Morgan Stanley: XPeng's AI Moat Established; Maintains Overweight with HK$96 Target Price

Overview

Morgan Stanley published a deep-dive review of XPeng Inc., reiterating its "Overweight" rating and HK$96 target price. Although recent stock prices have corrected due to macro uncertainty, the institution believes XPeng's increasingly clear AI technology roadmap is the core catalyst for valuation repricing. The report focuses on the physical world foundation model unveiled at CVPR 2026, noting the company is transforming from a pure EV manufacturer into a "Physical AI" company, with multiple key milestones including Robotaxi and humanoid robots expected in H2 2026.

Core views

AI Technology Architecture Builds Core Barriers: XPeng's VLA 2.0 deeply integrated with world models addresses shortcomings in traditional VLA models regarding implicit world understanding, sparse supervision, and dimensional constraints. Its "trinity" world model includes Active Reasoning (XMind), Controllable Generation (XWord), and Long-horizon Prediction (Foresight), achieving a 2.7x inference speed increase without retraining, laying the foundation for large-scale commercial deployment. The company emphasizes that real-world edge case data is the critical threshold for L4 generalization, reinforcing the necessity of its full-stack in-house development and vertical integration strategy. Dense Catalysts Arriving in H2 2026: The report outlines specific timelines, including the August World Model OTA push, Q3 pilot operations in Guangzhou, and cabin-drive fusion demos showcased later in the year. These events mark substantive progress toward embodied intelligence. Additionally, with the landing of super electric/hybrid systems and L4 autonomous driving, monthly sales are expected to rise significantly starting H2 2026 and further expand in 2027. "Car as Robot" Strategy Unleashes Synergies: XPeng unifies teams through a General Intelligence Control Center (GSC) to accelerate cabin-drive fusion. Due to shared infrastructure, foundation models, and system architectures, models deployed on vehicles require only minimal fine-tuning for application in robotics, with differentiation primarily at the motion control layer. This reuse capability allows new businesses like AI Turing chips, humanoid robots, and Robotaxi to unlock potential valuation upside as they enter mass production and promotion phases in 2026. Financial Outlook and Profitability Inflection Point: With scale effects of new models fully manifesting in 2026, gross margins are expected to improve meaningfully. Volkswagen's technical service revenue will support long-term gross margins, while rising ADAS penetration increases visibility for service revenue growth. In the base case, the company is projected to turn profitable in 2028, driven by economies of scale, ASP growth from overseas sales, and operating leverage release.

Analysis framework

The report adopts a comprehensive analytical framework combining "Technology Verification + Event-driven + Sum-of-the-Parts (SOTP) Valuation." First, it verifies the authenticity of the AI moat by analyzing CVPR 2026 technology releases and company tech day information at the underlying logic level. Second, it maps out specific product and technology implementation schedules for H2 2026 as triggers for short-term stock price repricing. Finally, at the valuation level, rather than relying solely on whole-vehicle PS multiples, it introduces SOTP methodology to separately price AI chips, humanoid robots, and Robotaxi businesses to capture the option value of non-vehicle businesses.

Methodology notes

• Valuation MethodologySOTP Segmental Valuation

  Sum-of-the-Parts Valuation

  For companies with multiple differentiated business lines, this method splits them into independent segments for separate valuation before summing. This report breaks XPeng down into four parts: Vehicles (1.5x P/S), AI Chips (6x P/S), Humanoid Robots (10x P/S), and Robotaxi (10x P/S) to more accurately reflect the standalone value of AI and high-tech businesses and avoid masking them with the lower valuation of vehicle businesses.

• Competition & Strategy FrameworkMoat / competitive advantage

  Data Layers and Vertical Integration as Moat

  The report points out that simulators cannot exhaust all scenarios; therefore, 'real-world edge cases collected by fleet vehicles' become the core barrier for advancing L4 autonomous driving. This dependency on scarce real-world data, combined with full-stack in-house development (chips + data pipelines + AI infrastructure), creates an irreplicable structural competitive advantage.

• Industry/Industrial Analysis Framework

  Physical AI and World Model Paradigm

  Distinct from traditional perception-decision separation architectures, the report focuses on the fusion of VLA (Vision-Language-Action) and world models. This methodology emphasizes that machines must not only 'see' but also understand physical laws and perform long-horizon predictions, serving as key technical indicators for evaluating generational differences in autonomous driving and robotics technologies.

Asset mapping & comparison

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

• XPeng Inc (9868.HK / XPEV.N)

  Core beneficiary: AI technology breakthroughs and new business implementations directly drive valuation repricing

  Strengths

  Leading VLA 2.0 and world models; deep accumulation of real-world data; full-stack in-house chips and infrastructure; Volkswagen technical service revenue supports gross margins

  Weaknesses

  Currently still loss-making; intense competition in medium-to-high-end markets pressuring pricing power; new business commercialization requires time for validation

  Comparison

  Higher AI option value and SOTP valuation elasticity compared to pure automakers; possesses mass-production hardware carriers and cash flow businesses compared to pure AI companies

  Risks

  Intensified competition in the 150k-300k RMB segment leading to market share decline; ADAS penetration falling below expectations; weak macro conditions suppressing industry valuations

Key data

• HK Target PriceHK$96.00Converted from ADR target price using 7.85 exchange rate; corresponds to 1.7x 2026E P/S in base case
• Bull Case ValuationHK$156.00Corresponds to 3.2x 2026E P/S, assuming rapid scaling of new businesses and accelerated sales ramp-up
• Bear Case ValuationHK$39.00Corresponds to 0.8x 2026E P/S, assuming intensified competition and deteriorating macro environment
• 2028E Net ProfitRMB 194 millionProfitability turnaround projected for 2028 under base case (ModelWare basis)
• World Model Inference Speed Increase2.7xAchieved via acceleration technology requiring no retraining; supports commercial deployment

Impact & implications

The report argues that XPeng Inc is undergoing a cognitive shift from a "new car manufacturing force" to an "AI technology company." If the Robotaxi pilot operations and humanoid robot demonstrations in H2 2026 proceed smoothly, the market will no longer value XPeng solely based on automotive manufacturer P/E multiples; the option value of non-vehicle businesses will gradually be realized. Simultaneously, technical service revenue and the volume growth of high-margin overseas models will help alleviate long-standing market concerns about EV companies showing "revenue growth without profit growth," paving the way for the 2028 profitability inflection point.

Risks

• Intensified competition in the medium-high-end sub-segment (150k-300k RMB) may slow market share growth
• ADDS/XNGP penetration growth rates falling below expectations, impacting long-term service revenue potential
• Macroeconomic uncertainty and slowing overall auto industry sales growth suppressing sector valuations
• Delayed improvement in profitability could lead to cash flow pressure

What to watch

• User feedback and experience improvements following the August World Model OTA push
• Actual performance and regulatory progress of the Q3 Guangzhou Robotaxi pilot operations
• Display effects of the H2 2026 cabin-drive fusion demo
• Whether monthly sales take off as scheduled in H2 2026
• Mass production and commercialization pace of new businesses including AI Turing chips and humanoid robots