The Cost to Time Shift Energy

The United States produces about 4.2 PWh (1 petawatt is 1x10^15) of electricity every year. Per day, that is about 12 TWh. Let’s say we wanted to time shift 25% of that energy from one point of the day to another, twice per day. For example, your wind farm output might be sagging in the early morning, and your solar output isn’t expected to ramp for another 4 hours or so. And then again in the afternoon.

That would require us to have 3 TWh of storage that went through two cycles each day.

Tesla’s Gigafactory is currently producing about 23 GWh of cells every year. Each cell can deliver 100% discharge for about 500 cycles. If you only discharge from 90% to 40%, though, you can get a thousand or more cycles. But let’s say the batteries will get discharged to the point that will yield exactly 730 cycles (365 * 2) and that will still give us 23 GWh of storage. It’s optimistic, but good enough for now.

To rely on cells to provide time-shifting for this 3 TWh of storage would require 125 Gigafactories, with 100% of their output making cells for the grid 24x7. And because the batteries would be used up after a year (700 cycles) you’d need to have those 125 factories making replacement batteries continuously.

A single 19 WH 21700 cell is about 70g. This means we’d need about 150B cells made per year, and about 11B kg of finished goods. Mining has roughly a 50:1 return by weight (gold is 150K:1, copper is 10:1, coal is very close to 1:1), which means you’d need to mine about 530M tons of earth, to begin the process of making these cells. Mining costs for ore in general are perhaps $10/ton.

The US mining industry processed 1200M tons of coal last year.So, the mining effort for batteries would be on par with that of coal today. And just as we see train cars chugging across the Midwest carrying an endless chain of coal, we’d see those same train cars carrying the other raw materials needed to make the cells. Non-stop. 24x7. Every year.

If the mining cost is about $5B annually, and the selling price of these cells at $100 kWh would be $287B/year. Thus the mining cost would be about 2% of the delivered cell cost.

For comparison, hydro storage would be similarly challenging. Hoover Dam produces about 4 TWh of energy each year. We’d need 365 strategically located Hoover Dams to time-shift the energy target above. That has a much smaller chance of happening than a lot of Gigafactories.

How about the cost adder? If we must spend $287B/year on cells (at $100/kWH) to backup the grid that provides 4.2PWH/year, that means a cost adder of $0.068 per kWH for a few hours of backup. That’s a lot. In practice, you’d need days or weeks of storage OR you’d need a parallel natural gas system that was dormant 99% of the time.

Other Cell Chemistries

Thus far, the cost adder assumes cells with killer volumetric energy density (Tesla’s NCA chemistry). But while that cell is great for cars, it might not be the best for grid storage. If you can devote more space to grid storage, then there are likely other technologies that can deliver the storage at a better cost. For example, LFP can deliver 5-10X more cycles than NCA, but it’s cost per KWH is perhaps 50% greater. But over it’s life time, it’s cost per delivered KWH is 1/5 to 1/4 that of NCA. This is where “flow” batteries come into play. This technology could be built as part of a building’s infrastructure during construction, and with potentially 10K or 20K cycles it could deliver costs 1/10th that of NCA.

But then why look at NCA? Because it’s here today, at scale, makes for an easy case study and the result can be quickly scaled for other technologies. If Tesla’s cells are delivering grid storage for $0.068 per kWH, and something comes along with 1500 cycles, then the cost falls to 0.034 per kWH. If flow batteries can deliver 20,000 cycles at the same cost of NCA, then the cost for grid storage falls to 0.0034 per kWH.


Time-shifting energy via batteries would be a massive undertaking. It’s doable, but it’s not something that could be started anytime soon. Just the mining aspect alone would prompt decades of fights in the court. That battle has already begun.

The cost for a few hours of nationwide backup would be staggering, easily doubling the wholesale cost of energy. And for events such as a polar vortex, where many wind farms were forced to shut off due to the low temperatures, the problem is even more vexing. It suggests that we will need a parallel gas network ready to handle the shortfall at any time. Thus, when consumer demand prompts us to build an extra 1 GW of generation, it will consists of 1 GW of renewable generation, backed by 1 GW of conventional generation. And that conventional generation will be largely unused 99% of the time. But when it is needed, it will be needed for life and death events. Without it, anyone that relied solely on renewable to make it through a polar vortex event would die.

A final thought: The US drives about 3T miles per year. At 300 wh/mile, that’s another 900 TWh of electricity needed, or about 20% of our current consumption. In other words, the demand side will be going up a fair bit over the next decade or two.

Puget Sound Energy Fuel Mix

Tesla and Motor Metrics