L2++ penetration rose to a new high of 26%, and Horizon led the L2+ market for the first time

Overview

Based on Gasgoo auto insurance registration data, Bernstein conducts quarterly tracking of China’s intelligent-driving chip market. In 2Q26, despite a 22% YoY decline in China passenger-vehicle sales, sales of vehicles equipped with L2+ or L2++ functions still grew 17.6% YoY, and overall NOA penetration rebounded to 42.6%. Growth was mainly driven by the accelerated deployment of urban NOA, with L2++ penetration rising to a record high of 26%, indicating that intelligent driving is shifting from a high-end differentiating feature to a broader mainstream function.

Core views

Over the next one to two years, the explorer camp consisting of BYD, Geely, and emerging local automakers is expected to contribute most incremental L2++ demand and become the core battleground for competition between automakers’ in-house solutions and third-party chip suppliers. Horizon has already achieved leadership in L2+, but its L2++ share remains below expectations; whether it can enter BYD’s expanding urban NOA supply system is the most important upside variable over the coming quarters. NVIDIA still leads L2++ with a 52% installation share, but it is facing challenges from Horizon, Qualcomm, automakers’ in-house chips, and other new entrants. As urban NOA becomes more widespread, chip computing power and semiconductor content per vehicle are expected to rise in tandem, and LiDAR will also benefit.

Analysis framework

The report uses end-vehicle insurance registrations as the basis to separately calculate L1-L2, L2+, and L2++ penetration rates, and measures chip supplier shares by installed vehicle count and total computing power. It also classifies automakers into pioneers, explorers, and traditional automakers to analyze penetration rates, outsourcing preferences, and changes in supplier wallet share across different camps, while assessing the investment value of relevant companies based on product ramp-up, customer structure, and valuation multiples.

Methodology notes

• Industry data trackingEnd-market insurance registration tracking

  Using vehicles sold to end users as the statistical basis

  The data comes from auto insurance registration records published by Gasgoo and, compared with wholesale or chip shipment data, more directly reflects the number of intelligent-driving vehicles that have achieved end-market sales.

• Technology tieringIntelligent-driving level penetration analysis

  Segmenting the market by no assisted driving, L1-L2, L2+, and L2++

  By comparing installed volumes and penetration rates across levels, the report identifies changes in demand, computing power, and semiconductor content brought by the upgrade from highway NOA to urban NOA.

• Competitive landscapeDual-dimensional share analysis by installed volume and total computing power

  Measuring both the number of vehicles installed and installed TOPS

  Installed volume reflects customer coverage and platform scale, while total computing power reflects high-compute chip deployment and potential content per vehicle; the combination is used to assess supplier competitiveness.

• Customer structureAutomaker camp classification

  Classifying automakers into pioneers, explorers, and traditional automakers

  The three types of automakers differ in intelligent-driving penetration, in-house development capabilities, and outsourcing tendencies, which can be used to judge near-term sources of incremental growth and the long-term serviceable market for third-party chip suppliers.

• Valuation methodsEnterprise value-to-sales multiple method

  Multiplying forward revenue by an enterprise value-to-sales multiple

  Horizon’s target price uses a 10.3x enterprise value-to-sales multiple, corresponding to RMB 11.5 billion revenue during the 2BF period; Black Sesame uses a 4.0x multiple, corresponding to RMB 2.3 billion revenue.

• Data limitationsSupply-chain lag adjustment

  Chip shipments lead end-vehicle sales

  After intelligent-driving chips are sold by suppliers to automakers or Tier-1 suppliers, it usually takes approximately 3 to 5 months or about 1 to 2 quarters before they are reflected in end-market insurance registrations. Therefore, the data is more suitable for judging medium- to long-term trends rather than real-time shipments.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Horizon Robotics (9660.HK)

  Core beneficiary

  Strengths

  Has integrated software-hardware capabilities in China’s intelligent-driving chip sector, became the No. 1 player in the L2+ installed-vehicle market with a 32% share in 2Q26, J6B and J6M are beginning to ramp, and its balance sheet can support sustained R&D.

  Weaknesses

  L2++ share fell from 5% to 3%, and HSD model sales performance and high-end design wins have been slower than initially expected.

  Comparison

  Its L2+ share has surpassed NVIDIA, but it still significantly lags NVIDIA’s 52% in L2++.

  Risks

  BYD cooperation or wallet share falling short of expectations, weak HSD model sales, and intensified competition from automakers’ in-house solutions and third-party suppliers such as Qualcomm.

• NVIDIA (NVDA)

  High-end intelligent-driving chip leader and competitive benchmark

  Strengths

  Still maintains a leading 52% share in the L2++ market, with a mature high-compute platform and broad customer base.

  Weaknesses

  Continues to lose share in both L2+ and L2++, and has been surpassed by Horizon in L2+.

  Comparison

  Its leading position in the high-end market is stronger than Horizon and Qualcomm, but it faces dual pressure from new third-party platforms and automakers’ in-house solutions.

  Risks

  Automakers such as BYD shifting to multi-supplier systems, accelerated customer in-house chip development, and substitution by new suppliers.

• Qualcomm (QCOM)

  Intelligent-driving chip supplier with rapidly rising share

  Strengths

  L2+ share rose to 20%; deployment of SA8797P puts it among leading L2++ platforms, and it benefits from volume growth across Leapmotor’s full model lineup.

  Weaknesses

  Current growth remains highly dependent on Leapmotor, with concentrated customer structure and a low base.

  Comparison

  Growth is faster than NVIDIA and Black Sesame, but scale, customer breadth, and L2++ share still need further validation.

  Risks

  Leapmotor sales volatility, slow customer diversification, and high-end solution mass production falling short of expectations.

• Black Sesame (2533.HK)

  Competitive pressure target

  Strengths

  Already has L2+ intelligent-driving SoC products and is attempting to expand into L2++.

  Weaknesses

  Lacks sufficient software capabilities, new customer acquisition and design-win progress are slow, wallet share at Geely continues to flow to Horizon, and it has not yet secured L2++ automaker design wins.

  Comparison

  Compared with Horizon, it is at a disadvantage in software-hardware integration capability, customer expansion speed, and scale.

  Risks

  Cash-flow pressure from sustained R&D investment, equity dilution from frequent financing, customer losses, and failure to commercialize L2++ products.

• Hesai (HSAI.US/2525.HK)

  Beneficiary of urban NOA and rising LiDAR penetration

  Strengths

  Maintains the No. 1 position in China’s LiDAR market with a 47% share, has global leadership, and can benefit from LiDAR configuration rates exceeding 80% in urban NOA models.

  Weaknesses

  Demand remains closely tied to intelligent-vehicle sales, urban NOA configuration strategies, and the pace of vehicle cost reduction.

  Comparison

  Its market share leads RoboSense’s 23%, Huawei’s 21%, and Seyond’s 9%.

  Risks

  Substitution by pure-vision approaches, automaker cost-reduction pressure, price declines caused by competition, and intelligent-vehicle sales falling short of expectations.

• BYD

  Largest future source of incremental urban NOA demand and determinant of the supply landscape

  Strengths

  Is shifting its intelligent-driving focus from L2+ to L2++ and is expected to become the largest source of new urban NOA demand over the next 12 months.

  Weaknesses

  There remains significant uncertainty regarding supplier allocation and in-house chip strategy.

  Comparison

  Compared with other explorer automakers, BYD’s sales scale makes its supplier choices have a greater impact on the market shares of Horizon and NVIDIA.

  Risks

  Urban NOA rollout pace, end-model sales, or supply-chain solution adjustments falling short of expectations.

Key data

• NOA penetration in 2Q2642.6%A significant rebound from 32% in 1Q26.
• L2++ penetration26%Up 7 percentage points QoQ and reaching a record high, the largest contributor to NOA growth.
• L2+ penetration16%Up 3 percentage points QoQ, supported by automakers such as BYD and Geely promoting economical highway NOA solutions.
• L1-L2 penetration37%As consumers migrate to more advanced intelligent-driving functions, the share of basic assisted driving continues to shrink.
• YoY sales growth of NOA models17.6%During the same period, total China passenger-vehicle sales declined 22% YoY.
• Horizon L2+ market share32%Surpassed NVIDIA’s 28% for the first time, becoming the market leader by installed vehicle count.
• Qualcomm L2+ market share20%Mainly driven by Leapmotor model sales and deployment ramp-up.
• NVIDIA L2++ market share52%Still ranks first, but automakers’ in-house solutions and emerging third-party suppliers are intensifying competition.
• Horizon L2++ market share3%Down from 5% in the previous quarter, with design-win progress slower than expected.
• LiDAR penetration20%Rebounded from approximately 15% in 1Q26; more than 80% of urban NOA models use LiDAR.
• Hesai LiDAR market share47%RoboSense, Huawei, and Seyond were 23%, 21%, and 9%, respectively.
• Horizon target priceHKD 10.00Based on a 10.3x enterprise value-to-sales multiple and RMB 11.5 billion revenue during the 2BF period.

Impact & implications

China’s intelligent-driving demand is accelerating its shift from basic L1-L2 and highway NOA to urban NOA, driving higher content per vehicle for high-compute SoCs, domain controllers, and LiDAR. Explorer automakers will dominate near-term incremental growth, while traditional automakers, due to weaker in-house development capabilities, will become the longer-term serviceable market for third-party intelligent-driving chip suppliers. For Horizon, its L2+ leadership provides a growth foundation, but further valuation upside requires validation of BYD cooperation, HSD model sales, and L2++ share; for NVIDIA, its L2++ leadership remains, but customer multi-sourcing and automaker in-house development will continue to erode its share.

Risks

• Auto insurance registration data has a lag of approximately 3 to 5 months or 1 to 2 quarters versus actual chip shipments, so short-term share changes may not reflect suppliers’ operating conditions in real time.
• Continued weakness in China passenger-vehicle demand could drag down the absolute installed volume of intelligent-driving models.
• Urban NOA penetration speed or consumers’ willingness to pay may be lower than expected.
• Automakers expanding investment in in-house chips and algorithms may compress the serviceable market and profit margins of third-party SoC suppliers.
• Horizon may fail to secure meaningful BYD L2++ wallet share, or HSD model sales may perform poorly.
• Intensified competition in the chip industry could bring pricing pressure, higher R&D investment, and product iteration risks.
• Smaller suppliers such as Black Sesame may face financing and equity dilution pressure due to sustained R&D burden.
• LiDAR cost reduction, reliability validation, or model configuration-rate increases may fall short of expectations.

What to watch

• Further disclosures in 2H26 on Horizon’s cooperation with BYD regarding HSD and urban NOA.
• Horizon’s wallet share at BYD, Geely, and Chery, as well as the installation ramp of J6B, J6M, and J6P.
• Whether Horizon’s L2++ share can rebound from the current 3%, and whether its L2+ share can approach the 55% to 60% potential implied by management’s target.
• The actual allocation as BYD shifts from NVIDIA dominance to a diversified supply system of NVIDIA, Horizon, and in-house chips.
• Mass production and customer diversification progress for Qualcomm SA8797P beyond Leapmotor.
• The pace at which Li Auto M100, Xheart X7, and other automakers’ in-house chips substitute third-party suppliers.
• Subsequent increases in L2++ penetration among explorer automakers and NOA penetration among traditional automakers.
• LiDAR configuration rates in urban NOA models and whether Hesai can maintain its 47% market-leading share.