Musk: FSD V14.3 is a Key Piece but Won't Immediately Scale Up Robotaxi
Elon Musk discussed the latest Full Self-Driving version during the earnings call, stating that FSD V14.3 is the "final piece of the puzzle for achieving unsupervised FSD." However, he emphasized that the team has a series of "significant architectural improvements" on the way, and that it is not reasonable to deploy Robotaxi on a large scale before these improvements are implemented. According to previous disclosures, V14.3 has rewritten the AI compilation and runtime based on MLIR on the vehicle side, improving overall response time by about 20%, and upgrading the visual encoder and reinforcement learning training link, which is seen as a system-level reconstruction rather than a regular iteration.
Source: Public Information
ABAB AI Insight
Musk refers to V14.3 as the "final piece of the puzzle," but he also actively puts the brakes on the timeline for Robotaxi, reflecting a combination of technological, regulatory, and commercial realities. Technologically, V14.3 addresses model inference efficiency and the consistency of the perception-decision link, allowing the system to "see clearly, react quickly, and behave more coherently" in more scenarios. However, transitioning from "high-level driving assistance" to an "unsupervised operational network" requires meeting engineering-grade standards in reliability, redundancy design, remote operation, and anomaly handling, not just single-vehicle software performance.
From a regulatory and liability perspective, once large-scale Robotaxi commercialization begins, accident liability will no longer be defined as "driver + assistance system," but will be redefined as "operator + autonomous driving system." This requires Tesla to not only prove that the system is statistically safer but also to provide auditable logs, remote control mechanisms, and upgrade processes so that regulatory and judicial systems can clearly reconstruct the decision-making chain after an accident occurs. Musk's decision to delay the "immediate rollout of Robotaxi" essentially acknowledges that a single software leap is insufficient to cross this institutional threshold; further architectural changes are needed to upgrade FSD from a "consumer-level feature" to an "infrastructure-level system."
On the commercial side, FSD subscriptions have already begun to contribute stable software revenue for Tesla, while the Robotaxi network signifies a shift from "selling cars + subscriptions" to "operating a mobility network," which requires significant upfront capital, fleet management, and city-level negotiations. If a nationwide Robotaxi is hastily launched while the architecture is still undergoing significant adjustments, any systemic flaws could have amplified effects on the brand and regulatory relationships. Musk's choice to define V14.3 as a "critical capability point" while postponing "scaled Robotaxi" until after the next round of architectural upgrades attempts to balance the delicate tension between "telling a sufficiently aggressive AI story" and "avoiding over-promising again."