The Bumpy Road to Self-Driving Cars: Who's Winning?

TL;DR

The self-driving car industry, born from military challenges, is a high-stakes race dominated by Waymo's massive head start and deep resources, while others struggle to keep pace. #VentureStep #AutonomousVehicles #Tech

INTRODUCTION

The promise of fully autonomous vehicles has felt just around the corner for years, capturing the imagination of tech enthusiasts and the public alike. Yet, for all the billions invested, the industry remains one of the few mature sectors still largely built on speculative value rather than widespread, revenue-generating services. It’s a landscape defined by incredible technological leaps, high-profile failures, and a relentless race to solve one of the most complex engineering problems of our time. ¹

In this episode of Venture Step, host Dalton Anderson takes us on a ride through the chaotic and fascinating story of self-driving cars. We start at the very beginning, in the sands of the Mojave Desert, where a government challenge laid the groundwork for the entire industry. From these ambitious, and initially unsuccessful, experiments, a new generation of technology was born, sparking a commercial gold rush.

Dalton breaks down the current state of play, exploring the companies that have risen to the top, like Google's Waymo, and those that have stumbled, like the Ford and Volkswagen-backed Argo AI. We'll dive into the strategies of key players, the critical difference between marketing hype and real-world capability, and what the future might hold for a technology that could fundamentally change how we live, work, and travel.

KEY TAKEAWAYS

• The modern autonomous vehicle industry has its roots in the DARPA Grand Challenges, military-sponsored competitions that pushed early innovators to solve fundamental problems in perception and navigation. ²²²²
• Waymo (Google) holds a significant lead in the space due to its early start, massive data and machine learning resources, and strategic partnerships with ride-sharing giants Uber and Lyft. ^3333333
• There's often a gap between marketing and reality, as seen with Tesla's "Full Self-Driving" feature, which is currently classified as a Level 2 driver-assistance system, not full autonomy. ⁴⁴
• The industry is highly volatile, with major automakers like Ford and GM pausing or shutting down ambitious projects (Argo AI, Cruise) after facing challenges in achieving profitability and scalability. ^555555555
• While a few companies pursue full autonomy, a profitable sub-sector has emerged for those "selling the shovels"—companies like Mobileye that provide the essential sensors and hardware, generating billions in revenue. ^666666666

FULL CONVERSATION

Dalton: Welcome to VentureStep podcast where we discuss entrepreneurship, industry trends, and the occasional book review. ⁷ Today we're going to be taking some time to take a ride through the fascinating and sometimes chaotic story of autonomous vehicles. So no pun intended, let's buckle up. ⁸ I'm your host, Dalton Anderson. Today, we're discussing the sometimes bumpy but fascinating journey of autonomous vehicles. ⁹⁹⁹⁹

An Industry Built on Speculation

Dalton: Autonomous vehicles feel like one of the few mature industries that is still built on speculative value. ¹⁰I don't think there are that many companies providing a service to the public, either B2B or B2C, where it's generating legitimate revenue to support all the resources going into the projects. ¹¹ I know that there are people selling shovels. We know that story. It's not the people digging for gold, it's the people selling the shovels that make the money. ¹²You could see those things with AI, with Nvidia selling the GPUs to these AI companies. ¹³

Autonomous vehicles, feel like is one of the few mature industries that is still built on speculative value. ¹⁴

Dalton: It's the same thing for companies that provide hardware. Tesla used Nvidia for the longest until COVID happened. 15151515They couldn't produce some of their Teslas because they didn't have enough chips. 16So Tesla was like, alright, well, I guess we'll just make our own chips, which is crazy because it's not that easy, but it worked out. 17 A similar thing could be said for Amazon during those peak shipping periods. Amazon was heavily reliant on UPS, and then UPS was missing the shipping times for the orders. 18So Amazon's like, fine, I'll just make my own logistics company. 19

if you're providing hardware but can't sustain it, some of these larger tech companies will just build it themselves. ²⁰

Dalton: My point was if you're providing hardware but can't sustain it, some of these larger tech companies will just build it themselves. ²¹ Today's agenda is we're going to talk about the early days and that deals with the DARPA challenges. Then we'll talk about the rise of self-driving cars, and where we're at today. ^22222222

The DARPA Challenge: Failure in the Desert

Dalton: DARPA is the Defense Advanced Research Projects Agency. ²³DARPA had an initial challenge in 2004 where they requested teams to compete in this autonomous driving vehicle challenge in the Mojave Desert for 150 miles. ²⁴No one finished in 2004. There were a couple of reasons why. ²⁵

Dalton: One, there were sensor limitations; they weren't as advanced as they are today. ²⁶ They had algorithm shortcomings. Their algorithms for processing data were not sophisticated enough to maintain control in complex environments like the desert. ²⁷ There were also mechanical and hardware issues. The heat caused things to overheat. ²⁸ Think about a car—the computers it takes to run a car through the desert. It's hot and it's already doing a lot, so it's overheating and not working. ²⁹They also had communication challenges, where instructions from a control center were only making it so far before the vehicle would disconnect. ³⁰

Raising the Stakes: Success in 2005

Dalton: So one year later, after everyone failed, the government said we need to figure this out. ³¹They did the same exact challenge, but made it a significantly more difficult race, and if you finished, you got a million dollars. ^32323232 That year they had sensor improvements. They started using LIDAR, which stands for light detection and ranging. ^33333333LIDAR can be used in low light situations, so you can see at night. ^34343434

Dalton: The teams started incorporating machine learning into their algorithms. That improved simulations and things on the fly because the models they were using could interpret real data that it was receiving. 35 They also did a whole bunch of collaboration. These other teams all banded together, sharing their knowledge after the 2004 event where no one finished. 36

I think they were just like, don't care who finishes. We just need someone to finish. Please. We can't be embarrassed like this again. We're supposed to be smart. ³⁷

Dalton: They applied all their lessons learned from the 2004 failures, addressed their weaknesses, and it led to five teams finishing, which is big. ³⁸

The Urban Challenge: From Desert to City Streets

Dalton: So then the government moved on from the desert to the cities. ³⁹ The city is a different challenge because you've got to obey traffic laws. This pushed the teams to apply these sophisticated algorithms and sensors to the real world. ⁴⁰ It moves it from this controlled environment, like the desert, to a complex urban setting. You have to abide by rules for pedestrians, intersections, parking, and other real-world scenarios. ⁴¹

Dalton: In 2007, the challenge was held at the George Air Force Base, so it was a more controlled environment. ^42424242 You would need advanced perception and decision-making to navigate these urban environments. You need object detection and tracking to know what objects are. What does a human look like? What does a stop sign look like? If a light is red, what does that mean? There are a lot of what-ifs going on here. ^43434343

Dalton: This also sparked public awareness. People are less comfortable with this autonomous vehicle thing. They've seen all these sci-fi movies where the autonomous vehicles take over. 44 People don't like change. They're not necessarily going to embrace it. It's sketchy. I feel like it's way less sketchy than driving itself though. People can doze off, get frustrated, distracted, and cause mistakes. 45This race inspired commercial applications and the development of the self-driving car industry bloomed. 46

The Rise of the Commercial Players: Waymo's Dominance

Dalton: The one player that has been constant in this regard is Google's Waymo. Waymo started in the mid-2000s and I think Waymo has the strongest position right now. ⁴⁷As of June 2024, Waymo is offering autonomous ride services, Waymo One, in Los Angeles, Phoenix, San Francisco, and is testing in Austin. ⁴⁸Uber is partnering with, guess who, Waymo. ⁴⁹Lyft is also partnered with Waymo to provide self-driving vehicles. ⁵⁰ So that really solved it. Waymo is ahead. ⁵¹

Dalton: Waymo should be in front, to be fair. Waymo started pretty much first, has the backing of Google, and they had a headstart. 52Google has massive resources in machine learning, deep learning, and the infrastructure to support these things. 53They've built up servers and TPUs to allow them to train data and run simulations. 54Google already has it all set up for them, and that gives them an advantage. 55

Where Does Tesla Fit In?

Dalton: Then there's Tesla, which I love, but I'm just really not sure about their full self-driving capabilities. It's been going on for years that they're talking about full self-driving. But is it full self-driving? I don't know. ⁵⁶Full self-driving is rated Level 2. The driver needs to stay ready and be attentive at all times. ⁵⁷

I'm not really sure how it's called full self-driving if it's not full self-driving though. ⁵⁸

Dalton: I'm kind of confused about why you would call it that. Just call it self-driving. People that don't know any of these things related to autonomous vehicles are going to be so confused. 59 Full self-driving was expensive. On Saturday, the software's price dropped to $8,000 from the previous price of $12,000. 60 It's also available as a $99 a month subscription. That license is tied to the owner, not the car. So if the owner buys a new Tesla, you don't have to buy another one; you can just transfer it to your new vehicle. 616161616161616161

The Human Cost and Future Promise

Dalton: There have been a couple of accidents. There have been people falling asleep at the wheel, and a couple of people have died. ⁶² People are concerned, and rightfully so. My personal opinion is people die a lot from car crashes. About 43,000 people died per year as of the last statistic in 2024. ^63636363

I think that the progression of this technology if done correctly could save a lot more lives than just the couple that have lost during the process. ⁶⁴

Dalton: I'm not advocating for anyone to die, but accidents happen. I think with the technology improving at the rate it is, if they do it correctly and in a safe manner, years from now we could have this technology to allow elderly people who can't see very well to get out of the house and experience life. 65You could have more time with families when you're traveling instead of having someone drive. 66

The Automakers Who Tried and Stumbled

Dalton: There were other traditional automakers that had their own go at it. GM had its Cruise division. ^67676767In 2023, they suspended operations for Cruise, and then in May 2024, they put the cars back on the road to begin testing. ⁶⁸BlueCruise is for Ford and Lincoln, but they closed out their venture with Argo AI. ⁶⁹ Argo AI was a startup that was also part-owned by Volkswagen. Both companies decided to shut down Argo in 2022, citing challenges to achieve autonomous vehicles at scale. ^{707070707070707070}

when you're really pushing the boundaries of what can be to reality...the side effects of these ideas...come into the public...and greatly affect their lives in a positive manner. ⁷¹

Dalton: After that, Ford was focusing more on near-term assistant driving features. ⁷²Volkswagen is still pursuing self-driving and they're partnered with a company called Mobileye. ⁷³ Then there is Motional, which is Hyundai's joint venture. They're in development for Level 4 and also have partnerships with Lyft and VIA. ⁷⁴

The Future of Autonomous Entertainment

Dalton: There was this other company that was like a Formula One for autonomous vehicles, and it was called RoboRace. It was the world's first autonomous vehicle racing series where teams compete with self-driving cars. ⁷⁵One of the cool features was you could race the car yourself. ⁷⁶On the track, you could be racing with the other AIs, basically playing a video game while the event is going on, racing against the AI in real life in real time. ^77777777It sounded really cool, but it folded during COVID. ⁷⁸

Maybe autonomous vehicles will take away some things, but a lot of things that will get back like the ability to connect and communicate with your loved ones and your partner and your friends while a car is driving you, which sounds really cool. ⁷⁹

Dalton: I hope that you enjoyed this rambling episode of me talking about the history of autonomous vehicles. It's a wild ride. You never really know what's going to happen. Companies pull out, come back, fade away, or emerge like a phoenix. 80I think there are exciting developments in the years to come. 81

RESOURCES MENTIONED

• DARPA (Defense Advanced Research Projects Agency)
• Nvidia
• Amazon
• UPS
• Tesla
• Waymo (Google)
• Uber
• Lyft
• George Air Force Base
• GM (General Motors)
• Cruise
• Ford
• BlueCruise
• Volkswagen
• Argo AI
• Mobileye
• Motional (Hyundai)
• VIA
• Pony.ai
• WeRide
• Luminar Technologies
• RoboRace

INDEX OF CONCEPTS

Dalton Anderson, VentureStep, Autonomous Vehicles, Self-Driving Cars, DARPA, DARPA Grand Challenge, DARPA Urban Challenge, Mojave Desert, Speculative Value, B2B, B2C, Nvidia, GPUs, Tesla, Amazon, UPS, LIDAR, Machine Learning, Deep Learning, George Air Force Base, Object Detection, Waymo, Google, Waymo One, Los Angeles, Phoenix, San Francisco, Austin, Uber, Lyft, Level 2 Self-Driving, Full Self-Driving, GM, General Motors, Cruise, Ford, BlueCruise, Lincoln, Volkswagen, Argo AI, Mobileye, Motional, Hyundai, VIA, Pony.ai, WeRide, Luminar Technologies, RoboRace