Skip to content
Home » Tesla Q3 2024 Earnings Call Reveals FSD Updates, MPI Improvements, and Expansion Plans

Tesla Q3 2024 Earnings Call Reveals FSD Updates, MPI Improvements, and Expansion Plans

Tesla

Tesla’s ambitious plans for Full Self-Driving (FSD) technology are taking shape with unprecedented momentum. Tesla’s has unveiled a roadmap that promises to transform autonomous driving from a sophisticated experiment into a mainstream reality.

In their Q3 2024 earnings call, Tesla outlined their most aggressive FSD reliability targets yet. The company’s metric of choice, Mean Miles Between Interventions (MPI), is set for a dramatic improvement – targeting a thousandfold increase in reliability throughout 2024.

The path to enhanced reliability starts with FSD V12.5, which aims to improve MPI by two orders of magnitude. V13 is expected to push these boundaries even further, with projected performance 5-6 times better than its predecessor, (Tesla AI and Autopilot: Close to Production AI Software that Will Ship Soon in FSD v13). By 2025, Tesla expects their systems to outperform human drivers in terms of safety metrics.

 training ahead of schedule on a 29,000 H100 cluster at Gigafactory Texas

Tesla isn’t just making promises – they’re already testing these capabilities in the real world. A new employee-exclusive ride-hailing service has launched in the Bay Area, marking the company’s first steps toward a broader Robotaxi network.

While safety drivers currently remain behind the wheel due to regulatory requirements and current MPI levels, Tesla’s expansion plans are already in motion. Tesla has identified California and Texas as initial markets for public launch, with Texas potentially leading the way due to its streamlined regulatory environment.

Tesla expects to roll out a ride-hailing app to the general public in Texas and California in 2025

Tesla’s AI infrastructure has seen a significant upgrade with the transition to AI4 (HW4) This platform supports models with five times more parameters than its predecessor, driving substantial improvements in decision-making capabilities.

For vehicles equipped with HW 3.0, Tesla maintains its commitment to backward compatibility. Through sophisticated model adaptation and increased training resources, the company aims to maximize performance on existing hardware.

Tesla’s computing infrastructure is expanding at a breathtaking pace. Their current 67.5 EFLOPS capacity is set to reach 90 EFLOPS by year-end, supported by a massive training cluster at Giga Texas.

The Giga Texas facility houses 29k H100 chips, with plans to scale to 50k shortly. The ultimate goal? A hybrid cluster of 100k H100 and H200 chips, creating one of the world’s most powerful AI training environments.

Tesla’s FSD ambitions extend beyond passenger vehicles. The company is developing specialized models for their Semi trucks, leveraging data from hundreds of vehicles already on the road. Each Semi comes equipped with FSD hardware, ready to deploy autonomous capabilities once the software matures.

Tesla full self-driving future isn’t just approaching – it’s accelerating.

Related Post

Tesla FSD v12.5.6.1: Completely End-to-End AI Technology Running on Both City Streets and Highway

Tesla Cybercab Testing Program in Palo Alto: A Look at the Autonomous Future

Tesla Robotaxi: How Removing Steering Wheels Could Drive Autonomous Acceptance

Will Grok-1.5V Become the Secret Sauce for Tesla FSD v13?