Supply Chain Overview > Supply Chain Convergence Map
Supply Chain Convergence Map
Every major electrification domain — electric vehicles, humanoid robots, grid-scale battery storage, AI datacenters, and industrial electrification — draws from the same upstream supply chains. The chokepoints in those supply chains are not sector-specific problems. They are shared constraints that simultaneously limit EV production, robot deployment, BESS buildout, and datacenter power delivery.
This page maps those convergence points at two levels: critical materials at the processing and refining tier, and strategic finished components at the system integration tier. Both tables represent the first systematic cross-domain chokepoint mapping of this kind in public content.
The organizing principle: supply chain risk is not determined by mining concentration — it is determined by processing and refining concentration. China mines a minority of the world's lithium but refines the majority into battery-grade material. The chokepoint is not the mine. It is the refinery.
Critical Material Chokepoints
Concentration percentages reflect processing and refining dominance — the tier where raw ore becomes usable industrial input. Mining concentration data is noted separately where it differs materially from processing concentration.
| Material | Processing / Refining Concentration | Affected Domains | Primary Chokepoint | Substitutability | Risk Level |
|---|---|---|---|---|---|
| Battery-grade lithium (Li) | China ~80% of global processing | EV, BESS, consumer electronics, humanoid robots | Lithium hydroxide and carbonate refining — not mining | Low near-term — sodium-ion emerging but not at scale | Very High |
| Spherical graphite (anode) | China ~90% of global spherical graphite production | EV, BESS, humanoid robots | Purification and spheronization — capital intensive, China-dominant | Low — silicon anode partial substitute, not equivalent | Very High |
| Battery-grade nickel (Ni) | Indonesia dominant mining; China ~70% of Class 1 nickel processing | EV (NMC), BESS (NMC), industrial | High-purity Class 1 nickel refining for cathode active material | Medium — LFP chemistry reduces nickel dependency | High |
| Cobalt (Co) | DRC ~70% mining; China ~80% refining | EV (NMC/NCA), consumer electronics, aerospace | DRC artisanal mining + China refining dual concentration | Medium-High — cobalt-free LFP and LMFP chemistries scaling | High — declining as LFP scales |
| NdFeB permanent magnets (REE) | China ~90% of global NdFeB magnet production | EV traction motors, humanoid joint motors, wind turbines, industrial motors, robotics | Neodymium and dysprosium separation + magnet sintering — vertically integrated in China | Low — induction motors avoid REE but sacrifice efficiency and power density | Very High |
| SiC wafers | US (Wolfspeed), Europe (Infineon, STMicro), Japan (Rohm) — Western-controlled but physics-limited supply | EV traction inverters, DCFC, BESS PCS, solar inverters, humanoid joint drives, datacenter power, industrial VFDs | Boule crystal growth yield — weeks per crystal, cannot be accelerated by capital alone | Very Low — no material equivalent for high-voltage high-frequency power switching | Very High — compound demand from all domains simultaneously |
| GaN epitaxial wafers | GaN-on-Si: standard silicon fabs; GaN-on-SiC: limited — Wolfspeed, MACOM | Humanoid joint drives, LiDAR pulsers, datacenter PoL converters, fast chargers, AV compute power delivery | GaN-on-SiC epitaxy for highest-performance applications | Low for MHz-range switching applications — no silicon equivalent | High — demand accelerating faster than epitaxy capacity |
| Copper (Cu) — refined | Chile dominant mining; China ~40% of global refined copper | EV wiring harness, motors, transformers, BESS busbars, grid T&D, datacenter power distribution, charging cables | Volume constraint — EVs use 3–4x copper of ICE vehicles; grid modernization adds structural demand | Low — aluminum partial substitute for some wiring, not for motors or high-current busbars | High — demand volume, not concentration |
| Grain-oriented electrical steel (GOES) | Japan (Nippon Steel), US (AK Steel), China (Baowu) — oligopoly of ~6 producers globally | Grid transformers, solid-state transformers, motor laminations, BESS magnetics | Specialty steel production capacity — 18–24 month lead times for large power transformers | Very Low — amorphous core partial substitute for some applications | High — grid modernization and datacenter transformer demand converging simultaneously |
| Battery-grade cathode active material (CAM) | China ~80% of global CAM production (NMC and LFP) | EV, BESS, consumer electronics | Precursor cathode active material (pCAM) processing — China dominant at every upstream step | Low near-term — Western CAM capacity scaling under IRA incentives but 3–5 year lag | Very High |
Strategic Finished Component Chokepoints
Finished component chokepoints operate at a different level than material chokepoints — they represent concentration in manufactured systems and sub-assemblies that require integrated design, process, and supply chain capability beyond raw material processing. These are harder to diversify because they require industrial know-how, not just capital.
| Component | Geographic Concentration | Affected Domains | Why It's a Chokepoint | EV Equivalent | Severity |
|---|---|---|---|---|---|
| Battery cells (LFP) | China ~85% — CATL, BYD dominant | EV, BESS, humanoid robots, marine | LFP is Chinese-originated chemistry with vertically integrated production — IP, materials, manufacturing all China-concentrated | Direct — same cells in EV packs and BESS systems | Very High |
| Cathode active material (CAM) | China ~80% | EV, BESS, portable | Upstream of cell manufacturing — controls chemistry performance and cost at the precursor level | Direct | Very High |
| Traction motors (PMSM) | China dominant — motor manufacturing and NdFeB magnet integration | EV, humanoid robots, industrial, wind, marine | China controls both the magnet supply and the motor manufacturing — vertical integration creates dual-layer concentration | Direct — same motor architecture in EVs and humanoids, different scale | Very High |
| Harmonic drive reducers / strain-wave gearboxes | Japan (Harmonic Drive Systems, Nabtesco) dominant; China scaling rapidly | Humanoid robots, quadrupeds, industrial robots, surgical robots | Precision gear ratio at compact scale requires decades of manufacturing process knowledge — not replicable quickly with capital alone | No EV equivalent — unique to robotic joint architecture | Very High |
| Actuator modules (integrated joint units) | China scaling rapidly — Unitree, Fourier, INNFOS; Western: limited production | Humanoid robots, quadrupeds, exoskeletons | Integrated motor + gearbox + encoder + driver in one unit — China achieving cost positions 60–80% below Western equivalents | No EV equivalent — new supply chain category | Very High |
| Tactile sensor arrays | No established production-scale supplier globally | Humanoid robots — hand and fingertip dexterity | Supply chain does not exist at production scale for any humanoid platform — dexterous manipulation blocked by sensor availability | No EV equivalent — nascent category | Critical — supply chain nascent |
| SiC power modules (traction-grade) | US, Europe, Japan — Western-controlled but capacity-constrained | EV traction inverters, DCFC, BESS PCS, solar inverters, industrial VFDs, SST | Crystal growth physics limits supply expansion — compound demand from all domains simultaneously outpacing capacity additions | Direct — traction inverter core component | Very High |
| Utility-scale power transformers | ~6 global producers — ABB, Siemens, GE, Hitachi, SPX, Prolec; 18–36 month lead times | Grid modernization, datacenter grid interconnect, gigafactory utility service, BESS grid-tie | Oligopoly production + long manufacturing cycle + simultaneous demand surge from grid modernization and AI datacenter buildout | Indirect — required for EV charging infrastructure at scale | Very High — active supply crisis 2024–2027 |
| HBM memory (AI compute) | SK Hynix ~50%, Samsung ~30%, Micron ~20% | AI datacenters, autonomous vehicle compute, humanoid robot AI inference modules | 3D stacked memory production requires advanced packaging capability concentrated at 3 producers — NVIDIA GPU production directly constrained by HBM allocation | No direct EV equivalent — AV and robot compute dependency | High — active constraint on AI chip production |
| Advanced packaging (CoWoS, HBM integration) | TSMC dominant for CoWoS; Samsung and ASE for alternatives | AI datacenters, AV inference compute, edge AI modules | AI chip performance gains increasingly driven by packaging architecture — TSMC CoWoS capacity directly gates NVIDIA H100/H200/B200 production | No direct EV equivalent — AV and datacenter compute dependency | High |
| BESS power conversion systems (PCS) | China dominant at cell level feeding PCS; Sungrow, CATL, BYD leading PCS producers | Grid storage, datacenter backup, FED energy management, microgrid islanding | PCS integrates SiC power electronics, controls, and grid interface — China dominance at cell level extends into system integration | Analogous to traction inverter — DC-AC conversion at system level | High |
| Electric motor controllers / VFDs (industrial) | ABB, Siemens, Danfoss, Yaskawa — established but SiC transition creating new concentration points | Industrial electrification, port equipment, mining, HVAC, pumps and compressors | SiC transition in VFDs creates new SiC module dependency — same supply constraint as traction inverters | Analogous to traction inverter — different voltage and duty cycle profile | Medium-High — SiC transition accelerating |
Key Convergence Observations
The processing tier is the real chokepoint — not mining. China mines a minority of global lithium but refines the majority. The Democratic Republic of Congo mines most of the world's cobalt but China refines most of it into battery-grade material. Diversifying mining without diversifying processing does not reduce supply chain risk.
Humanoid robots introduce entirely new chokepoints with no EV equivalent. Harmonic drive reducers, integrated actuator modules, and tactile sensor arrays do not exist in EV supply chains. The embodied AI supply chain is a parallel superset of the EV supply chain — not a derivative. Its most critical components have no precedent production base outside Japan and China.
SiC is the only component appearing across every electrified domain simultaneously. EV traction inverters, DCFC chargers, BESS power conversion systems, solar inverters, humanoid joint drives, datacenter front-end rectifiers, industrial VFDs, and solid-state transformers all require SiC power modules. Compound demand from all domains simultaneously creates a supply pressure no single domain analysis captures.
The transformer crisis is the most underreported near-term constraint. Utility-scale power transformers have 18–36 month lead times from an oligopoly of six global producers. Grid modernization, AI datacenter interconnects, gigafactory utility service, and EV charging infrastructure expansion are all simultaneously increasing demand. This constraint gates electrification deployment more immediately than any material shortage.
Western policy addresses the visible chokepoints — not the structural ones. The US IRA and EU Net-Zero Industry Act incentivize battery cell manufacturing and critical mineral processing. Neither addresses harmonic drive reducers, actuator modules, or tactile sensors — the components where Chinese and Japanese producers have structural advantages that capital alone cannot overcome on a 5-year policy horizon.
Relationship to the Six Autonomy Framework
The two tables above map directly onto the first two layers of the Six Autonomy Framework. Table 1 corresponds to Materials Autonomy — the freedom from critical materials concentration that determines whether production can be sustained when upstream supply tightens. Table 2 corresponds to Silicon Autonomy and the broader component autonomy question — freedom from finished component concentration at the system integration tier.
An organization or system that achieves MA-3 and SA-3 status under the Six Autonomy Framework has resolved the chokepoints in both tables for its critical operational inputs. No organization has achieved this across all domains simultaneously. Tesla's stack — mapped in the Tesla Six Autonomy Case Study — represents the most complete attempt currently underway.
See also: Supply Chain Overview · Battery Supply Chain · Motor & Drivetrain Supply Chain · Power Electronics Supply Chain · Humanoid Robot Supply Chain · Six Autonomy Framework
