Back before we called them memes, the US comics strip Peanuts introduced one of the greatest memes of all times – Lucy, Charlie Brown and the Football. Charlie Brown asks Lucy to hold an American Football so he can kick it. She promises that this time she will hold it, unlike all his previous attempts. He runs up to the ball and at the latest second, Lucy pulls the football away and Charlie Brown falls down spectacularly. Again.
We are reminded of this proto-meme every team we look at the photonics industry.
Photonics or Optical components have been an incredibly powerful part of modern networking technology for decades. They underpin the Internet, making the whole thing possible. Despite this, optical has been a terrible business for as long as we can remember. One investor we know regularly described Optical as the “Airline Sector” of the Tech sector, a capital intensive industry with perennially poor returns.
How can optical be both fundamental and uneconomical? Part of the problem is what we describe as the PhD to manual labor ratio. Optical componentry requires R&D teams staffed with serious engineers comfortable with pushing the laws of physics, but the products they design involve a high degree of hand assembly. We once visited a brand new, state-of-the-art optical factory and remember clearly the first room had what we would best describe as a mud sink that workers used to craft ceramic components. The very opposite of the “clean rooms” we imagine when we think of high tech assembly.
Another explanation can be summed up in one word – packaging. Optical systems have a number of very highly specced components – lasers, sensors, fibers, etc. These have to be assembled and packaged in some very tight form factors. As widespread as optical systems are, the total units sold every year are measured in the millions, which is fairly small scale for modern electronics, hampering the economics of investing in more automation.
Moreover, the behavior of optical signals is very different from those of electrons. These two are not interchangeable, and need different systems to manage them. The big problem in optical systems has traditionally been how to handle moving data streams from fiber to electronics and back again. For example, a high capacity router can have dozens of incoming optical strands. Each signal has to be translated into electric signals, interpreted and then translated back into optical signals for transmission to its next destination. Historically, this has involved some fairly complex sub-systems, and the end result is that optical systems are very expensive – with the optical cables often costing more than the router or other hardware they are plugged into.
The solution to these problems is something of a Holy Grail for the industry. How to package optical sub-systems so that they are ‘seamlessly’ integrated into the electronics. Ultimately, everyone would like to use semiconductor manufacturing technology to manufacture these Silicon Photonics systems.
The good news is that we are likely getting closer to this. Luxterra, a start-up acquired by Cisco a few years back, had a promising approach, albeit one which still required a fair amount of manual assembly. Now Ayar Labs seems to be closing in on something fairly significant. There are other start-ups out there as well showing varying degrees of promise. Accomplishing this could unleash a huge change for the industry – not just for networks but for data centers (aka the Cloud) as well. It seems that the industry is getting close, and this time they totally promise to hold the football.