The Moore’s Law Sonic Boom
Chances are, you’ve heard of Moore’s law, the observation that transistor density doubles every two years. There’s a significant blip coming that’s not about density but about availability.
For decades, Moore’s Law has proved accurate and become integral for everyone from computer enthusiasts to the semiconductor industry itself to plan for the long term.
Autodesk CEO Carl Bass has pointed out another step-change we’re approaching in terms of the economics of computing power:
We’re approaching the point where one second of computing time on 10,000 computers working in parallel is as inexpensive as 10,000 seconds of computing time on a single computer. (To save you the arithmetic, that’s nearly three hours’ worth of computing delivered in one second.) If machine learning can be made highly parallel, the cloud can deliver results almost instantaneously.”
As parallel computing becomes cheaper than scheduled computing, a huge number of processes, and even economics, will change all at once.
Everything that’s queued for processing will no longer necessarily need queuing. Running extensive check-lists, considering complex what-if scenarios, and preparing reports, will no longer take so much time that business processes need to work around them.
Take report preparation for example. If reports can be created and modified on demand, strategic meetings no longer need to wait for regular intervals, but can happen as required. If reports can be modified at whim, and designed to only appear if there’s a problem to action, do you need regular management meetings any more?
Product improvement processes may no longer need to collect customer feedback over several visits, since their design responses to feedback could be instantaneous.
In the past, as computing power has doubled, new applications have become affordable, and those new applications have arrived at a rather steady, predictable rate. For example, it was clear with the introduction of the iPhone that our most powerful personal CPU would live in our pocket, not on our desktop. These waves of new applications were evenly spaced.
As parallel computing power becomes equally priced to linear power, those waves will overlap at once, like a sonic boom. Small shifts towards real-time results using parallel computing will, when combined, create a multiplying effect.
You don’t need to create some kind of parallel computing product or service to benefit from this — any organization that can adapt its processes in a few small ways to make use of this shift will find themselves outmaneuvering others.