Feb 26, 2019, 2:44 PM ETInterview, Fireside Chat

a16z Podcast | Capitalizing on an Autonomous Vehicle Future

Current State Analogy: The autonomous vehicle (AV) industry in 2019 is compared to the mobile phone industry circa 2005, characterized by high valuation of talent and early-stage technology rather than mainstream consumer readiness.
Talent Dynamics: Top-tier engineers, specifically roboticists and AV developers, are currently as "highly coveted" in Silicon Valley as mobile engineers were in 2010–2011, with major firms like Uber, Waymo, May Mobility, and Voyage aggressively recruiting from specific university labs (e.g., CMU, UC Berkeley, University of Michigan).
Hype Cycles: Public and media perception oscillates rapidly between "early days," "trough of despair," and "imminent arrival," mirroring the volatility of the mobile hype cycle.
The "iPhone Moment": Defined not as a single product launch (e.g., the 2007 announcement), but as a 12–24 month period where multiple transformative applications (e.g., Uber, Snapchat, Instagram) mature simultaneously; this moment is anticipated to occur when AVs move beyond early adopters to general public use in constrained environments like campuses and airports.
Market Entry Strategy: The speakers argue that 2019 represents the optimal time to launch an AV company, as the path to mainstream adoption will be incremental rather than a discrete event, similar to how the iPhone era arrived through a accumulation of innovations rather than a single "Eureka" moment.

Level 0: Standard production vehicles today (e.g., anti-lock brakes, traction control) with no world-sensing automation.
Level 1: Basic automation such as adaptive cruise control where a single system (e.g., radar) manages acceleration or braking.
Level 2: Combined lane-keeping and adaptive cruise control (e.g., Tesla Autopilot), requiring constant human supervision and attention.
Level 2+: Extended Level 2 features allowing for automated freeway interchange navigation and merging while the human remains in the loop; a key focus for current OEMs like Mobileye.
Level 3: A technically "dubious" classification where the vehicle manages driving but requires the human to take over immediately upon alert; often indistinguishable in practice from Level 2 due to safety risks of delayed takeover.
Level 4: Fully autonomous operation within a specific Operational Design Domain (ODD), such as Waymo's geofenced zones in Arizona; includes campus shuttles and delivery robots where no human driver is present in the vehicle.
Level 5: Theoretical full autonomy in all conditions and weather; described as an idea rather than a near-term reality, with no industry player currently close to this goal.

Commoditization of Hardware: The smartphone wars created a "peace dividend" where expensive military-grade sensors (GPS, IMUs) became cheap, commoditized parts available for recombination in the automotive sector.
Vertical vs. Horizontal: The trend favors a horizontal ecosystem where companies buy off-the-shelf commoditized components (sensors, mapping) and focus differentiation efforts solely on core algorithms, contrasting with the early automotive industry's vertical integration (e.g., Ford owning rubber plantations).
Cloud Analogy: The AV industry is experiencing a "AWS moment," where infrastructure companies provide the "rails" (simulation, mapping, data) allowing new entrants to bypass building proprietary stacks from scratch.
Co-evolution: The industry involves a non-linear feedback loop where real-world vehicle data continuously improves simulation tools, which in turn improve the next iteration of vehicle software.

National Sentiment: Unlike the software industry, AV adoption is deeply tied to national interest; governments in Germany (Daimler/BMW), Japan (Toyota/Honda), and China are unlikely to allow foreign competitors to displace domestic manufacturers.
Regulatory Arbitrage: U.S. states act as "laboratories of innovation" (e.g., Arizona, Indiana), with varying regulatory strictness creating a patchwork environment that balances risk against the cost of stifling innovation.
Safety Trade-offs: States accepting AV testing (like Arizona) face the risk of being the first to experience fatalities, creating a political trade-off where citizens must elect representatives who accept these risks.
Legal Consequences: Unlike pure software, AV hardware failures carry severe legal repercussions, including potential prison time for engineers, as seen in the Volkswagen diesel scandal.

Shift in Simulation Use: Simulation has evolved from a tool to finalize hardware specifications to a continuous infrastructure for developing software products that evolve indefinitely in the wild.
"Good Enough" Threshold: Simulations do not need to be perfect replicas of reality but must reach a threshold where they provide high confidence that simulated behaviors translate to real-world scenarios.
Reality-in-the-Loop: A shift from linear development cycles to a circular process where real-world data informs the simulator, which then feeds improved data back into the vehicle system.
Data as Differentiator: While tools are becoming commoditized, the value lies in the "know-how" of algorithm development and the unique datasets required to handle edge cases and specific ODDs.

Inevitability: The transition to autonomy is driven by the convergence of three factors: cost reduction, increased convenience, and enhanced safety, making it an inevitable market shift regardless of specific timelines.
Second-Order Effects: The industry will face significant disruption in related sectors, including real estate, insurance, and infrastructure design, similar to the unexpected social impacts of mobile phones (e.g., selfies, social media).
Democratization: In 10–15 years, the ecosystem may reach a point where building autonomous vehicles is as trivial as creating a mobile app today, leading to a proliferation of niche, specialized AV applications.
Silicon Valley and Detroit Merger: Success requires the integration of Silicon Valley's software capabilities with Detroit's manufacturing, supply chain, and brand distribution channels; the future is an "and," not an "or."