Rob walks through the same calculations I've been doing on capex for battery prod capacity. We both get to $93B for 3 TWh or $31M per GWh annual capacity.
Yeah, I just watched that. I think there's more to the story. GF1 is reported to have cost about $6B in CapEx to date, and is producing just 36GWh (let's say for ease of estimate).
That's $167M/GWh/yr of production capacity. Ugh. Even now Pana adds just 3.5GWh/yr for $100M of CapEx, or ~$28M/GWh/yr capacity. Tesla's bty day slide stated they expect a 69% reduction in CapEx. Panasonic's latest line is already at an 83% reduction over the intial CapEx, so Tesla's slide can clearly not be referring to that comparison.
Roadrunner is all about developing production equipment and tech to build cells from raw materials at the receiving dock to battery cells delivered out the gate. I doubt RR's cost reduction comparison is to the $2T estimate for world-wide battery capacity buildout.
But that's all in the past. I reinterate my statement that,
going forward, Tesla will choose to build new
G-cubes initially in jurisdictions with tax-friendly policies with available low-interest, non-recourse local loans such as the deal Tesla negotiated in China.
In China, Tesla has basically only needed to pay for the production tooling and equipment. Why would they accept a lesser deal in the future? Certainly, this looks to have influenced the recent announcement about Model 3 exports coming from China. It's all about lowest net cost of a vehicle delivered to the hands of a consumer.
But soon, as the number of
G-cubes (an individual unit of a modular bty factory) grows, their growth will be become self-sustaining from FCF. I estimated here on TMC back in Jul 29, 2019 that a 2-yr payback period for CapEx allows Tesla to grow battery production geometrically:
- I predicted that by 2022 Tesla would have bty production capacity of 119 GWh/yr
- I predicted that each G-Cube must be able to produce 25GW/yr in finished bty cells
- G-Cubes must produce cash to pay for another G-Cube in 2 yrs (41% CAGR)
- I predicted the cost to build each G-cube would be $2B, based on $100/kWh product
So I should now revise sales estimates to $50/kWh @ 20 gWh per year for the product, allow for operating expenses (roughly half in raw materials/labor), specify a 2 yr payback for each G-cube, and calculate the max allowable CapEx. Once that's paid for, each G-cube produces roughly $500M in FCF per year. The 2 issues are how fast does it pay for itself, and how much FCF do you need to accumulate to pay the CapEx for the next G-cube (presumably lower CapEx with each iterations). Whew!
All the above assumes high levels of automation for all operational tasks inside the G-cube, with human operators acting mainly as safety, maintenance, and logistics operators. Dojo/AI whittles down that expense each year going forward.
Cheers!