Battery recycling economics are determined at the sorting stage, not the hydromet plant. A single NMC cell contaminating an LFP black mass stream reduces cobalt and nickel recovery yields in the downstream acid leach — and the problem doesn't become visible until the precursor precipitation step. Niobia AI brings AI-driven chemistry identification and cross-contamination detection to the point where correction is still possible.
Two Processing Routes, One Sorting Problem
Battery recycling runs two main processing routes. Pyrometallurgical smelting is simple and tolerant of mixed chemistries, but energy-intensive and unable to recover lithium economically. Hydrometallurgical processing (black mass → acid leach → selective precipitation → precursor synthesis) recovers lithium, cobalt, nickel, and manganese at high purity but requires chemistry-sorted input streams. A mixed NMC/LFP black mass batch requires separate precipitation steps for iron and phosphate removal that a pure NMC stream does not — adding cost and reducing yield.
The economic case for hydromet depends on receiving chemistry-pure black mass. LFP black mass (iron, phosphate, lithium) has lower intrinsic value than NMC (cobalt, nickel, manganese, lithium) but is growing rapidly as EV volumes shift toward LFP. Cross-contamination of 5% NMC in an LFP stream, or vice versa, can reduce precursor purity below customer specification.
What Needs to Be Sorted and How
Incoming battery packs arrive in three formats (cylindrical, pouch, prismatic) and multiple chemistries (NMC various grades, LFP, NCA, LCO). Discharge before shredding is a safety requirement — incompletely discharged cells risk thermal runaway in the shredder. Chemistry identification uses X-ray fluorescence (XRF) for cobalt, nickel, and iron signature, and computer vision for format classification. Both need to operate at sort speed on a conveyor line.
The hardest sorting problem is distinguishing NMC chemistries (622 vs. 811 vs. 532) from XRF signals alone — the Ni:Co:Mn ratios overlap in the measurement uncertainty of handheld XRF at sort speeds. Niobia AI addresses this by combining XRF with pack-level geometry features from vision to narrow the chemistry identification before the pack-level data is available from a battery passport system.
Where Most Recyclers Get This Wrong
Recycling lines inherit a data problem from the cells they process: chemistry is rarely labeled on the cell itself, and the incoming stream is mixed. Human operators achieve 80–90% accuracy on chemistry identification under ideal conditions; under shift-end fatigue or with unfamiliar pack designs, that drops to 60–70%. At the throughput volumes recyclers need to be profitable — hundreds of tonnes per day — that error rate creates a contamination problem that degrades hydromet yield across entire production shifts.
The second failure is the absence of a yield prediction model that connects incoming black mass composition to hydromet output. Without that model, operators have no way to adjust leach parameters for a lower-cobalt batch until after the precipitation step has already run at sub-optimal conditions.
What AI Process Intelligence Changes
AI-driven XRF + vision sorting achieves 97–99% chemistry ID accuracy at conveyor speed — the threshold needed to protect hydromet yield. Real-time cross-contamination flagging catches misfed cells before they reach the shredder, at the point where diversion costs seconds rather than contaminating a black mass batch worth tens of thousands of dollars.
Hydromet yield prediction from incoming composition — built from 3–6 months of production data — enables dynamic adjustment of leach acid concentration, residence time, and precipitation pH for each incoming batch. The yield improvement from optimised parameters rather than fixed recipes is typically 3–8% on cobalt and nickel recovery, which translates directly to margin at current metal prices.
Battery recycling process intelligence has a shorter payback period than almost any other manufacturing AI application because the value of the recovered material is directly visible in the output. Every percentage point of cobalt recovery improvement has a dollar value that is trivial to calculate — and every contamination event that is prevented has an immediate cost avoidance that justifies the investment.
About the author
Dr. Gaurav Jha is the Founder of Niobia AI. His PhD focused on fast-charging niobium pentoxide (Nb₂O₅) based nanostructured anodes. At Intel he worked on wet etch defect reduction in 5nm and 7nm chip fabrication. He developed one of the first large-scale lithium-sulfur cathode coatings at Lyten, then moved to Sila Nanotechnology for silicon anode particles. He founded Niobia AI to bring manufacturing and materials science experience into an AI platform built for the production floor.