This is all an artifact of non-dedicated circuits. With a dedicated T1, for example, you know what you are getting and the circuit is idle when you are not usi

This is all an artifact of non-dedicated circuits. With a dedicated T1, for example, you know what you are getting and the circuit is idle when you are not using it. With a shared circuit, advertised "peak" and "average" bandwidth is all virtual and is managed via QoS, and is based on expected usage patterns and leveraging the anticipated lulls and peaks. It's very similar to financial arbitrage. So it may be the case that reducing the risk of sudden, simultaneous peak events allows the overall network to be much lower cost and more performant for other uses. T-Mobile is trying to squeeze out the most value for its users at the lowest cost. If a user wants a dedicated circuit, those are available from other vendors. Perhaps in the future carriers will offer circuits with QoS specifically tailored to different categories of users. It seems like custom QoS could be offered to individual users and priced accordingly. I'd happily tolerate lower max video download bandwidth in exchange for a lower price b/c I rarely watch video on my phone.