The processing power of super computers, cloud servers and ordinary computer processors are still following Moore’s law.

This means that the transistor density in integrated circuits doubles every approximately two years. This results in a larger processing power being available for scientific research, but also for ordinary industrial and consumer applications for the IoT. As the sensor prices have dropped, the number of sensors also increase and thereby the data, which is produced by these sensors, and as the connectivity solutions for the devices are also getting cheaper so does the data being available on the web. This facilitates centralized processing that also enables to search for correlations between multiple sensor data sources to aggregate the value from the single sensor. These data of course require storage, but also processing power to harvest the value hidden in correlating the data from multiple sources. In addition, the advances in processing algorithms such as artificial intelligence by using artificial neural networks and blockchain technologies processing distributed ledgers to validate transactions, requires substantial additional processing power.

Previously, it was primarily possible to have simple threshold algorithms from a single sensor source, but as processing power in microcontrollers became more available in the 80’s and 90’s, it became possible to perform Fourier transforms such that spectral analysis could be used for instance for predictive maintenance. Here rules for decisions based on the individual frequencies was constructed by engineers, based on experience from a limited dataset and integrated into the product. Today, all the data is sent to a cloud server continuously. The cloud server then uses AI to construct optimum rules for the system, which can then be transferred back to the IoT device for optimized decision making. The advances in processing power both facilitates the AI on the cloud and supports the more complex algorithms on the IoT device itself, for edge computing.

Image source: http://www.singularity.com/charts/page71.html.