Why Supply Can't Catch Up
Demand for battery metals can rise in a few years. The two other sources of supply cannot follow it that fast.
These are teaching toys, not research. The first is a deterministic stock-and-flow sketch with order-of-magnitude parameters; the second is calendar arithmetic on a published average. Neither represents a real supply chain, and no number either produces should be cited. Each exists to make one mechanism visible.
A · The loop back arrives late · B · Too late to decide
A · Recycling
The loop back arrives late
Battery demand → critical mineral demand → primary mining and end-of-life recovery. Move the knobs.
What this is showing
The recovered-supply band is the whole point. Its size in any given year is set by how many packs were built ten to fifteen years earlier, multiplied by how many of them are actually collected and how much metal the process recovers. Those are decisions made long before the shortfall appears.
Try it: push demand growth up until a gap opens, then try to close it by raising the recovery yield. You cannot, not quickly: the material simply is not in the waste stream yet. Now lower the pack lifetime and watch the loop arrive sooner but carry less. Then set LFP share to 100% and watch the nickel and cobalt problems vanish while the lithium problem does not.
That is the argument for studying these systems end to end and back rather than one stage at a time. A model of mining alone cannot see the recovery loop; a model of recycling alone cannot see what feeds it. The interesting behaviour lives in the coupling, and it is a decade-scale lag.
B · New mines
Too late to decide
If recycling cannot close the gap in time, the other door is a new mine. A mine takes about sixteen years. Pick a year you want metal, and see when the decision had to be made.
Show the numbers
What this is showing
S&P Global put the average at 15.7 years from discovery to production for mines that opened in the 2010s, and the IEA uses a comparable figure. Sixteen years is longer than most policy horizons, longer than most political terms, and longer than the interval between a target being announced and the year it falls due.
Which produces the result worth sitting with: for any target inside the next decade and a half, the decision is already behind us. A 2030 target needed work under way in 2014. A 2035 target needed 2019. The date does not move because the target is important.
Try it: drag the target year toward the present and watch the required schedule slide backwards past the dashed line marking today. Then push Permitting to its upper end, the widest stage and where most of the difference between two otherwise identical projects lives.
Read together
Panel A says recovery cannot ramp; panel B says new mines cannot arrive in time. What is left inside the window is expansion of mines that already exist, substitution toward materials that are easier to get, or less demand, and how much is needed in the first place depends heavily on which technologies get built.
What they leave out
Almost everything that makes supply hard. Projects fail, and most discoveries never become mines: the sixteen-year average is drawn from the ones that made it, which flatters it. There is no price signal, and price is what actually starts and stops these projects. No community consent, no water, no tailings, no ore body worse than the drill core promised. Expansions of existing mines are faster than greenfield and are not modelled, and neither is the chance that permitting reform shortens the middle of the chain. The recovery model does not say where primary supply comes from; it arrives as a line on the chart.
Recovery model: src/scripts/linkedModel.js, about eighty lines,
all parameters at the top. Lead-time model: src/scripts/leadTimeModel.js,
with stage durations in src/scripts/leadTimeData.js, each
labelled with its source. The 15.7-year total is from S&P Global Market
Intelligence; the stage split is an order-of-magnitude apportionment made
for this page that sums to it.