So the new @BloombergNEF battery _price_ report has come in, and the goal of @ARPAE GRIDS has been hit, assuming that a pack _price_ $137/kWh is close to a pack _cost_ of $100/kWh. This is a big deal.

https://arstechnica.com/science/2020/12/battery-prices-have-fallen-88-percent-over-the-last-decade/

But _why_?

(disclaimer: I don't know) 1/herewego

Is it because of (waves hands wildly) "China"? I have taken this as shorthand to say, engineering at breakneck speed to market-acceptable acceptable quality (i.e. "good enough"). Advances skipping the "premium", going straight to high volume / low margin markets? 2/n

Is it because NMC was so road-mappable that it provided a clearly analog way of reducing cost while increasing performance as Ni goes up and Co goes down? 3/n

Is it because the spectre of the EV s-curve drove scale to get the learning curves in place to motivate incredible acceleration of 1 + 2? 4/n

Because AFAICT, if I time travelled back to 2005 and showed myself a modern NMC811 battery, nothing in the composition would be too surprising, I'd just look at a 50 Ah format and say "are you sure this thing won't blow up?" 5/n

When I (and a bunch of other people) embarked on $100/kWh projects in 2010, the implicit assumption is that what I described above simply could not exist _safely and reliably_ and TBH I bet too much of my tenure motivation on this (now clearly faulty) premise 6/n

But perhaps all the above is the tactical why? What are the lessons learned? Well 7/n

Energy density & cost _can_ positively couple. Before 2010 this wasn't realized, but after, assuming no using Au (& less Co), the more energy we can store per unit mass, assuming all that mass roughly costs the same price, the less mass we need, the cheaper the battery is. 8/n

What's happened with power density is amazing: cost has tracked down while power density has slowly increased. To first order porous electrode theory says these should be negatively coupled. 9/n

What this means to me is that strong use of DFN (and beyond) has really worked to _power optimize the hell_ out of energy density maximizes electrodes. The rate capability that has been wrung from dense, thick MO cathodes is amazing and surprises by 25 year old self. 10/n

Which brings us to LFP. LFP entered life as a chemistry productized to realize _power premiums_, (these premiums were never captured). While its power performance is astounding, it finds itself in roles increasingly for its low _cost of energy_ (Fe, P & O are cheap). 11/n

Oversimplifying: MO's got to sufficient power faster than LFP could capture that premium, so LFP lost round one, but emerged as a value play. Amazing. I'll be spicy and say compare A123 of 2008 to QS today. (not really fair b/c A123 had very good commercial cells by '08). 12/n

Clear losers? Spicy again: Flow batteries. The paper advantaged of energy/power decoupling has yet to be realized at costs that beat current power derated energy/power coupled systems. Again, see LFP. I didn't see this coming, my 2008 self is reeling a bit. 13/n

That and the O&M simplicity of swapping cells after X years generally beats routine and more expensive maintenance on flow batteries. This is not speculation in 2020, this is empirical. Again, it shouldn't be, but it is. LFP is just that cheap, at least to 6 hours. 14/n

I can't overstate this. The asymptotic line for duration on energy/power coupled systems in 2010 was routinely cited as 4 hours. I did this as well. WRONG. Luckily I smelled this coming in 2012 and shifted research priorities. 15/n

This boils down to two aspects that are surprising to my 2008 self and one that is confirming: one I should have seen coming, one that I'm allowed to be surprised by still, and one that was becoming apparent but is still amazing. 16/n

The first is that battery cost battery asymptotes at materials cost. Processing cost goes to zero at large enough scales. The incorrect floors for Si PV in the early 00's presaged this, and I should have seen this coming for batteries when PV prices dropped in 08. 17/n

This I attributed to my academic bias training as a "process engineering person" (I'm worth more than $0 right!). I've come to understand since 08 that it's my job to teach people to drive the cost down to zero and move on. 18/n

The second is the amazing power capability of the lithium ion battery: nearly unlimited at the particle level, and surprisingly flexible at the pack level. Need a "free" charge acceptance boost? Warm to 40˚C. "D" goes up by almost an order of magnitude. Incredible still 19/n

Finally: relentless and ever improving SEI (and now CEI) engineering, I hedged on this in 2010 but it's clear that interfacial stability can be tuned to "good enough" even for systems pretty far from equilibrium. 20/n

The shotgun approach of the 00's has given way to ML/AI (which is just the computer shooting the gun, hot take), but we actually have interfacial understanding coming in, which is incredible. 21/n

Anyhow, all of these, and probably a bunch of other things I missed, combined in a "perfect oasis" to get to where we are in terms of cost. A fantastic amount of collaborative engineering world wide got us there. 22/n

But the bounding system of lithium ion has had flexibility akin to that of Si for microelectronics, and has proved to be an amazing canvas. Hitting these numbers a year after the Nobel Prize was award only confirms its significance. 23/n

FINALLY: what remains to be seen is how learnings in the decosting of lithium ion apply to other chemistries? Can they? Do they even matter? Should we reset the bar to $10/kWh for 2030 to see? Why not. n/n