One view of the AV timeline, links

View of the AV timeline from a former hedge fund manager, following technology – Michael Dempsey’s view of the AV timeline

What’s the timeline?

Context: When do we have fully autonomous vehicles on the road?


A) Many say 2020–2025 because every OEM, tech journalist, and thought leader has said that time. In practice, if you look at adoption of new technologies over time within auto, as well as how regulations are moving, this “a few years out” timeline makes sense, but a lot is based on self-imposed deadlines from major automakers.

B) Much longer than we think. Until we can get autonomy magnitudes more perfect than human drivers are, it won’t be mainstream, because the public won’t be able to accept computers crashing and killing drivers at nearly the same rate that humans do.

Further Reading: Forecasts, Autonomous vehicle implementation predictions, Why driverless cars will screech to a halt, A self-driving vehicle in every driveway but when?

2. How Does This Change Car Ownership?

Context: If cars can drive themselves all around, how does ownership change?


A) UberPool model. 96%+ of the time vehicles sit unused today. Vehicles will change from owned assets to leased assets. It becomes a race to $/mile charged and the majority of people will just utilize UberPool type rideshare programs where they call a vehicle that is on a continuous trip. Whether or not the ride is actually shared with others is inconsequential. Some of the wealthier people will own their own, single-family AV.

B) Car ownership won’t fundamentally go away because people like owning their own cars (status symbol?) and things like suburban infrastructure + commute times will make it so that people will want their own cars to be on their perfect schedule and to service their family. Also, people love driving.

C) All of these ownership model changes also could mean big problems for OEMs. Light vehicle sales could fall 40% while newer players like Tesla, Faraday Future, Zoox, and more could emerge and massively disrupt the incumbents. In addition, OEMs might not be able to move from being hardware companies to a combination of hardware/software companies. Massive consolidation will occur with many OEMs dying a slow death.

Further Reading: Barclay’s Report, Columbia Sustainable Mobility, The coming nightmare for the car industry

3. How Does This Change Cities?

Context: Autonomous vehicles are going to have different infrastructure requirements than those with drivers in major metropolitan areas.


A) Infrastructure will be great! Parking lots could move to outskirts of a city or even not need to exist as people’s vehicles will “work for them” a la Tesla Master Plan Part Deux. There’s a massive problem with this as it doesn’t fix for cleanliness and on-demand readiness (if I want my Tesla in the spur of the moment but it is driving someone across town, I’m going to be upset).

B) Traffic could be hell. With the ownership model potentially shifting to shared vehicles, cars will most likely only need infrastructure as long as they need to charge and instead of finding parking, could just continue to circle/drive around cities empty when not in-use. Robin Chase puts it best in the Backchannel article:

“Right now, our “congested” roads and cities are mostly filled by individuals driving alone in their cars (75 percent of all trips). Just imagine our streets and your frustration when 50 percent of the cars have no people in them at all.”

Further Reading: How shared self-driving cars could change city traffic, Self-Driving cars will improve our cities, if they don’t ruin them , Urban Planning and Autonomous Vehicles

4. What about impossible choices? The Trolley Problem

Trolley Problem

Context: Who is the vehicle loyal to, society or the “owner”? The trolley problem depicts this as a situation in which a vehicle must choose between two certain deaths. The car is skidding towards a group of children and has to make the decision to either kill the children, or course correct and kill the driver.


A) Car should continue on path and let nature take its course.

B) Car should optimize for minimal societal damage. Goodbye driver.

C) Punt on the question entirely. Will vehicles even operate at this level of understanding in the near-term?

Further Reading: Self-driving cars and the trolley problem, Why self driving cars must be programmed to kill, Let the policy makers handle the trolley problem

5. Who is responsible and what do we do about insurance?

Context: With no drivers, somebody will have to be responsible for the actions of cars.


A) The OEMs/service providers — Volvo and others have come out and saidthey will take full responsibility of their vehicle. Other OEMs should too. Maybe they even take over insurance pending regulations.

B) The Consumers at a micro level— Insurance could be on a per trip to the user or a per vehicle to the service provider.

C) New insurance providers arise with very specific products for AVs.

Further Reading: KPMG on auto insurance for AVs

6. Will it be illegal to drive in the future?

Context: Humans are simply not as good drivers as computers. Would it benefit society as a whole if it were illegal for us to drive?


A) It should be based on the facts that human drivers are significantly more dangerous. How does our perfect world of autonomous vehicles work if there are random data points of bad human drivers. Almost an “only as strong as your weakest link” situation.

B) Humans view driving as a right. In no near future will it be illegal to drive. The lifecycle of cars can last years, and it’s not fair or practical to assume that all old vehicles won’t be able to be on the road. With that said, modifications could be required a la seatbelts being required in all cars today.

Further Reading: Would we ever ban human driving?, Driving a car will be illegal by 2030, Elon Musk on driving being illegal,

7. What about all the jobs?

Context: With all of this automation, will the millions employed as professional drivers become unemployed? What happens to the jobs?


A) Many jobs will be repurposed. Instead of drivers, some will be fleet operators/managers. But many jobs will be lost. This is a problem that we’ll have to solve over time, however with each new technological innovation that has allowed humans to be more efficient, they have been able to come up with new problems to solve that only humans can work on. Driving is a highly robotic task, “when humans are replaced by machines, it frees humans up to do more human things.”

B) Basic income. Similar to point A, but free up people to work on different problems.

Further Reading: What are the impacts on the economy with AVs?,Autonomous vehicles will replace taxi drivers. But that’s just the beginning,Self-driving trucks and basic income,

8. Let’s Talk About Tesla

Context: Tesla was the first player to have some sort of autonomy in the market, has recently released its Master Plan Part Deux, and is the leading new-school OEM disrupting incumbents.


A) Elon wins BIG. Tesla has a massive advantage over any other player due to deep technological expertise on both hardware and software stack,Gigafactory, and ability to capture more data around driver behavior than any other company in the world. The company caters to 100% of the future ownership model market but selling its vehicles to service providers, selling direct to consumers who still want to own their own vehicle, and also to the in-between class that wants to own but make $ by turning their under-utilized asset into a ridesharing vehicle. In addition, the move into public transit vehicles + utility vehicles allows them to penetrate markets that may take longer (or shorter) to reach mass adoption of AVs.

B) Let’s take another look at that master plan. Tesla’s deep technological expertise has been partially based on a third-party provider in Mobileye (which they just severed ties with) and has been very stubborn around putting expensive LiDARs in their cars, which many believe are necessary for full autonomy in the near-future. If low-cost LiDARS like Quanergy, Innoviz,TriLumina, or others don’t make it to market as promised, and Tesla isn’t able to develop their own, they could be in a precarious situation in getting to market first. Especially if their cars keep crashing.

But they have all of that data? It turns out that the majority of the data that Tesla gathers across those millions of miles isn’t actually as valuable as many believe. They log images every few frames and save video data/multiple frames only during accidents/failures. So maybe that data moat isn’t as big of a head start as many think.

And loaning out your $30K+ asset to strangers? Sure vehicles are under-utilized but you pay for the convenience of being able to use your car instantly, which could be a problem if it’s across town. Also, who is going to pay to have your car cleaned? Will the extra money made by the average person drastically outweigh the depreciation and maintenance of your asset?

Further Reading: Has Tesla Lapped Google in Self-Driving Car Race?,Understanding the huge gulf between the Tesla Autopilot and a real robocar,Tesla’s Master plan — some expected, some strange

9. Now, Let’s Talk About Uber (and tangentially Lyft)

Context: Uber is one of the most polarizing private companies in the world with a known desire to move to an autonomous vehicle-centric model. There are a lot of questions surrounding what happens to them when driver supply is no longer a constraint.


A) Uber wins. Uber has built the pre-eminent ridesharing brand and will be able to roll out an autonomous fleet for monetization faster than most. They are already mapping, have bought vehicles, and have some of the brightest minds in AVs working on this. They can turn on/off autonomy for data gathering purposes based on road conditions (perhaps only during the day when its somewhat cloudy in suburbs to start) and they will still be able to monetize this autonomy. Uber understands how cities ebb and flow better than any other company in the world and that data will be invaluable for fleet optimization as they potentially move to an asset-heavy model.

B) Supply-chain dynamics kill Uber. Brand loyalty doesn’t exist in the sharing economy and OEMs (like GM/Lyft) can optimize vehicles and vehicle prices for the new ownership models. An OEM can go from selling a car at an LTV of $50,000 to leasing it/utilizing it for an LTV of $100,000+. The software approach to auto will proliferate from Tesla to all existing OEMs (though consolidation will happen) and will make OEMs more bulletproof to disruption. If the ultimate metric is $/mile, owning the entire vehicle stack wins, and Uber does not own the vehicle manufacturing.

Further Reading: Uber’s Achilles Heel, @ajt, NPR on Lyft/GM, Autonomous Cars Break Uber

Other Questions:

Is the car the next screen?

It started with TVs, expanded to computers, then to phones, and the next screen will be the car.

Everyone continues to talk about how we are going to be freed up with an extra 40 minutes per day during our commute to consume content, view ads, and do anything else our heart desires in our autonomous vehicles.

This could impact everyone from Netflix to Google (see more below) and even to companies like McDonalds which won’t be able to entice a robot to pull off the highway on road trips to go eat at their restaurants.

What is Google doing?

Google has had large talent attrition recently and has made strong statements surrounding the difficulty of manufacturing its own vehicles. It seems increasingly likely that they could partner with an OEM (potentially acquire one?) and control the software stack in order to create an Uber/Lyft competitor.

Another alternative is that they build out the vehicle OS, take advantage of the “cars are the next screen” narrative, and monetize with ads. Startups likePolysync are positioned to fight this battle.

What is Apple Doing?

Apple has lost major people within their stealthy car team. They most recently were rumored to be moving away from creating their own car and instead are focusing on the software side. Unclear what that means yet, however it is a strange move for the most powerful hardware company in the world.

I had a conversation with a friend who is a founder of an autonomous vehicle software company about where the future of startups/emerging companies are within his space and we started to really get somewhere when he asked me “well, what kind of future do you believe in?”

It’s an obvious framework in theory, but that single question encompasses so much that it helps tease out a lot.

For autonomous vehicles I start to think about the ownership model being completely upended which in turn trickles up (?) to the manufacturers due to the needs and desires of shared car service providers. This includes things like updatable/modular vehicles that last longer, require very little maintenance, and OEMs that are able to iterate on their product (in this case, the vehicle) quickly without the same infrastructure costs. This is just the beginning of the massive list of opportunities and frankly, my friend helped spur many of those initial thoughts.

In this conversation as we talked and I subsequently learned which future that I believe in related to autonomous vehicles, I realized that this is the question I should directly ask about every emerging industry that I look at within Frontier Tech. It helps build more than just a thesis around a particular segment of a market. It forces you to think through which ancillary parts are affected, where opportunities may lie, and what parts you still haven’t fully thought through.

Decision Tree 1: Will Autonomy Commoditize?

As I continue to build out various theses and thoughts around different areas of tech, I’ve started to build out a framework (with the help of smart friends + colleagues) of looking at key decision trees within a given area.

The idea is that if you can identify the pivotal decisions or moments within a given area, you don’t have to be able to predict the future fully, but utilizing a portfolio approach, as an investor you can make bets on both sides and be comfortable with the risk that a few investments will not work out.

One that I’ve spent a fair amount of time on recently is the question that my super smart friend Kevin and I talked about for a few hours the other week:

Will autonomy within driverless cars become commoditized in the near future?

There are tons of things that can spawn from this but let’s simplify it a bit:

If the answer to the question is yes, then companies that are working on the “autonomy software” layer, will see consolidation or early M&A (a bet I believe many VCs are making). This assumes that these autonomy software/algorithm companies don’t become huge layers that sell into multiple OEMs, which doesn’t seem likely.

It also means that nobody will buy a Mercedes vs. a Toyota because one can drive itself better. This inherently makes sense from a regulatory standpoint, despite the fact that some cars today are indeed safer than others to drive.

Thus, if we believe that autonomy software becomes commoditized, other areas such as routing software/algos (which Uber probably owns the best of) could prove to be a key competitive advantage, along with proprietary maps, which a few OEMs (Daimler, VW, BMW) have already collaborated on in the form of the$3B HERE acquisition. The secondary and tertiary effects go on and on, but this is where I’ll leave this “yes” path.

Where things get more terrifying (for some) is if the answer to the question is no and autonomy doesn’t commoditize in the next 10-20 years.

We’re already seeing just about every major OEM tout how they are going to put driverless cars on the road in some form in the next 5, 10, or 20 years. How realistic that is, is debatable. But if autonomy doesn’t commoditize and only the most innovative and adaptive OEMs are able to figure out full level 4 autonomy at the production level, then we will see massive consolidation across multiple billion-dollar enterprises (some of the top public auto & truck manufacturers currently hold a cumulative $87.8B market cap).

Car sales are already expected to fall due to autonomy, and with a smaller overall pie available, a lack of full autonomy would be the death spell for many OEMs. We would see some M&A for manufacturing/supply chain (though even *that* could change as ownership models change) as well as certain IP around the hardware, but bargaining power of a fledgling OEM that couldn’t make the transition from a hardware business to a combination of hardware/software business could be low.

From there the next-generation of auto OEMs will want to control other aspects, such as the map layer or better decision making or more advanced driving behavior. This in turn could lead to an increase in speed limits for different types of vehicles and down the rabbit hole we go.

Honestly, all of this is probably a bit overzealous in this case as I’ve been told by some very smart people that these battles could play out in the 20-30 year timeframe, and not so much the 7-15 year future. However, I believe the decision trees we build (and dynamically adjust) can prove to be an optimal framework for investing in emerging technologies which possess massively disruptive opportunities with uncertain paths of how we get there.

Portable Power + Infrastructure

Power does not need infrastructure..maybe

One of the main issues I wrote about for online grocery delivery moving forward has been infrastructure within most of India. While around 80% of the middle class that will be benefiting from traditional eCommerce services in India will reside in major cities, how can India (and other, less developed nations) develop and get online?

There are many factors in this, one of which is a stable power grid.

Up until now we’ve always thought of power as a static infrastructure – long power lines, large power plants nearby, and lots of permanent investment to connect cities/towns/houses/etc. But what if we thought of power as a mobile component? Using fuel cell technology or renewable energy sources, we could setup power grids in tier 2 and 3 cities that are more stable and can exist independently of one another. This will start a cycle that goes initial power -> increased demand and building on infrastructure -> more scalable/stable power.

This may not have worked in India for things like airports, which the government built many of in hope’s of spurring travel + increasing economic growth of further out cities, it could help level up tier 2 cities in a certain way. While India’s road and air infrastructure may not get outsiders to move, a stable or upgraded power grid can attract key business players to stay there permanently, and thus an improved ecosystem of sorts is born.

And a lot on drones development/transport…

Voice Controlled Drones

Majority of commercial use cases within the drone space require some form of collaboration with the drone, and not just pure set and forget. If we want to lower the barrier of entry for drone flight (effectively allowing users to buy drones and fly them off the shelf vs. hiring operators) natural language interaction is what could eventually drive forward the “drone assistant”.

To be fair, touch-screen enabled drone flight is great for point A to point B, but for precision commands we will first see developer/command line level control (i.e. “camera movement: North 25 mm”, pivot camera 25mm up) which companies like dronee are working on, and then in the future as CSAIL is betting on, pure natural language commands (”drone fly to the top of the closest structure and persistently fly 360 degrees around the object.

This will be enabled by good object identification + avoidance, and then NLI.

Building Smart Drone Infrastructure

TL;DR – Drones are getting smarter, and this “smart” drone tech should be the basis for any future drone infrastructure development.

Drones are a key component to my Data 2.0 thesis for their ability to do things that previously weren’t possible at a commercially viable price point, as well as their value-prop that allows them to replace other areas of the imagery and geospatial analysis stack.

Drone funding has exploded in the past year, and we are probably only a year or two away from another Unicorn in the space (most likely Airware) with the potential for many more in the future. But with these fundings have come a bevy of “me too” drone entities targeting the same enterprise use-cases in the same ways (often basic capturing of aerial imagery with a technology-enabled, but somewhat manual analysis product).

The opportunity still exists for these V1 (or “dumb”) drone companies, as no players have won anything close to a majority market share across the applicable verticals, many which represent large markets.

But this perceived opportunity in the new tech seen today opens the door for an overeager market to overfund, and overestimate the “dumb” drone industry in order to fully develop the market leaders in the first waves of the commercial drone movement.

As a counterpoint, one could argue that by establishing early winners, it opens the door for further innovation and consolidation. However, I believe the inevitable existence of “smart” drone tech companies, both on the manufacturer and service side, could upend these incumbents and cause a wasteland of overfunded or once-promising “frontier tech” companies founded between 2012-2017.

It is for these reasons that as we look towards the future, public and private entities must not build infrastructure that is made for the drone companies of today, at the cost of stunting innovation in the “smart” drone technology of 7-10 years from now.

For the first time in our history, humans are beginning to accept that machines can better handle dynamic, complex tasks – leading to a willingness to put our lives in the hands of machines like self-driving cars.

As drone usage becomes more widespread and advanced for this newly formed tier of UAVs, the conversations around how countries will utilize and regulate drones at scale are only going to increase. Today, drones are primarily manually flown, or pre-programmed to fly GPS coordinates in open space. But this industry must not follow the path that the automotive industry did. Innovation is happening at a much faster rate than ever before possible with cars, and thus, we must think about sunk costs when it comes to building the infrastructure for drones.

With companies like Amazon already working on delivery, we should capitalize on this future adoption (whether it happens in the US or elsewhere) and build out a completely new form of infrastructure with an ideal system in mind: One that supports autonomous drones and the advantages that future technology will afford us including precision flying, machine vision, real-time video analysis, and a connected platform that will enable drones to be smarter and more aware than ever before. Amazon shouldn’t ever need to pilot their drones, and drones shouldn’t have to be afforded a 1-2 meter margin of error.

A well thought-out infrastructure could even forward the Drones as a Service model – evoking a vision very similar to electric cars moving autonomously throughout cities waiting to be called. On-demand drone utilization should become as close to a commodity product as Uber, Lyft, or car-hailing has become in just 6 short years.

As stated above, this movement starts with machine vision and awareness, and the first-movers may already be here.

Skydio was founded on the basis of developing the “visual cortex” of drones, allowing drones equipped with their technology to “navigate the world intelligently”. PreNav, uses computer vision (via LiDAR) deep neural networks and 3D path planning software to allow a similar maneuverability, initially targeting high-res industrial monitoring, but with use-cases that can expand to multiple industries. Airware could power the enterprise drone fleets of the future and SkySpecs’ technology platform could enable all drones to coexist on a broader platform without accidents.

Smarter drone technology is progressing and has surpassed the MVP stage, so why push off to tomorrow what we can accomplish in the next 5-10 years?

When thinking about a future city with a drone infrastructure built in the skies, I always come back to the transportation systems of Minority Report. In that fictional city, at some point the entire infrastructure had to be re-done for the self-driving vehicles. The costs must have been massive, but the benefits had to of made sense. We don’t need to withstand these massive “change” costs. Let’s build the companies and the infrastructure together the right way the first time.

Building Out a Thesis on Data 2.0

I firmly believe that we are on the brink of reaching escape velocity in the way we gather, categorize, and analyze both proprietary and existing data sets.

Over the course of my professional life, I’ve developed a few key theses that have driven both my career and personal interests strongly.

One of my first theses is one I’ve called the Niche Consumer. You can read more about it here, but the gist of it is that consumers are being driven to verticalized and specialized experiences. I believe this will reverberate throughout the United States across multiple socio-economic classes and industries ranging from food to travel to material good commerce to even real estate. Other theses I’ve written less about revolve around the staying power of suburban America, as well as personal wealth management.

My latest thesis is centered on what some are calling Data 2.0. I firmly believe that we are on the brink of reaching escape velocity in the way we gather, categorize, and analyze both proprietary and existing data sets that were previously un-attainable or largely too expensive for widespread, private sector use.

Innovation around things like satellite technology, drones, mobile phones, connected sensors, geospatial analysis and more broadly big data analytics has been built by or enabled a new crop of companies founded in the past 5-7 years to capitalize on this. And while their use-cases are extremely wide-ranging, they often serve deep-pocketed customers in both the private and public sectors, and many can enable change on a worldwide scale for good (examples here andhere). In short, their potential customer base is large.

In addition to the primary use-cases, there are entire economies that will rise with these sectors.

For satellites, there are innovators like Accion Systems, a company focusing on electronic propulsion systems specifically for CubeSats, or Spaceflight Industries, which as part of its business offers “rideshare” services to aggregate demand and bring down the most costly part of commercial nano-satellite process (the launch) or space de-commissioning (”cleaners”) companies likeAstroscale that see the inevitable rise of space debris and are proactively looking to service it.  Even Image analytics companies like Orbital Insight andWindward are able to serve as customers for the satellite operators by purchasing the imagery, while also providing a new, targeted layer of analytics to potential buyers across multiple industries. There even is an independentSPACE DETECTIVE AGENCY. And that’s just a few of the many companies working with satellites!

When I look at commercially-viable drones, I start to see primary use-cases around imaging (whether it be in agriculture, real estate, maritime or other areas) and delivery (an upcoming burden for governments worldwide) but also anti-drone detection companies like DeDrone, or drone-as-a-service companies like Dronebase. As the hardware continues to improve, I envision that drones will reach a point where they can deliver a layer of granularity that is a far-greater cost-proposition for specific use-cases versus low-orbit CubeSats, because as Yael Maguire says, “launching a drone is always going to be cheaper than putting a satellite into orbit.

Moving to mobile, rising economies are less clear, however companies likePlacemeter, which uses open cameras feeds and mobile phones to capture data about cities and the assets moving within them, has huge implications on both private-sector traffic modeling, as well as public-sector city planning.

Others, like Premise Data utilize the rising worldwide mobile phone adoption to gain a previously unparalleled view into consumer goods pricing and derive detailed economic data in emerging markets that allows a better economic understanding of our world than ever before. Data which is again, incredibly valuable to both the private sector (companies like P&G to hedge funds) and public sector (worldwide CPIs, economic development groups).

As I continue to build out my thesis around “data 2.0″ I believe the opportunity both present and future for many of these companies is massive. In fact, I’vegone on record calling many of these companies future winners (or “Unicorns” if you’re into that). And what is most exciting is that, while some areas of tech in the past have taken years of growth in a specific geographic concentration, the data 2.0 trend is inherently global. Multiple companies mentioned above (and many others that I’ve left out for now) are being founded and doing incredible things in countries all around the world. This is a global movement.

In the coming weeks I’ll be synthesizing these thoughts into a series of reports (similar to my report on India online grocery delivery). The first will be around the rise, opportunities, use-cases, and the companies operating within the satellite space, some of which I’ve outlined here.