In a 4G network each operator will have a single network. This of necessity will be a general purpose network supporting the range of use that we all use from browsing the Internet, phone services, streaming and more recently IoT. But many of these applications have different demands on a network Maybe they require very low latency, have a high download data demand, or support many millions of low data devices. Network slicing allows the network to be split into different virtual networks, slices, that provide an end to end virtual network. Each of which can be tailored to the needs of the applications. It is envisaged that some companies would be able to "purchase" their own 5G network and tailor it to their need, in particular in smart factories.

• Autonomous vehicles require low latency but not a large amount of data
• Mobile Broadband
• Massive IoT network
• Smart factory network
• Streaming TV (or film) service does not really need to worry about latency but there is a large downstream data demand and no real upstream data demand
• Online Video Games
• General purpose mobile phone network

The network operator could run these different networks themselves, with different revenue models, but also sell virtual networks. Netflix, or similar, may be interested in a streaming virtual network which could be the future of TV. Ford (or another car maker) may wish a virtual network for their autonomous cars. Large scale automated factories could be interested in their own virtual network. Some of these virtual networks will be built on mmWave frequencies, and some on the more general 700Mhz or 3.5GHz.

Network slicing is seen as a major discriminator that makes 5G much more than an upgraded 4G, whereas 4G was really a faster and better 3G