The software developers have a hard time to make their programs to run slowly with each new release as the modern multi-core CPUs are too powerful.
Hopefully the new age is coming soon to help. The IoT market drives the power consumption demand down to the lowest possible values. The mobile phones, which have less and less to do with phones of old times are the most wide spread IoT devices actually, but not really recognized as such. The on-battery time remains in the list of the highest demand features since the introduction of the first smartphone. The AI is on the other hand requires the highest computing power while remaining within the reasonable electrical power consumption range. These two and possibly others related areas like autonomous driving could be those pushers which can bring the CS technology back to the game.
We consider that some work on CS side has to be done too, otherwise the coming waves will be served as usual by Si-only. Therefore we started that project to promote a strategic inflection point like mentioned in the book of a former Intel CEO Andrew S. Grove “Only the Paranoid Survive”. Our main target is the major part of CS production a MOCVD process. We aren’t the first to see the problem, but so far MOCVD process remained untouchable or considered as not flexible enough for an upgrade and most researchers focused on its replacement for a given special material and process step. That is not a real improvement. It will lead to the technology fragmentation, while technology integration is a real goal.
We aren’t wizards yet and we fully understand, that a significant cost reduction can be only achieved, if also a mental shift of what is really required will be made as well. Otherwise an improvement is still possible, but relatively limited. Generally speaking we have to ask ourselves if the MOCVD technology will be still existing 10-20 years later how it would look like? Would it be the same like we know it today? That is the question our company is looking answer for.