Geely’s recall of 93,000 vehicles across its Galaxy and Lynk & Co brands exposes a systemic vulnerability in the automotive-grade power electronics that feed LiDAR sensors – a defect that disables driver-assistance features and signals deeper supply-chain stress as Chinese EV makers race to standardize laser-based perception. The fault, traced to process variation at a LiDAR supplier that damages the unit’s power-management chip, highlights how the push for L2+ autonomy is outpacing the maturity of the high-voltage, high-reliability silicon that makes those sensors work.
Why LiDAR Power Architecture Matters for EV Range
LiDAR units in current Chinese production vehicles – primarily spinning mechanical or MEMS-based designs from Hesai, RoboSense, and Livox – typically draw 12-20 W each at 12 V or 48 V, fed through a dedicated DC-DC converter and power-management IC (PMIC) that steps down from the vehicle’s high-voltage bus. A single front-facing long-range LiDAR plus two or three blind-spot units can add 40-60 W of continuous load. Over a 600 km WLTP cycle, that translates to roughly 0.8-1.2 kWh of energy consumption, or 1.5-2.5 % of a 75 kWh pack – non-trivial for OEMs chasing every kilometer of certified range.
The recalled Geely models (Galaxy M9, Lynk & Co 900, 10 EM-P, 07, 08) each carry at least one roof-mounted long-range LiDAR and multiple side units. The defective PMIC sits inside the LiDAR housing, converting the vehicle’s 12 V or 48 V supply to the 3.3 V and 1.8 V rails that power the laser driver, photodetector array, and FPGA. If the PMIC fails open, the entire sensor goes dark; the vehicle’s sensor-fusion stack then disables Navigate-on-Autopilot (NOA) and advanced emergency braking, reverting to camera-only fallback. That fallback is precisely what Chinese regulators and NCAP protocols increasingly penalize.
Geely has not named the LiDAR supplier, but the affected model mix points to Hesai AT128 or RoboSense RS-LiDAR-M1 units, both of which integrate the PMIC on the sensor’s internal PCB rather than relying on a centralized domain-controller supply. That architectural choice – distributed power conversion at each sensor – reduces harness weight and voltage-drop losses but multiplies the number of automotive-qualified PMICs in the bill of materials. A single process drift at the PMIC fab or assembly house can therefore brick an entire model line’s autonomy stack.
Cross-Cutting Analysis: Semiconductor Supply-Chain Stress Meets Autonomy Arms Race
That points to a broader collision: Chinese EV brands are committing to LiDAR-as-standard across mid-range models (150,000-300,000 RMB) two to three years ahead of European and U.S. peers, yet the automotive-grade PMIC supply base remains concentrated in a handful of fabs – primarily Texas Instruments, Infineon, Onsemi, and a rising cohort of Chinese foundries (SMC, Sinochip) still climbing the AEC-Q100 qualification curve. In 2024, China’s domestic PMIC self-sufficiency for automotive was roughly 35 % by revenue, with the highest-reliability grade (Grade 0, -40 °C to 150 °C) still heavily imported.
If this trend holds, every new LiDAR-equipped model adds 4-8 additional Grade-0 PMICs per vehicle. At Geely’s 2024 volume of ~2.1 million units, a 30 % LiDAR attach rate implies ~2.5 million PMICs annually just for LiDAR – before counting domain controllers, zonal gateways, and 48 V mild-hybrid systems. A single fab excursion at a key supplier (e.g., TI’s 300 mm line in Chengdu or Infineon’s Villach facility) could constrain LiDAR deliveries across multiple OEMs simultaneously, creating a choke point analogous to the 2021 MCU shortage but narrower and harder to substitute because PMICs are co-qualified with the LiDAR module’s safety case (ISO 26262 ASIL-B or ASIL-D).
By comparison, Tesla’s vision-only FSD stack avoids this entire power-electronics chain, saving an estimated 30-50 W per vehicle and eliminating a qualified-supplier dependency. The energy-efficiency gap compounds: a Model Y Long Range (75 kWh usable) achieves ~185 Wh/km on the China CLTC cycle; a LiDAR-equipped Lynk & Co 07 with similar battery capacity averages ~195 Wh/km in independent testing. Roughly 5-7 Wh/km of that delta is attributable to the LiDAR suite’s sensor and compute load. Over a 15-year, 300,000 km lifecycle, that’s 1.5-2.1 MWh of extra electricity – equivalent to 20-30 full battery cycles – purely for perception redundancy that may be disabled by a $2 PMIC failure.
The recall also illuminates a regulatory feedback loop. China’s 2025 revision of GB/T 38056 (automotive functional safety) and the upcoming “Intelligent Connected Vehicle” mandatory standards (draft MIIT 2024) will require OEMs to demonstrate sensor-fusion fallback performance and supply-chain traceability for safety-critical components. Geely’s recall remedy – likely a firmware update that reconfigures the PMIC’s over-voltage/under-voltage thresholds plus a selective hardware swap for units outside spec – will become a template for how Chinese regulators evaluate over-the-air (OTA) remediation versus physical recall. If the fix rate exceeds 95 % via OTA, it could accelerate acceptance of software-defined recall pathways, reducing the energy and logistics cost of future campaigns (each physical recall visit consumes ~15 kWh per vehicle in transport and shop energy).
Who This Affects
- EV battery/range engineers: Re-evaluate LiDAR power budgets in next-platform architectures; consider centralized 48 V zonal supply with qualified bulk DC-DC to reduce per-sensor PMIC count and improve fault containment.
- Semiconductor procurement managers: Audit PMIC qualification status for all LiDAR suppliers; negotiate dual-source agreements for Grade-0 PMICs and require wafer-lot traceability to fab level for safety-critical rails.
- Autonomous-driving system architects: Model sensor-fusion degradation paths when LiDAR drops offline; validate camera-radar fallback performance at night and in rain to meet 2025 CNCAP/GB standards without LiDAR.
- Policy analysts tracking China EV regulations: Monitor MIIT’s response to this recall for signals on OTA remediation acceptance, mandatory LiDAR durability testing, and domestic PMIC substitution timelines.
- Investors in LiDAR pure-plays (Hesai, RoboSense, Innoviz): Factor recall-related warranty reserves and potential OEM pressure to shift PMIC integration to Tier-1 domain controllers, which could compress sensor-module ASPs by 8-12 %.
What to Watch Next
- Geely’s 30-day recall completion rate and OTA fix deployment metrics – reported to SAMR (State Administration for Market Regulation) – will indicate whether software-only remediation is viable for PMIC threshold drift.
- Hesai and RoboSense Q3 2026 shipment guidance revisions – any downward adjustment signals broader PMIC yield issues across their customer base (Li Auto, NIO, XPeng, Zeekr).
- Infineon and TI automotive PMIC lead-time data from distributors (DigiKey, Mouser, Arrow) – extension beyond 26 weeks would confirm supply-chain tightening.
- MIIT’s final “Intelligent Connected Vehicle” mandatory standard (expected H1 2026) – watch for clauses on sensor-power redundancy, OTA recall eligibility, and domestic-chip content requirements.
- Solid-state LiDAR (OPA, FMCW) pilot production launches – these architectures typically integrate power management on-chip (SiP), potentially eliminating the discrete PMIC failure mode entirely.
Bottom line: The Geely recall is not an isolated quality escape – it is the first visible fracture in a power-electronics supply chain that Chinese EV makers have stretched to its limit in the race to make LiDAR standard equipment. Until automotive-grade PMIC capacity catches up to the sensor-volume curve, every new LiDAR-equipped model carries a latent energy-efficiency penalty and a systemic recall risk that vision-first architectures simply do not face.
Read the full report at CnEVPost
Note: facts and figures attributed above to CnEVPost (China EV & new-energy industry) reflect that outlet's original reporting. Broader context, cross-sector connections, and forward-looking scenarios reflect independent analysis by our editorial team.
About this article: Drafted by Energy Ai with AI-assisted research and writing based on public reporting, then reviewed under our editorial process before publication.
Leave a Reply