Quite simply three years of continued progress in the historical 5-8% yearly range could give around 20% or so of cost reduction so GF only needs to get another 10% or so. Contrary to what he suggests I don't consider 20 years of production to be a "mature" technology with all economies of scale already achieved. Putting the equivalent of the current world production in a single factory and dealing directly with material suppliers will certainly increase cost savings and efficiency. Plus his quoted battery expert is quite wrong about the potential for future advances in lithium chemistry, there is more to be had before we get to lithium air.
Also, keep in mind that all of the investor prospecti, conference calls, etc are using 2020 as the reference date for when all of these metrics will be valid. So 35GWh of cell production in 2020 might mean only 27GWh of production in 2017 with all the same machining, processes, etc.
So when we model costs to set up the factory we need to perform the same translation from current costs, so if Tesla was hypothetically putting this factory online tomorrow it would need enough CapEx to produce ~22GWh of batteries, and then natural improvements in chemistry or whatnot would "ramp" up the capacity of the factory to the promised levels in 2020.
Focus Should Be At Cell Level Economics
Because of how these changes in capability occur over time, the better way to model all of this is to focus on the individual cells. The basic manufacturing costs haven't changed in some time, and the materials costs vary only depending on the specific chemistry.
For some time now those costs have averaged out to ~$1.70/cell +/- some modest amount. In fact, the manufacturing costs of a 2.2Ah cell is
virtually the same as the 3.0Ah cells right now, which neatly illustrates this rule of thumb in real time. There is every reason to suspect that cell costs in 2020 will still be somewhere close to $1.70/cell, with minor reductions due to scale.
Mob Economics 101
Anyways, when Elon promises a ~30%+ reduction in per kWh pack costs by 2020, my assumption is that the real improvement at the cell level is greater than that, and the balance is the Vigorish that Tesla is paying Panasonic to entice them into a huge investment.
When
Panasonic counters that they're not sure they can make a 30% reduction, this is just the public face of a negotiation where Panasonic is demanding a greater profit.
Plus, if you read the Panasonic quotes closely you see that they are offering to "help" Tesla with development of
electronic controls in addition to battery-cell cost reductions as a means to lower the overall cost of the battery pack.
Note that the article also states that
One of the main growth drivers for Panasonic will be an expansion of its auto-parts business, which hinges in part on its key client Tesla.
When you factor in the likelihood that most of the commentary in the press about achieving production "efficiencies" is actually just the public face of a Panasonic shakedown, it becomes much more comprehensible. There are clearly efficiencies to be had, but the natural increase in energy density predicted by 2020 totally dominates when you are talking about super-mature cylindrical cell production processes.
The real story is, and always has been, what it will take to get Panasonic to take a huge risk on Tesla. The simple answer is, and has been, money. Gobs and gobs of money. A 30% "decrease" in cost per kWh in 2020 implies a major increase in profit margins over what Panasonic is making right now, once you factor in the possibility that the batteries are likely to be 40%+ better.
That story makes sense when you consider the bargaining power of the participants. Panasonic wanting to "help" with power electronics is
exactly like Samsung wanting to "help" with touchscreen technology during the failed negotiations for the 2 billion cell contract for the S Platform.
Essentially they are trying to use their leverage in the battery negotiations in order to capture a larger percentage of Tesla's business, up to and including discussion of power electronics, an area where Tesla likely has greater expertise than Panasonic does. This would allow Panasonic to grow their parts business with Tesla beyond the cell level and into the drivetrain which is Tesla's core area of expertise and value.
The vanilla interpretations of public announcements that are floating around fail to properly characterize the relative bargaining power dynamics of the negotiation, as well as the very well studied rates of improvement in Li-Ion tech likely to be realized by 2020.