Venture Bytes #109: Robotaxis Moving Closer To Reality

Robotaxis Moving Closer To Reality

‍

Self-driving cars have been anticipated for years, but their moment might finally have arrived. The recent approval of Waymo's autonomous taxis by the California Public Utilities Commission and strong customer response – exemplified by 300,000 sign-ups – demonstrates a tangible example of significant changes in urban mobility. It is also an indication of Waymo progressing towards a scalable self-driving solution.

Self-driving technology’s path to commercialization has taken substantially longer than expected due to the slow pace of technological evolution. Developing autonomous capabilities is inherently complex, with the biggest challenge being the vastness of the operational design domain (ODD). Autonomous vehicles must navigate a near-infinite number of scenarios, far beyond what physical testing sites can replicate. Consequently, relying solely on physical tests limits the ability to train vehicles to respond with human-like adaptability.

However, recent progress in AI, machine learning, and sensor technology has accelerated development. Improvements in light detection and ranging (LiDAR) and RADAR systems, high-resolution cameras, and powerful onboard computing platforms are bringing us closer to realizing the potential of self-driving cars.

At CES 2024, Advanced Micro Devices unveiled its Versal AI Edge "XA" SoCs, delivering up to 171 TOPS of AI performance and high-bandwidth programmable logic for vehicle signal processing. These chips manage sensor fusion for LiDAR, radar, and cameras, handling complex computations to map surroundings and detect obstacles. Nvidia’s Thor, expected in 2025, is expected to have 2,000 TOPS of processing power.

Additionally, research on AI algorithms for autonomous vehicles has soared from 302 studies in 2014 to 2,718 in 2023. Advances in deep learning have dramatically improved perception, prediction, and planning tasks, with object detection accuracy increasing from 70% to 95%, semantic segmentation from 50% to 90%, and trajectory prediction from 60% to 90%. Additionally, semiconductor advancements and the demands of the autonomous driving industry have driven radar technology to evolve, with transmitter and receiver channels expanding from 1×2 to 12×16 by top Tier 1 companies and 48×48 arrays by leading startups. Given the rapid pace of technology adoption, the probability of fully autonomous Level 5 vehicles operating "anywhere, anytime" in the next five years is very high, in our opinion.

A recent study highlighted the potential of fusion algorithms in overcoming sensor limitations and enhancing the overall safety and efficiency of autonomous driving systems. The study showed that integrating LiDAR with camera sensors to improve object detection in autonomous vehicles leads to improved detection performance, especially in challenging scenarios. This is particularly important to attain level 5 autonomy, which means that vehicles can drive anywhere in road traffic and under all conditions without human beings.

Waymo’s approval is a bright spot in the autonomous vehicle sector, signaling a potential shift from trials to actual commercial deployment. With its broader rollout, Waymo can demonstrate the viability of driverless taxis as a sustainable venture. The company, with its fleet of 700 driverless cars with over 3.8 million miles driven in San Francisco, is also addressing longstanding safety concerns.

By all metrics, the safety of AVs is improving, and it is already safer than human-driven vehicles. According to IDTechEx data, human drivers in the US are involved in collisions about once every 200,000 miles, and this rate nearly doubles in cities like San Francisco. In contrast, driverless Waymo cars have a collision rate of once every 52,000 miles, and Cruise cars once every 63,000 miles. However, only six out of 33 collisions involving Cruise vehicles were due to the autonomous system, translating to one fault per 344,000 miles. Similarly, Waymo's system was at fault in five collisions or one per 238,000 miles. A study conducted by Waymo in collaboration with Swiss Re, based on over 3.8 million miles of fully autonomous driving in California and Arizona, also revealed a 100% reduction in bodily injury claims, compared to Swiss Re’s human baseline of 1.11 claims per million miles. With increasing safety, consumer acceptance is also growing with a recent S&P Global Mobility consumer survey demonstrating that 65% of buyers want Level 2+ hands-off automated highway driving.

The regulation of robotaxis is becoming increasingly crucial as companies like Zoox, Waymo, and Cruise advance their autonomous vehicle programs. Zoox is testing in Austin, Miami, Las Vegas, and Seattle, operating robotaxis without steering wheels and pedals, and recently began carrying passengers in Foster City, CA. They have a new fleet of retrofitted Toyota Highlanders in Austin and Miami, though not yet open to the public. Waymo is operating in Phoenix, Los Angeles, and San Francisco, with plans to expand to Austin in 2024. Cruise, after pausing operations due to a pedestrian accident in San Francisco, plans to restart testing in Dallas with a driver and has recently cleared a key hurdle to get robotaxis back on California roads.

Investment Opportunities in Enabling Technologies

A McKinsey study estimates that autonomous driving could create $300-400 billion in revenue by 2035. Also, UBS predicts that 80% of people will use robotaxis, and car ownership in urban areas will fall by 70% by the same period. In addition to industry leaders like Waymo, Cruise, and Zoox, companies in the broader ecosystem supporting robotaxis will also be beneficiaries of the mass adoption of robotaxis. In addition to companies like Nvidia, Mobileye, and AMD, these include various private start-ups developing critical hardware and enabling technologies.

California-based Pony.ai develops and deploys autonomous driving technology for robotaxis and commercial logistics. The company, valued at $8.6 billion in its Series D round in October 2023, offers full-stack solutions including localization, mapping, perception, prediction, planning, and control systems for Level 4 autonomous vehicles. Founded in 2019, Autobrains offers self-learning technology that mimics human driving perception. With $120 million in funding from BMW i Ventures and Temasek, its adaptive technology enhances autonomous driving efficiency and reliability. Autobrains secured its first commercial order for Liquid AI software from a Chinese EV manufacturer, with production set to begin in 4Q24, signaling strong market validation.

California-based Applied Intuition develops simulation software and tools for autonomous vehicle development and testing. With $590 million in funding from Elad Gil and Lux Capital, Applied Intuition provides simulation and validation tools that accelerate the development of safe autonomous vehicles. The company has partnered with 18 of the top 20 automotive OEMs and with its clientele including major automotive companies like Toyota, Daimler, and General Motors.

Israel-based Nexar develops AI-powered dashboard cameras for real-time road monitoring and incident detection, enhancing driving safety. The dashboard camera market is expected to grow from $4.54 billion in 2024 to $8.32 billion by 2029 at a 12.90% CAGR, per Mordor Intelligence. With $149 million in funding from GE Ventures and Aleph, Nexar's technology is vital for collecting valuable driving data and improving road safety.

Not Your Father’s Wars Anymore; Doing More With Less

‍

The world faces heightened geopolitical tensions. Conflicts in Ukraine and Gaza, escalating issues in the Red Sea, and concerns surrounding Taiwan underscore the urgent need for the US to strengthen its defense capabilities. Given this backdrop, the call to action would be to ramp up investments in defense tech and beef up the size of the armed forces. But that is not what we see. Venture capital investment in defense tech went down from $35.8 billion across 800 deals in 2022 to $34.9 billion across 627 deals in 2023, per Pitchbook.  Furthermore, the US Army is downsizing its personnel, cutting 10,000 positions in engineering and counterinsurgency roles and reducing personnel in non-deployable units and training posts by 9,200.

This restructuring reflects a strategic pivot towards advanced technology and preparing for future large-scale combat scenarios, that are expected to minimize direct combat and maximize long-range strike capabilities. The changes aim to align resources with modernization efforts and enhance readiness against sophisticated adversaries, amidst ongoing recruitment challenges across the military services. Investments in sophisticated drones, unmanned aerial vehicles, and eVTOLs, among others, are now part of the arsenal.

Israel employed drones for long-range strikes against Hamas, minimizing direct engagement in close-quarters urban combat. Skydio, a California-based startup, offers AI-powered autonomous drones for public safety, fire and rescue, enterprise inspection, defense, and more. In October 2023, BETA Technologies, a Vermont-based electric aircraft startup, delivered its ALIA aircraft to Eglin Air Force Base in Florida. This marked the first deployment of an eVTOL to the Air Force. Such startups are poised to become highly valuable companies in the coming years.

‍

“The successful integration of small, autonomous drones will provide our unit-level leaders with enhanced situational awareness, arming our Marines with a unique combat advantage.” -      Lt. Col. Frank P. Mease, CO 1/1 First Battalion, First Marine Division

Additionally, the backend technologies such as data labeling and cybersecurity that bolster national security are also getting an investment boost to counter the surge in geopolitical tensions. Incidents of cyber sabotage and cyber espionage have driven an increase in cyber militarization. Several nations now recognize 'cyber' as the fifth domain of warfare alongside land, sea, air, and space. Many countries have allocated substantial budgets to enhance their military cyber capabilities, comprising both offensive and defensive strategies. Nearly 50 states possess or are developing offensive cyber capabilities.

Scale AI, for instance, secured a $1 billion round led by Accel, valuing the data labeling and evaluation startup at $13.8 billion. While not a conventional defense tech firm, Scale AI's technology supports defense applications. Cybersecurity startups raised $2.7 billion in 1Q24, marking a 69% surge from the previous quarter, per CrunchBase. In 2023, the US Cybersecurity and Infrastructure Security Agency partnered with SentinelOne, a California-based cybersecurity company, for the Joint Cyber Defense Collaborative (JCDC). This collaboration aims to enhance the US government's cybersecurity strategy for safeguarding the cyber ecosystem and critical infrastructure. AI-powered cybersecurity startups are well-positioned to help tackle cyber threats on national security. Cyera, a New York-based AI-powered Data Security Posture Management (DSPM) startup, delivers a cloud-native, AI-driven, agentless platform for securing data across diverse environments. Cyera stands out by offering comprehensive oversight of data throughout its lifecycle, from creation to storage and use. A significant 180% valuation surge in its latest funding round reflects strong investor belief in Cyera's future growth potential.