Li Auto CEO: Cars Aren't Phones, Safety Can't Be Rushed

Li Auto CEO Pushes Back on Product Cycle Critics

Li Auto CEO Li Xiang has fired back at critics questioning the company's 4-year development cycle for the new L9 Livis, arguing that automotive development fundamentally cannot — and should not — mirror the rapid iteration cycles of the smartphone industry. In a public statement, Li emphasized that cars are life-safety products requiring exhaustive physical testing that no amount of AI simulation can fully replace.

The response comes amid growing debate in China's fiercely competitive EV market, where consumers and analysts increasingly expect Tesla-like over-the-air updates and shorter refresh cycles. Li's defense highlights a tension that extends far beyond one Chinese automaker — it touches on the broader question of how AI and software-defined vehicles are reshaping expectations for the entire automotive industry.

Key Takeaways

• Li Auto's new L9 Livis follows a 4-year product development cycle, which some critics call too slow
• CEO Li Xiang argues automotive safety validation requires extensive real-world road testing that cannot be compressed
• AI tools have limited effectiveness in replacing physical crash tests, durability trials, and real-road validation
• The company must invest significant time acquiring high-value road test vehicles before mass production
• The debate reflects broader industry tension between software-speed expectations and hardware-safety realities
• Li Auto's stance contrasts with smartphone makers like Apple and Xiaomi, which refresh products annually

Why the 4-Year Cycle Sparked Controversy

China's EV market moves at a blistering pace. Companies like BYD, NIO, and Xpeng routinely announce new models, facelifts, and platform updates on aggressive timelines, creating consumer expectations that mirror the smartphone upgrade cycle.

The original Li Auto L9 launched in mid-2022 as the company's flagship full-size SUV, quickly becoming one of the best-selling large SUVs in China. When Li Auto unveiled the next-generation L9 Livis in 2026 — roughly 4 years later — some industry observers and consumers questioned whether the gap was too long in a market where competitors refresh lineups every 18 to 24 months.

This criticism reflects a uniquely Chinese market dynamic. With over 100 EV brands competing for attention, product freshness has become a key differentiator. Consumers in China's premium EV segment often compare vehicle update cycles to those of flagship smartphones from Apple, Samsung, and Xiaomi, which typically follow 12-month refresh cadences.

Li Xiang's Core Argument: Safety Cannot Be Simulated Away

In his public response, Li Xiang laid out a detailed defense centered on the irreducible complexity of automotive safety validation. His core points include several critical distinctions between automotive and consumer electronics development:

• Crash safety testing requires building and destroying dozens of physical prototypes across multiple configurations
• Durability validation demands hundreds of thousands of miles of real-world driving across varied climates, altitudes, and road conditions
• Powertrain and battery testing involves extreme-temperature cycling, abuse testing, and long-term degradation analysis that cannot be meaningfully accelerated
• Regulatory compliance across multiple markets requires sequential certification processes with fixed timelines
• Supply chain validation for new components — especially safety-critical parts — adds months of lead time

Li specifically addressed the role of artificial intelligence in the development process, acknowledging that AI has transformed certain aspects of automotive design — such as aerodynamic simulation, generative design, and software testing — but insisting that it remains fundamentally limited in physical validation.

'AI can help us simulate thousands of crash scenarios digitally,' Li's argument essentially states, 'but regulations and physics still demand that we crush real cars, drive real miles, and validate real hardware.'

The AI Simulation Gap in Automotive Development

Li Xiang's comments touch on a nuanced reality that the broader tech industry sometimes overlooks. While digital twin technology, finite element analysis (FEA), and AI-driven simulation have dramatically improved the efficiency of early-stage automotive design, they have not eliminated the need for physical validation.

Companies like Tesla, BMW, and Toyota all still maintain extensive physical testing programs despite being at the forefront of simulation technology. Tesla, for example, uses sophisticated neural network-based simulation for its Full Self-Driving (FSD) software but still conducts millions of miles of real-world testing.

The gap is particularly pronounced in several areas:

• Material behavior under real-world stress — simulations approximate but cannot perfectly predict how new materials fatigue over years of use
• NVH (Noise, Vibration, Harshness) — subjective ride quality assessments still require human evaluators in physical vehicles
• Battery thermal management — real-world thermal runaway scenarios involve complex chemical reactions that simulations can model but not fully replicate
• Sensor calibration for ADAS — advanced driver assistance systems require extensive real-road validation across edge cases that simulations may not capture

This reality check is important context for an industry increasingly enamored with the idea that AI can compress every timeline. In automotive development, the physics of safety validation creates a hard floor below which cycle times cannot safely drop.

Industry Context: The Software-Defined Vehicle Paradox

Li Auto's defense arrives at an interesting moment in the global automotive industry. The concept of the software-defined vehicle (SDV) has dominated industry discourse for the past 3 years, with companies from Volkswagen to General Motors investing billions to transform cars into continuously updatable platforms.

The promise of SDVs is that major vehicle improvements can be delivered via over-the-air (OTA) software updates, theoretically reducing the importance of hardware refresh cycles. Tesla has pioneered this approach, with vehicles receiving significant capability upgrades years after purchase.

However, Li Xiang's argument exposes the paradox at the heart of the SDV vision: while software can be updated continuously, the underlying hardware platform — the body structure, suspension geometry, battery architecture, and thermal systems — still requires traditional development timelines.

This creates a two-speed development reality. Software features like infotainment updates, autonomous driving improvements, and energy management optimizations can iterate on weekly or monthly cycles. But the physical vehicle platform that houses all that software still demands 3 to 5 years of development — a timeline that has remained remarkably consistent across the industry for decades.

What This Means for the Broader EV Market

Li Xiang's public stance carries implications beyond Li Auto's product roadmap. It signals a potential maturation of the Chinese EV market, where the initial land-grab phase of rapid model proliferation may be giving way to a more quality-focused approach.

For Western automakers watching China's EV market with a mix of admiration and anxiety, the message is somewhat reassuring. The laws of physics and safety engineering apply equally in Changzhou and Detroit. Companies that have felt pressure to match the apparent speed of Chinese EV development can take some comfort in the fact that even China's most successful EV executives acknowledge that certain timelines cannot be compressed.

For consumers, the takeaway is more nuanced. A 4-year development cycle for a ground-up new vehicle platform is actually quite aggressive by traditional automotive standards — legacy automakers like Toyota and Mercedes-Benz have historically operated on 6 to 8-year platform cycles. Li Auto's 4-year cadence already represents a significant acceleration compared to industry norms.

For AI and tech companies working on automotive simulation tools, Li's comments represent both a challenge and an opportunity. The acknowledgment that AI has limited impact on physical validation timelines identifies a clear gap in current technology — one that companies specializing in digital twins, physics-based simulation, and synthetic data generation may seek to close over the coming decade.

Looking Ahead: Can AI Eventually Close the Gap?

The long-term question Li Xiang's comments raise is whether advances in AI and simulation technology will eventually compress automotive development cycles more dramatically. Several trends suggest partial convergence is possible.

Generative AI is already accelerating the design phase, enabling engineers to explore thousands of structural configurations in hours rather than weeks. Digital twin platforms from companies like Siemens and NVIDIA Omniverse are creating increasingly accurate virtual replicas of entire vehicles. And synthetic data is enabling autonomous driving systems to train on billions of simulated miles.

But Li's fundamental point is likely to remain valid for the foreseeable future: when lives are at stake, physical validation is not optional. Regulators worldwide — from NHTSA in the United States to Euro NCAP in Europe — show no signs of accepting purely simulation-based safety certification.

The most likely outcome is a gradual compression of timelines from 4 years to perhaps 3, enabled by better simulation tools that reduce — but do not eliminate — the number of physical prototypes required. Full replacement of real-world testing remains a distant prospect, measured in decades rather than years.

For now, Li Xiang's message is clear: building a safe car still takes time, and no amount of industry hype about AI acceleration should change that calculus.