Waymo Takes Action Ahead of Tesla’s Cybercab Rollout

Last week, Waymo asserted that achieving fully autonomous vehicles requires a combination of sensors, emphasizing that “pure end-to-end” AI systems fall short in safety — implicitly targeting Tesla without directly naming the company.

These comments were made in a blog post and during an interview with Axios, just a week before Tesla’s anticipated September 3 event to officially launch its Cybercab, a two-seater vehicle, into its growing robotaxi fleet. Additionally, Waymo announced the introduction of three new markets on Tuesday morning, further broadening its robotaxi services, which currently operate in over a dozen U.S. cities.

Although the discussion was quite technical, Waymo’s statements sparked a lively social media debate that spanned the entire weekend.

Pierre Ferragu, an analyst and managing partner at New Street Research covering Tesla, criticized Waymo’s arguments on X, stating, “Their arguments are weak. My interpretation: Waymo has built a driving gas plant that large-scale AI renders obsolete, and now resorts to the rhetoric of incumbents. Classic innovator’s dilemma.”

Ethan Teicher, a spokesperson for Waymo, shared an image of a John Wick promotional poster featuring Keanu Reeves’ character with multiple guns aimed at him, commenting, “It shows how convinced they are that their data and success record indicate that driverless mileage, AI interpretability, and a mix of sensor types are essential for safe, truly autonomous driving at scale.”

Tensions are high because if Tesla successfully demonstrates that the Cybercab can operate effectively at scale, it would intensify the competition between the two companies and their differing strategies for developing autonomous driving technology. The potential market at stake could be worth hundreds of billions.

In the past, this rivalry was more theoretical than practical. However, Waymo recently showcased its extensive real-world experience.

“Cameras are remarkable, but they aren’t sufficient,” noted Srikanth Thirumalai, VP overseeing Waymo’s driving software, in the company’s blog post. “For years, the debate over the capability of cameras alone for achieving full autonomy has persisted. Now, with over 200 million real-world miles under our belt, the evidence is clear: safe, fully autonomous operations at scale require more. By leveraging inputs from cameras, lidar, and radar, the Waymo Driver constructs a comprehensive, redundant worldview that no single sensor can replicate.”

Thirumalai added that a solely end-to-end neural architecture that processes raw pixels and produces steering commands — which reflects Tesla’s approach — risks encountering “black box failures.”

“Even the most advanced AI models with trillions of parameters can still misinterpret,” he explained to Axios. “In physical AI, there’s no option to click reboot or refresh. You must face the consequences.”

Waymo has consistently employed a combination of cameras, radar, and lidar sensors in vehicles manufactured by third parties. While this conservative approach has allowed the company to expand to a fleet of approximately 4,000 robotaxis across 14 U.S. cities, providing 500,000 paid rides weekly.

In contrast, Tesla’s Elon Musk has long criticized lidar as a “crutch.” The company focuses on achieving full autonomy using only cameras and AI. Tesla has spent recent years developing the Cybercab, a specially designed two-seater sedan intended to be autonomous from the outset.

The Cybercab lacks a steering wheel and pedals and has a comparatively small battery, with plans to produce thousands — a recent filing indicates the company aims to manufacture over 125,000 annually.

However, Tesla must still demonstrate that its self-driving software can achieve full autonomy. The company has faced delays; Musk previously promised 1 million robotaxis would be operational by 2020 — though Tesla has spent the past year testing its own Robotaxi network in select Texas and Florida cities with modified Model Y SUVs.

These trials remain limited in scope, as Tesla emphasizes safety over rapid scaling. Just in the past few weeks, Tesla has started removing safety monitors from most of those automobiles.

Nevertheless, that situation may soon shift. Tesla has begun registering Cybercabs with the Texas DMV in anticipation of the September 3 event. It remains uncertain how swiftly the company intends to scale the fleet following the launch of the new two-seater, although social media users have reported seeing numerous units parked in lots throughout the country.

Should Tesla validate that its AI-driven method for self-driving can function on a large scale, it would represent a significant achievement for the company and a success for its engineers.

Nevertheless, Tesla will have to tackle various challenges associated with operating a robotaxi service — challenges that Waymo continues to navigate with its fleet of thousands of self-driving vehicles. These challenges include managing adverse weather, responding to emergency situations, and ensuring safe movement in school zones.

Beyond the technical aspect, a financial competition is also emerging: cost. Waymo’s approach is inherently more expensive, as it employs additional sensors and integrates them into vehicles from various manufacturers. This means Waymo incurs costs for purchasing these vehicles and, in some cases, pays import taxes before implementing self-driving technology.

Tesla, on the other hand, manufactures its own vehicles and is betting that its AI will be proficient enough to navigate the environment using just a few cameras. It’s a bold wager — one that Waymo is skeptical will succeed. However, if successful, it could enable Tesla to outcompete Waymo or even Uber on pricing.