Robotaxi Rollout | The Brainstorm EP 107

Key Concepts

Full Self-Driving (FSD): Tesla's advanced driver-assistance system aiming for autonomous driving.
Robo Taxi: Autonomous vehicles operating as a taxi service, a key future product for Tesla.
Humanoid Robots (Optimus): Tesla's development of general-purpose humanoid robots.
Open-Loop vs. Closed-Loop Simulation: Different methods for testing AI systems in simulated environments.
AI Compute: The processing power required for artificial intelligence tasks, a significant cost factor.
Parameters: In machine learning, these are the variables that a model learns from data.
Hardware 3 vs. Hardware 4: Tesla's generations of onboard computer hardware for FSD.
Unit Economics: The cost and revenue associated with producing and operating a single unit (e.g., a robo taxi).
Commercialization Threshold: The point at which a technology or product becomes viable for widespread market adoption.
Prediction Markets: Platforms where users can bet on the outcome of future events.

Tesla Earnings Call and Robo Taxi Ambitions

Tesla's recent earnings call revealed an increased confidence in their robo taxi service. Elon Musk stated that improvements in Full Self-Driving (FSD) and the shift towards unsupervised FSD have led them to believe they can scale vehicle production capacity as fast as possible. This confidence is linked to a projection of reaching an annualized production rate of 3 million vehicles within the next 24 months.

A significant development mentioned is the addition of "reasoning" to the car's AI. This concept, explained by Ashoke, Head of AI at Tesla, differentiates between open-loop and closed-loop simulations. In closed-loop simulation, the actions of the simulated car can influence other elements within the simulation, allowing for a more dynamic and realistic testing environment. Adding reasoning to the car means it can better predict how its actions will affect the surrounding traffic and interpolate these outcomes.

Tesla has committed to removing safety drivers in parts of Austin by the end of the year and launching robo taxi services in 8 to 10 metro areas by the same deadline.

Key Points on Robo Taxi:

Increased Confidence: Tesla is more confident than ever in its robo taxi service.
Production Ramp: Confidence in robo taxi enables scaling production to 3 million units annually within 24 months.
Reasoning Capability: The addition of "reasoning" to the car's AI is a step towards unsupervised FSD.
Simulation Advancement: The distinction between open-loop and closed-loop simulation highlights improved testing methodologies.
Launch Targets: Removal of safety drivers in Austin by year-end and launch in 8-10 metro areas by year-end.

FSD and Hardware Generations

The discussion touched upon the current state of FSD and its relation to hardware generations. While Tesla is pushing forward with advanced AI models, the question of support for older hardware, specifically Hardware 3, was raised.

Tesla's CFO indicated that Hardware 3 is not being abandoned. The perspective is that Hardware 3 vehicles will eventually achieve the same capability as Hardware 4, albeit with a perpetual one-year delay, assuming Tesla continues to support these older systems. This presents a value judgment for owners: continue with Hardware 3 or upgrade to Hardware 4 for earlier access to FSD advancements. Tesla is also incentivizing upgrades by allowing FSD license transfers to new vehicles during specific promotional windows.

Nick shared his experience with FSD version 14, noting subtle but significant improvements over version 13. He described V14 as feeling more "intentional" in its actions, leading to greater user confidence, even if some edge cases still require refinement. The overall sentiment is that FSD is approaching a point where it can function as a reliable chauffeur, a significant human uplift.

Key Points on FSD and Hardware:

Hardware 3 Support: Tesla is not abandoning Hardware 3, but it will lag behind Hardware 4 in FSD capabilities.
Perpetual Delay: Hardware 3 vehicles are expected to be perpetually one year behind Hardware 4 in FSD performance.
FSD License Transfer: Tesla incentivizes upgrades by allowing FSD license transfers to new vehicles.
V14 Improvements: FSD version 14 offers qualitative improvements in intentionality and user confidence.
Chauffeur Capability: FSD is nearing a point where it can reliably act as a chauffeur.

The Connection Between Robo Taxi and Humanoid Robots (Optimus)

A significant portion of the discussion focused on the complexity of developing humanoid robots and how the lessons learned from robo taxi development are relevant. Brett argued that robo taxi is an "incredibly difficult, challenging problem," citing the long development cycles and the shift to full AI stacks required.

The complexity of the humanoid robot problem is estimated to be 200,000 times harder than robo taxi. This vast difference stems from several factors:

Actuated Outputs: Humanoid robots have a significantly higher number of degrees of freedom (20-40) compared to a robo taxi's limited outputs (steering, acceleration/braking).
Environment Complexity: Roads are structured and designed for ease of navigation, with a limited diversity of objects (cars, pedestrians). A home environment, conversely, contains hundreds or thousands of distinct objects requiring different interactions.
AI Compute Requirements: Solving the humanoid robot problem is projected to require an estimated $100 billion to $200 billion in AI compute spend. This is a substantial investment, and Tesla is exploring efficient training methods, including utilizing distributed chips in their vehicles.

The timeline for commercializing Optimus is estimated to be around 2029-2030, contingent on this massive AI compute investment. The success of robo taxi commercialization is seen as a potential gating factor for Optimus development, as robo taxi cash flow could fund the necessary AI compute. If robo taxi scaling is delayed, Optimus commercialization could be pushed back by approximately four to five years.

Key Points on Humanoid Robots:

Robo Taxi as a Precursor: The challenges and solutions in robo taxi development inform humanoid robot development.
Exponential Complexity: Humanoid robots are estimated to be 200,000 times more complex than robo taxis.
Factors of Complexity: Higher degrees of freedom, diverse object interaction, and unstructured environments contribute to this complexity.
Massive AI Compute Investment: An estimated $100-$200 billion in AI compute is needed for humanoid robot software training.
Commercialization Timeline: Optimus is projected for commercialization around 2029-2030.
Capital Constraint: Robo taxi cash flow is crucial for funding Optimus development; delays in robo taxi could impact Optimus timelines.

Regulatory Hurdles and Market Predictions

The conversation also touched upon regulatory challenges for robo taxi deployment and predictions for future milestones.

Robo Taxi Regulation:

State-by-State Approval: Robo taxi rollouts require approval from regulators, with varying ease across states.
California's Complexity: California is noted as having particularly complex regulatory hurdles for robo taxis.
Political Influence: The widespread adoption and perceived benefits of robo taxis (e.g., parents using them for children's transport) are expected to create political pressure on regulators to approve their expansion.

Prediction Market Insights:

Optimus Release: A prediction market suggests a low probability (around 14%) of Tesla releasing Optimus for public purchase by June 30, 2026. The consensus is that manufacturing and internal use will likely precede public sales.
California Robo Taxi Supervision: A prediction market on whether Tesla's robo taxi will operate without supervision in California this year (by the end of 2024) had a 10% chance of "yes." While capability might exist, regulatory approval is seen as the primary bottleneck. Tasha predicted "no" due to regulatory timelines, while Nick was more optimistic about the possibility.

Conclusion and Future Outlook

The discussion highlights a period of significant technological advancement and ambitious goals for Tesla. The company's increasing confidence in its robo taxi service, driven by FSD improvements and the addition of reasoning capabilities, is poised to accelerate vehicle production. Simultaneously, the development of Optimus, while facing immense complexity and requiring substantial investment, is seen as a long-term endeavor that will benefit from the foundational work in autonomous driving.

The interplay between technological development, capital allocation, and regulatory approval will be critical in shaping the timelines for both robo taxi and humanoid robot commercialization. The overall sentiment is one of excitement for the future, with Tesla facing "good problems to have" in terms of deploying capital into these transformative technologies. The discussion also underscored the importance of AI compute and the potential for rapid progress once certain technological thresholds are crossed, drawing parallels to advancements in language models.