Argo.ai Is a Canary in the Coal Mine for “Robotaxi or Bust”

The market for Autonomous Vehicle (AV) platforms is as competitive as the underlying technology is complex, with dozens of companies, ranging from startups to established Original Equipment Manufacturers (OEMs) vying to develop vehicles that can safely assume driving tasks on our behalf, to the benefit of our safety and our convenience. While technology choices over sensor stacks, Artificial Intelligence (AI) frameworks, and System-on-Chip (SoC) compute configurations vary from one supplier to the next, in general terms, every developer of AV platforms falls into one of two camps—“scalable autonomy” or “robotaxi or bust.”

Scalable Autonomy: Developers of “scalable autonomy” have placed flexibility at the heart of their platforms, ensuring that their core hardware and software modules can be configured to support the full spectrum of autonomous applications, from active safety to semi-autonomous driving, to fully driverless robotaxis. These developers have been driven by a strategy to monetize AV technologies in the real world as it exists today, not waiting for leaps forward in legislation or consumer enthusiasm to begin introducing their core technologies to the market. “Scalable autonomy” players have been key in formulating the Level 2+ (L2+) strategy, which combines driver control and responsibility with high-end autonomous features—a formula that has been welcomed enthusiastically by many passenger vehicle OEMs.

In addition to the short-term revenue opportunities of semi-autonomous driving in the passenger vehicle market, these vendors keep one eye on the long term, fully driverless goal. By leveraging the scale of the millions of passenger vehicles shipping each year, “scalable autonomy” developers can bring down hardware costs, refine software and algorithms, and crowdsource vital maps to place themselves in the optimum position to deploy fully driverless technology once the regulatory and technical framework is in place. Players in this camp include Mobileye, NVIDIA, and Qualcomm.

Robotaxi or Bust: Most AV startups fall into the “robotaxi or bust” camp, focusing their resources on the Holy Grail of autonomous driving—fully driverless robotaxis. “Robotaxi or bust” developers often (and correctly) point out that the real value of autonomous driving is to be found when drivers can be done away with altogether, unlocking not only safety benefits, but entire business models and value propositions founded on low-cost, high-utilization driverless transport. Furthermore, these vendors often shied away from addressing the very real Human-Machine Interface (HMI) challenges that arise when trying to balance an autonomous driving system with a human driver—often concluding that this was a problem hardly worth solving if a jump from Level 2 (L2) to Level 4 (L4) could be achieved within the next few years.

In some cases, but not all, developers of “robotaxi or bust” platforms determined that a fully driverless platform could be their ticket to revolutionizing personal transit without the need to partner with or supply an established automotive OEM. Indeed, if ridehailing platforms like Uber could rapidly rise to become staples of our everyday personal mobility, then imagine what an Uber without the driver could achieve. Suppliers of “robotaxi or bust” platforms include some of the biggest names in the AV industry, including Waymo, Cruise, Zoox, Aurora … and Argo.ai.

Argo.ai is not the first AV startup to disappear, and it will certainly not be the last. Already, vendors such as Starksy Robotics and Drive.ai have folded, while ridehailing giants such as Uber and Lyft sold off their own in-house robotaxi efforts. If a badly run or poorly funded AV startup shuts down, it may grab headlines, but it shouldn’t cause too much lost sleep in the industry. What will concern many in the industry, however, is how things at Argo.ai could have gone wrong so quickly when, by some measures, things seemed to be going so right.

Argo.ai was among a tiny handful of players offering bona fide driverless rides to regular passengers, and not just a preselected “panel” of first adopters/testers. It had expanded into Miami and Austin, and enjoyed good working relationships with local legislators. Through partnerships with established brands Lyft and Walmart, Argo.ai had a strong and proven path to market for its AV platform in both personal transit and goods delivery, respectively. Of course, this falls far short of where many expected the “robotaxi or bust” market to be by 2022, but compared to many of its contemporaries, Argo.ai seemed to be in a strong position.

Since early 2019, ABI Research has emphasized the value of the “scalable autonomy” strategy, particularly as it relates to the low hanging fruit, L2+ opportunities. In its October 2022 financial release, Ford’s Chief Executive Officer (CEO) Jim Farley confirmed the importance of L2+ to its AV strategy, and how this ultimately led to the decision to wind down Argo.ai: “ … there’s a huge opportunity right now for Ford to give time – the most valuable commodity in modern life – back to millions of customers while they’re in their vehicles. It’s mission critical for Ford to develop great and differentiated L2+ and L3 applications that at the same time make transportation even safer. We’re optimistic about a future for L4 ADAS, but profitable, fully autonomous vehicles at scale are a long way off and we won’t necessarily have to create that technology ourselves.” (ABI Research emphasis was added.)

While Argo.ai’s commercially-offered rides and path to market partners might have set it apart from other “robotaxi or bust” suppliers, that pales in comparison to the 83,000 customers for, and 21 million hands-free miles delivered by Ford’s L2+ BlueCruise product, just 1 year after launch.

Every autonomous platform provider must now have a solid L2+ strategy. ABI Research’s Smart Mobility and Automotive Research Service provides guidance on how to develop SoCs, sensor configurations, and mapping platforms capable of addressing a key L2+ opportunity.

Ultimately, Argo.ai’s fate can be explained by a fatal mismatch between its AV philosophy and the needs of its primary investors. As a product of the peak of the AV hype of the 2010s, Argo.ai favored a “robotaxi or bust” development approach, while its primary investors, Ford and Volkswagen, needed a partner that could also supply their L2+ and Level 3 (L3) needs. The only viable path to market for AV technology over the visible horizon is the current batch of established automotive OEMs. Therefore, it is essential that providers of AV technologies pivot to prioritize scalability and flexibility in each of their hardware and software components.

• Sensor Fusion: AV platform suppliers must be able to extract maximum insight from visions and radar sensors, which are low cost and already widely adopted in active safety Advanced Driver-Assistance Systems (ADAS). An early sensor fusion approach that leans heavily on exotic sensors, such as Light Detection and Ranging (LiDAR) will not prove scalable into L2+.
• Driver Monitoring: In L2+ deployments, it is essential that the core role of the driver as supervisor be guaranteed through a robust driver monitoring system. L2+ cannot become untested or unverified L4 by the back door—to attempt this would be to provoke an enormous backlash by consumers and legislators.
• Common SoC Architecture: A common and scalable architecture that maximizes hardware and software component re-use between L2+ and L4/Level 5 (L5) will be key to translating short-term L2+ revenue into a competitive advantage when rolling out future L4/L5 vehicles.