TL;DR 21700s produce less power (not necessarily energy) for the same cycle length and weight, so Tesla will either have to reduce power consumption of their high end vehicles -or- add more batteries, which adds weight. Both of these things reduce performance.

I'll be honest: I've got much more knowledge about batteries than I do about Tesla's strategic direction.

I do know that there are currently problems with Teslas sustaining their power output when on high performance mode. Since the power density of 21700 cells will almost certainly be less than 18650 cells (this is basic physics) [0] I think they will use 18650s for their performance vehicles. However, I could be wrong; they -could- simply increase the number of cells used and make the vehicle heavier. Strategically, I would use 18650s (at least with the performance cars) if I was Tesla. However, for PR reasons they may choose to use 21700s and compromise the performance of the performance cars.

My claim that they have a low energy density is on less strong ground, of course. As Musk said during the Q3 2014 conference call, "show me the cell, not the powerpoint." I only have powerpoints and no cells. However, the powerpoints I do have show that the energy density will be between 220 and 260 Wh/kg. They may be up to 280 Wh/kg. But I doubt that they'll beat the current maximum of 310 Wh/kg that you can get today.

Tesla does a lot of really cool stuff, but there's a bit of a Tesla reality distortion field going on. For example, this post here [1]. Some genius speculates that batteries grow at a 5.94% compound rate, and then extrapolates to assume that Tesla's cells will reach 350 Wh/kg. Here's the kicker: li-ion battery cell capacity has been practically stagnant for 11 years. In some of these 11 years, the cells have gotten cheaper, but not by much. If we truly had grown at that 5.94% compound rate for the past 11 years, we'd currently have a 528 Wh/kg cell today.

Li-ion battery energy density grows in spurts. Realistically, there have been a few major improvements, with very small incremental improvements in between. For example, in 2001 we were at 180 Wh/kg. In just a handful of months we jumped to 260 Wh/kg, then to 280 Wh/kg as manufacturing processes improved. In the last 11 years, however, improvements have been maybe a few Wh/kg per year, and there have been literally zero improvement whatsoever in the past four years, by anyone.

I think Tesla - may - get a bit north of the 310 Wh/kg we can get today, and I'm confident in their ability to reduce costs, but I don't expect anything revolutionary (energy density wise) like Musk implies in his presentations. I'd set the threshold of 'revolutionary' at 400 Wh/kg, a 29% improvement from today.

For future-me-check-if-I-was-right purposes: these estimates are for Tesla's cells released in early 2017.

[0] The larger a cell is, the lower its theoretical power density must be. Power density is a function of power dissipation, and larger cells are worse at dissipating power. These new cells are only a few millimeters wider, but that's enough to increase their volume by 50%, which means their power density will almost certainly suffer.

[1] https://teslamotorsclub.com/tmc/threads/21-70-cell-informati...