The e-scooter has strong pricing ratios between CapEx + OpEx compared to its bike, e-bike and e-moped counterparts. That’s why it is often perceived as the sweet spot of the free-floating mobility world. And yet, when digging into the unit economics of the first generations of e-scooters, we can begin to understand why operators are struggling to cut the fierce cash burn.

The breakeven for e-scooters takes almost 4 months, and with hardware durability issues and vandalism, operators are unprofitable in many cities. While we expect new generations of e-scooters to have longer lifetimes, we’ve seen that these issues remain as some of the key hurdles that makes current e-scooters uneconomical.

The first generations of e-scooters have struggled in reaching profitability

At Aster, we’re really into electric mobility and we’re happy to see it’s picking up pace! E-micromobility is adding new layers of flexibility to urban transit, and more than half of all passenger vehicle sales are projected to be electric by 2040. That said, one key hurdle remains for EVs of all sizes: charging.

It goes without saying that the operating time for an electric vehicle is limited due to charging. For passenger cars, this means people are worried they won’t be able to use their vehicle when they need it. For micromobility operators, charging creates hefty operational expenses and a pool of untapped value for when their vehicles are recharging and off the street.

In comes battery swapping, enabling EV owners and operators to replace depleted batteries with fully charged ones in less than a couple of minutes. But which players are developing battery swap solutions? Where does swapping struggle and where can it add value? What could battery swapping look like in the future?

We’ve studied up on battery swapping over the past couple of months, and after digging into the value chain and exchanging with startups in the space, we’d now like to shed some light on these questions.

The battery swap startup landscape and value chain

We found that the battery swap value chain could be broken down into three main segments: (1) vehicle design and manufacturing, (2) battery swap station design and installation and (3) micromobility operators using battery swapping in their business. Most of the startups we came across operate in one segment of the battery swap value chain, except for a few players (such as Gogoro and NIO) that are integrated in both vehicle manufacturing and the swap station design and installation.

The companies we’ve come across offer battery swap solutions for two main use cases, including passenger cars and micromobility.

Companies have struggled in scaling passenger car battery swapping

Some players have attempted making passenger cars with swapping capabilities, the most infamous of which being Better Place. After raising roughly $800 million, the Israeli startup ended up bankrupt in 2013. Tesla also tested swap stations in 2013 for Model S owners, which it ultimately scrapped and deployed supercharger networks instead. Now NIO is taking a swing at passenger car battery swapping, but has had a tumultuous year with significant staffing cuts and struggling financials.

So why is it so challenging for companies to scale electric passenger cars with battery swap capabilities?

We believe there are 3 main reasons:

1. High installation costs: Building a reliable battery swap network is CapEx intensive. For Better Place, each swap station was initially projected to cost $500k but ended up with a price tag of roughly $2 million. These costs are remarkably higher than direct-current fast charging (DCFC) stations, where installation costs vary from roughly $25k to $200k.
2. Complex go to market strategies: OEMs must agree with battery swap startups to build vehicles that are specifically suited for the design of the swap station. Even more, as there is no battery standardization between different electric cars, the stations will likely have unsustainable inventories of multiple battery types as new cars are released.
3. Little acceptance from end–users: It seems increasingly unlikely that electric car owners will use swap stations as plug-in charging networks are becoming faster and more ubiquitous. In the case of Tesla’s pilot program, only 4–5 out of the first 200 participants were interested in swapping batteries. The low interest level was likely a result of the fact that a Model S could be charged for free and at high speeds (80% capacity in about 40 minutes), while in the swapping pilot, Tesla owners had to pay and book appointments to get a full charge, making the overall swap experience less compelling compared to Superchargers.

However, battery swapping can add a lot of value for money-losing micromobility startups

Micromobility is expected to boom in the coming years, with estimates of the e-scooter market alone reaching $40-$50 billion by 2025. But startups in the space need to move quickly to improve their unit economics to capture this market, as achieving profitability has been a significant hurdle so far.

Charging is a large operational expense for e-scooter companies, especially for those using a “crowdsourced” model, where individuals are paid to locate, recharge and rebalance scooters. Bird is an example of a company that has suffered as a result of this model. At one point in 2018, 47% of their total cost of revenue went towards charging expenses as a result of the crowdsourced model.

Other e-scooter companies have opted for hiring operational teams instead to collect the e-scooters and take them to charging and maintenance facilities. However, this model still requires that the scooters are off the streets for several hours to recharge, reducing overall revenue potential.

This is where battery swapping can make a meaningful difference. It adds two layers of value for micromobility operators, as it can reduce charging costs while also maximizing vehicle uptime. Swapping can be used in two different models:

1. Operational battery swap teams: Companies can hire operational teams that are tasked with swapping batteries and taking depleted ones to a charging facility. This model has recently caught the attention of operators in the space and is currently used by Uber’s JUMP. Skip recently announced they plan on switching to this charging model as well.

2. Swap stations: Operators can also opt to deploy stations or kiosks in retail stores or city streets, where users are incentivized in the form of ride credits to swap the batteries on their own.

Operational swap teams and swap stations can reduce charging expenses as a percentage of revenue per ride

But why would battery swapping work better in micromobility than it has with passenger cars? For one, it’s much easier for micromobility operators to agree with light vehicle OEMs to create battery swap capable scooters and bikes compared to passenger car OEMs. Also, battery swapping in micromobility can enable an incentive system where end-users are encouraged to exchange batteries to earn ride-credits, a mechanism that did not exist for passenger car swap stations.

So what’s next for battery swapping?

Passenger car battery swapping remains complicated due to high installation costs, complicated go to market strategies and unclear user acceptance. However, as the e-scooter and broader micromobility markets grow and consolidate, operators will be on the lookout to boost margins using operational innovations. We expect battery swapping to play an important role in helping these companies remain competitive by beefing up their unit economics and maximizing vehicle uptime.

But which battery swapping model will be adopted by operators? What will the role of hardware manufacturers be as e-scooters and e-bikes move towards battery swapping? Will operators develop their own swapping technology, or rely on 3rd parties such as Pushme or SharingOS? Will customers be swapping batteries in the future?

JUMP recently hinted they are considering moving towards user-based battery swapping. Although this charging model makes sense from a unit economics perspective, several hurdles remain, including user acceptance, creating a sustainable incentivization strategy and finding locations to place swap stations.

Dozens of start-ups have emerged over the past few years boldly announcing we have the technology; flying cars are here (or almost). Serious money is being thrown at electric and hybrid aviation, and credible backers are not lacking. Even the incumbent aircraft manufacturers are backing start-ups and launching their own in-house development programs. At the same time, the detractors say yes, maybe one day we’ll fly electric, but not anytime soon, mostly due to limitations of current battery performance.

Over the past 30 years I’ve experienced some fundamental industry transformations as a banker, investor and entrepreneur. I’ve seen technology innovation fundamentally transform massive industries such as telecommunications, computing and energy.

Today, I am seeing the signs of such a transformation in aviation. This article looks at the background to electric aviation, the challenges it faces today, and some indications about where attractive investment opportunities are likely to emerge.

1884: Electric Airships

It is well known that the electric car made a brief appearance at the beginning of the 20th century, only to be overwhelmed by the superior performance range of gasoline-powered vehicles.

Less known is the fact that the first powered flights were electrically powered and with Vertical Take-Off and Landing (eVTOL) capability The first powered aircraft, the dirigible “La France,” flew an 8 km trip around Chalais Meudon in 1884, using bespoke electric motors powered by a zinc-chlorine battery with an energy density of around 30 Wh/kg. Despite a series of successful flights, the aircraft’s range was too limited by its batteries to be further developed, and the glory days of the Zeppelin era were powered by diesel. Gasoline and other oil derivatives have reigned over aviation ever since.

Source: Public Domain

Why? Because fossil fuels simply pack in so much more energy than batteries for a given weight and aircraft range is highly sensitive to the energy density of its propulsive fuel, as described by the Breguet range equation. Electric “fuel” is further disadvantaged compared to conventional fuels because batteries do not lose weight as they discharge, further reducing range compared to combustible fuels.

1973: Electric Airplanes

It was not until the first oil crisis that electric powered aircraft made its reappearance: in October 1973 the first electric airplane (fixed-wings) flight took place in Austria based on a repurposed glider and was powered by VARTA nickel-cadmium batteries and a Bosch electric motor (borrowed from a forklift). The flight lasted for 14 minutes and achieved an altitude of 360 meters. As the oil crisis abated, so did interest in electric aviation. Only half dozen or so electric airplane projects had flown by the end of the last century.

Source: HB-Flugtechnik

In more recent times, an important catalyst for the revival of electric aviation has been NASA’s Centennial Challenges program, a series of innovation prizes launched to mark the Centennial of the Wright Brother’s first powered airplane flight in 1903. In particular, the Green Flight Challenge, designed to push the limits of energy efficiency of flight and which featured the largest cash prize ever awarded in aviation history (total purse of $1.65 million, sponsored by Google), drew 14 entrants. The challenge was to fly 200 miles in less than 2 hours using less than a gallon of fuel (or energy equivalent) per passenger. In 2011, the contest was won by the Slovenian Pipistrel flying an all-electric airplane, using lithium polymer cells with an energy density of 173 Whr/kg. Pipistrel remains one of market leaders for electric light aircraft and is currently developing an eVTOL (as a partner of the UBER Elevate initiative).

Source: Pipistrel

The Green Flight Challenge illustrated that electric aviation had come of age; and that battery technology and electrical propulsion systems had become sufficiently performant, at least for small electric aircraft.

2009–11: Electric Helicopters

Vertical take-off and landing aircraft are more demanding on power sources than conventional aircraft, which benefit more from the forces of lift to stay aloft. It was not until 2009 that a fully electric helicopter was developed and flown in 2011 by Pascal Chrétien and Solution F in France, as a demonstrator. The aircraft was powered by Kokam lithium cobalt pouch cells which had an energy density of 160 Wh/kg. So even for helicopters, battery technology had reached a degree of maturity sufficient at least for short flights.

Source: Solution F

Flying Cars, electric and hybrid airplanes

These clear demonstrations of the capabilities of electric propulsion and battery performance heralded a new era in aviation and drew entrepreneurs, capital and new ideas to the sector, many driven by a “clean tech” ethos. For example, Joby Aviation (“…advancing the transition to sustainable aviation”) was founded in 2009, Kitty Hawk in 2010 (“…we will improve our urban and natural environments with efficient and reliable mobility”), Volocopter in 2012 (“a green helicopter”), and Zunum in 2013 (“…get anywhere fast, and leave behind a healthy planet”).

Some of the incumbent manufacturers took notice of the potential of electrification around this time as well. Sikorsky (now part of Lockheed Martin) attempted to retrofit an existing product, replacing the thermal engine with an electric motors and batteries. And at the end of 2010, a “skunkworks” project was initiated at AgustaWestland in Italy (now Leonardo) to develop a fully electric tiltrotor eVTOL as a technology demonstrator. After test flights in 2011, the aircraft was revealed to the public in 2013. Airbus developed a 2 seater all electric demostrator (the “E-Fan”) which flew in 2014.

During this period played a key role in promoting the idea of electric and distributed propulsion, releasing several conceptual designs and funding a sub-scale tilt-wing hybrid demonstrator (“Greased Lightning”) which flew in 2015.

Drones

And another force came into play during this period: drones. As consumer drones increased in capability with better batteries and higher payloads, the commercial potential became apparent. New markets in aerial photography and surveying, inspection, agriculture, and security have been evolving quickly, and market overlap with helicopters is already apparent. Off-the-shelf drone technology also allows the rapid prototyping of new eVTOL concepts, and the push for delivery drones is forcing regulatory bodies to put in place air traffic management systems — which will be essential for the eventual deployment of “flying cars”.

UBER Elevate

Perhaps the biggest catalyst for electric aviation has come from the UBER Elevate initiative. Launched by the publication of a White Paper in 2016, UBER proposed to develop an ecosystem for urban air mobility as an extension of its existing “on-demand” ground mobility services. Subsequent conferences and industry gatherings have fleshed out a full vision of the ecosystem and design requirements (eVTOL common reference models) have been published. UBER has engaged closely with aircraft manufacturers (including EMBRAER, BELL, BOEING/Aurora, Pipistrel and KAREM) and the cities of Los Angeles and Dallas-Fort Worth, with the aim of launching commercial operations in 2023. UBER’s vehicle requirements will only accommodate fully electric aircraft and vertiport specifications do not allow for combustible fuel.

The promise of a huge market facilitated by UBER has drawn dozens of new entrants into the market since 2016. The website evtol.news keeps a running tab of eVTOL projects and start-ups — there are 167 of them listed as of May 14 2019, and many innovative designs are emerging.

Source: Neoptera

Big Questions remain

Above all, it’s about the performance of batteries. If it took engineers over one hundred years to progress from 30 Wh/kg to 160 Wh/kg, how long is it going to take to make the next leap? Aviation fuel (Jet-A) has an energy density of 11,000 Wh/kg!

The answer is that such a leap might not be necessary. Electric motors are inherently more efficient than thermal engines, by as much as a factor of 2. Moreover, electric propulsion scales linearly, enabling distributed propulsion. Distributed propulsion allows better airframe designs with superior lift/drag ratios than conventionally designed aircraft.

Hybrid designs, especially where a generator is used to power distributed electric motors, can allow aircraft to benefit from some of the advantages of distributed propulsion, and obtain longer ranges or higher payloads without having to wait for battery technology to improve. Zunum in the US and Voltaero in France are taking this approach.

But limitations in battery performance is unlikely to put much of a brake on this new sector for the simple reason that most flights are relatively short. Many routes (especially short helicopter hops, like Nice to Monaco) are well within the range of existing technology. Many more routes will soon be in range as improvements are beginning to be commercialized.

Source: OpenFlights / hasty data 2015

UBER Elevate specification calls for 300Wh/kg for aircraft put into commercial service in 2023, and it has assured its aircraft partners that it is working with battery partners to ensure this can be achieved. With the current growth of the terrestrial EV market, there is no lack of focus on improving batteries. Additionally, new chemistries, such as solid-state lithium (Bolloré, SAFT and others), lithium metal (Pellion Technologies for example) and lithium sulphur (Oxis Energy) are beginning to be commercialized. Oxis Energy recently announced a 450 Wh/kg cell, and funding for a factory to build them in Brazil. Even though battery technology remains in flux and certification challenges remain, it is unlikely that batteries will constrain the development of at least short-haul electric flight.

Other Forces at play

What’s really driving electric aviation and the market for “flying cars” and electric airplanes?

1) Environmental concerns. The initial impetus for electric aviation came from the “Green Flight Challenge.” Although the industry incumbents try to make light of it and claim that aviation accounts for merely 2% of global greenhouse gas (GHG) emissions, most observers agree that it’s one of the fastest growing sources of GHG. Many suspect that the 2% figure is a very significant under-estimate, especially if the warming effect of contrails are taken into account. The carbon footprint of commercial aviation has long been a pet-peeve of environmentalists and some have even vowed to stop flying completely, or at least severely limit it. A “shame of flying” or flygskam has entered into the public consciousness in Sweden to the point that domestic air travel demand declined in 2018, with a commensurate increase in demand of intercity train travel. Governments are beginning to respond to these concerns; the Norwegian government has mandated all-electric domestic flights by 2040. Environmental concerns may also become a key driver for the deployment at scale of cargo and delivery drones.

Source: Roland Berger

2) The flying car fantasy. No one likes getting stuck in traffic and if a similarly priced alternative is available, as promised by the likes of UBER, what is there not to like?

3) The promise of “distributed aviation” and regional air taxi services. The mega-hub system is designed to serve the operational efficiency of airlines flying large fossil fueled aircraft. Many people (in mature developed economies) live within a few miles of a local airfield, and their final destination is within a few miles of a local airfield. Regional aviation has been constrained by high costs and noise. If both of these constraints are removed, a renaissance in regional aviation may ensue.

4) Cargo drones are paving the way for eVTOLs. Easily accepted by society initially for the enhanced delivery of emergency medical services and transport of medical samples, it will be a small step to introduce air ambulances, especially if they are quieter and less costly than existing helicopters. Similar acceptability will be found for fire-fighting. As functional drones are normalized, last mile delivery is likely to become more acceptable, especially if lower carbon footprints can be demonstrated.

Turbulence

Let’s assume that incremental improvement in batteries is sufficient to economically serve short urban/suburban and even regional routes.

What could still prevent this market from emerging at scale anytime soon?

1. Noise — all of the new market entrants are promising quieter aircraft, sometimes much more quiet than existing ones. If this benefit does not materialize, public acceptance for new airport facilities and increased usage of existing ones may prove to be very difficult.

2. Costs — as with noise, everyone is promising a big reduction in costs, especially in operating and maintenance costs. While it’s clear that “electric” fuel is usually less expensive than the hydrocarbon equivalent, it may not be the case in all refueling locations. Also, battery life for this new application remains uncertain, and they may need to be replaced more often than anticipated.

3. Infrastructure — many new infrastructure elements need to be put into place for this industry to emerge, including the obvious “vertiports” for flying taxis, but also the electrical charging infrastructure. According to some estimates, even a modest recharging facility at an airport will require a doubling of existing electrical power availability. Existing air traffic management systems will need to be upgraded or supplemented to accommodate increase traffic, and traffic flying in new corridors. Who will pay for these infrastructure elements remains to be seen.

4. Certification — it goes without saying that safety is of paramount importance. Commercial air transport has an exemplary safety record, and it’s highly unlikely that standards or passenger expectation will be lowered. Certification bodies face major challenges to evaluate and certify new technologies, and are likely, given recent events surrounding the 737MAX, to be overly cautious.

5. Pilots — a renaissance in regional aviation and the emergence of ultra-short-haul eVTOLs will require a significant increase in licensed pilots. Certification agencies also need to adapt licensing to accommodate the new technologies of eVTOLs and electric propulsion. Autonomous systems can overcome this constraint, if and when certification and regulatory authorities agree.

Paradigm shift or not, the race is on to build and deploy these new electric aircraft. Long term success will depend on how well and how quickly market participants navigate through these areas of turbulence.

For investors, deciding where to place bets is challenging. Let’s first consider who makes money in aviation today; the most profitable links in the air transport industry value chain are the ticket sellers and the computer reservation systems. This throws an interesting light on UBER’s positioning in the industry. Other attractive segments include all freight forwarders, navigation and other services provided to the airlines.

To what degree this value chain holds in the transition to an electrified industry remains to be seen, and a big question hanging over the sector is: who is going to take the risk to purchase, finance, insure, and operate these new aircraft?

Attractive investment themes in electric aviation

Aster has been investing in early and growth stage innovators driving industry transformations for the past 20 years. We’ve developed a number of investment themes in electric aviation and urban air mobility which leverage and complement our existing investing activity in mobility, energy and industry. Our main themes in aviation include:

· Ticket sales and computer reservation systems. There is already a brisk business in fractional ownership and reselling capacity on private aircraft, including helicopters.

· The supply chain of the new aircraft manufacturers — batteries, battery management systems, electrical control systems, motors; avionics, sensors, navigation aids, autonomous pilot systems. High fidelity design tools for electric propulsion.

· Infrastructure and services — including vertiports, warehouses for delivery drones, air traffic management systems, battery charging, maintenance.

· Cargo and delivery drones — buoyant existing market for equipment and services, fewer regulatory constraints.

· Aircraft manufacturers — rigorous assessment of management and technical teams required, as well as technical review of design concept and execution capability. Entry barriers seem deceptively low; many start-ups will hit the wall as they attempt to undergo certification of their aircraft.

Paradigm Shift?

We most probably are in the midst of a paradigm shift: the underlying assumptions about the air transport industry may no longer hold. Market entrants arriving with new technologies present both challenges and opportunities for the incumbents. Airport planners and civil aviation authorities need to consider that growth may also come from regional airports rather than just mega-hubs. Cities and regions need to consider the benefits and risks of introducing urban air mobility and cargo drones. For the old guard a paradigm shift represents both threats and opportunities: existing airlines may be surprised by how quickly new entrants begin to erode the edges of their already fragile businesses.

While the annual growth of e-commerce is expected to reach 11.7% by 2025[1] with the result that the number of parcels shipped in Europe will double, urban logistics has to juggle with ever more constraints: legal constraints on the one hand, with many European cities setting access restrictions on their city centers to diesel vehicles; and economic constraints on the other hand. For example, the closure of Foodora’s meal delivery company business in France has highlighted the fragility of platform models for last-mile delivery because they do not allow the pooling of flows (for those most interested in this subject, I recommend reading this McKinsey analysis: The urban delivery bet: USD 5 billion in venture capital at risk.

In this context, can trains, subways, tramways and the Paris RER, for instance, which criss-cross large cities daily with a low carbon footprint[2] answer the current challenges of urban logistics?

While the potential of co-modality may be emphasized by Jerome Libeskind who quotes some examples in his work Urban logistics in Japan, this expert reserves his opinion on the ability of rail transport to provide a solution to urban logistics: “Rail mode, which is very rigid and often saturated, will not bear up to the competition […] in the face of highways made “clean” again and which bring flexibility of use”[3].

For our part, we wish to point out the intrinsic strengths of rail mode and the resulting opportunities in this growing market.

Stations are natural hubs within urban areas

Places of transit par excellence, the world’s largest stations every day see hundreds of thousands of people pass: between 500,000 and 700,000 people per day pass through Paris Gare du Nord, London Waterloo, Hamburg Hauptbahnhof, etc., while the figure for Tokyo’s Shinjuku Station is said to be more than 3,000,000[4].

“The station of the future […] will connect people, commercial flows and data all at the same time”. This is the conviction of Patrick Ropert, as the former CEO of SNCF Gares & Connexions explains in his book Citybooster[5] published in 2017. For this reason, picking up parcels at a collection point in a station looks to be an interesting compromise between economic efficiency and fluidity of the customer journey. The experiences of self-service pick-up lockers in rail stations, such as those operated by the La Poste Group company, Pickup, and Amazon Locker, are promising.

Digital tools are an opportunity for the train to be put (back) into the urban supply chain

Let’s stimulate our imagination by analyzing some start-up models that could be taken and applied to our subject:

• E-merchant X uses the Byrd (D) platform to manage its shipments. The Byrd algorithm — depending on package size and destination — determines the most suitable partner carrier.
• Then there is suburban train operator Y which can perform real-time tracking and predict numbers in its trains thanks to the Opencapacity(UK) platform. The latter merges data from ticketing, surveillance cameras and weight sensors and extrapolates them with artificial intelligence. The operator sells seats available at off-peak times to its network of parcel carriers.
• Carrier Z receives the transport order from e-merchant X. Thanks to its Wise (US) routing optimization software, it decides to consolidate its flows on the day via rail mode and organizes the last-miles with its network of cyclists equipped with K’ryole (FR).

A clever mix of Internet of Things, artificial intelligence and digital platform will allow operators to share their available transport capacity and the delivery status of packages in near-real-time. Customers will then have an ecological, capacitive and inexpensive mode of transport if they make use of the spaces available in off-peak hours. Resolution of the economic and operational equation will undoubtedly greatly depend on how the modal disruption and last-mile are organized. Rail station layout will be fundamental in this respect as will finding the right combination with “soft” transport modes. In these areas, the creativity of start-ups and the proactivity of our cities do not mean there won’t be surprises in store.

The train, new e-commerce platform?

Beyond the possibility of conveying parcels by rail, these means of transport could themselves become places of life in which purchase and collection actions are an integral part of the passenger experience.

Last November, the Telegraph reported that deliveries of Amazon packages would be possible on board the trains being proposed by Bombardier and Hitachi for the London-Birmingham line. According to the article, passengers could, for example, order a pair of sneakers on board the train and have them delivered right to their seat at the train’s next in-station stop. True bill or fake news in the context of a close struggle between bidding manufacturers for this huge tender?

In an “economy in which attention is the scarcest commodity and the most precious source of value”[6], the idea of encouraging consumers to shop while on their journeys is gaining ground. Start-ups are positioning themselves to capture and monetize some of the attention available from passengers during a flight, an Uber ride or a train journey. While Skydeals (FR) has been exploring the lie of the land with airlines for a few months, Enroute (IL) has just started an experiment with Deutsche Bahn with an original model: passengers who agree to do their shopping on board a train get a free journey!

Thus it may be seen that the rail passenger transport system, with its structuring axes, its stations and its thousands of potential consumers using the network, in a way focuses on the major issues of urban logistics: ordering, routing and withdrawal. To take it out of its marginal role, all that remains to be done is to convert stations into multimodal platforms. 

[1] Grand View Research, September 2017

[2] https://www.oui.sncf/aide/calcul-des-emissions-de-co2-sur-votre-trajet-en-train

[3] What future for urban rail freight? Blog Logicités, Sept 2018

[4] https://www.worldatlas.com/articles/7-of-the-busiest-train-stations-in-the-world.html

[5] Citybooster, Patrick Ropert, 2017

[6] The economy of attention, the new horizon of capitalism? Under the auspices of Yves Citton, Edition La Découverte, 2014

After the first episode “SW is eating the world”, the second opus “AI transforms all sectors” is now out. But let’s not stop to this noisy tagline. AI does not transform all companies. And AI does not transform all sectors as fast as one could think. The main reason behind that is simple: deploying artificial intelligence is resource intensive and challenging for most businesses. Specifically, companies have a hard time attracting and recruiting talents in data-science. Lack of speed, lack of budget… startups look much more attractive to many young graduates. Large corporates such as Air France or Renault make a try to spread AI within their organizations by launching some agile and well-funded internal entities (aka. digital factories), mimicking the start-up codes. One thing is for sure, for those who can’t afford to create such entities, meaning most SMEs and plenty of large corporates that are lagging behind, hiring experienced data-science profiles is a nightmare. And transforming their organization with AI seems a distant dream.

Things could change: computing performances continue to increase as fast as costs drop, AI continues to be a field where the open source is never that far from the state of the art, and AI requires less and less training. These three factors have enabled the development of platforms that facilitate or even automate the creation of machine learning applications. A kind of AI to create AI applications. The so-called Automated Machine Learning (or autoML) platforms aim at enabling data-scientists or even any engineer to build and use AI models.

To understand how AI automation can be enabled, one needs to go back to the basics and understand how AI applications are usually created:

• Step 1. Design and Discovery: identify a business opportunity, estimate the value creation potential, and estimate the feasibility based on the data-sets that are available.
• Step 2. Development of ML Models: clean the data-sets, annotate them if needed, define the output of your application, find the right models and algorithms to estimate the output with the best accuracy.
• Step 3. Deployment of ML Models: create a prototype that turn your algorithm into a usable application, test it, integrate it into the workflows of your customers or employees, and change behavior.

Finding, testing, and training a lot of algorithms at Step 2 is time-consuming. And finding the right one with a limited number of iterations is the privilege of a few experts. The promise of autoML is to automate or accelerate these repetitive, costly, low added value tasks. It is enabled thanks to a significant increase in computing capabilities and in the availability of algorithms. Additionally, having a unique tool covering the last two steps reduces time-consuming data manipulation.

DataRobot, a US-based company that recently raised an impressive $100m Series D Round, or H20.ai, are the first leaders in this new category of autoML tools. And the autoML trend is not expected to vanish as fast as it came. Three reasons, and three recommended readings:

• Billions have been invested in start-ups developing specific chips for AI computing in 2017 and 2018, without mentioning the endeavors of cloud giants to build their own chip, such as Google with its TPU. Moore’s Law is outdated, but the computing capabilities to tackle AI challenges are still progressing at a very high pace. AutoML platforms that test and learn different algorithms will get faster as hardware performances increase.[1]
• The AI Community is mostly built around open-source frameworks provided by tech giants such as Google and its TensorFlow framework. Open-source framework attract contributions or research papers from the community that fasten the development of the framework and thus increase user adoption, enabling a virtuous cycle. Auto ML platforms will benefit from the wide availability of this intense research activity.[2]
• Transfer Learning technologies reduces data requirements. Models that have been trained on extensive data-sets for a specific task might be suitable, as a starting point, for another problem. Models that have been trained for visual recognition are sensitive to shapes, edges, etc. They are thus an interesting common ground for any visual recognition application. By reducing the need to tag important data-sets, Transfer Learning lowers the barriers for any autoML platform.[3]

As the market gets larger, subcategories are thus emerging: some platforms focus on learning from specific data types (Clarifai, images), some focus on one specific sectors (Owkin, medical sector).

We do expect the creation of industry specific platforms, that would stand out from others in their ability to deploy the freshly created AI applications in bad connectivity environments, “on the edge”, to enable the creation of applications from smaller datasets, and to avoid the black-box effect by pinpointing the root-causes of a prediction. One thing is for sure: if these platforms might be of great help to accelerate the job of data-scientists, the simple user-interface of these tools should also widen the target group of users to business analysts and any “dataholic” engineer, solving in some cases the tricky hiring challenge of many companies.

[1] More info here from Patrick Shafto, Associate Professor @ Rutgers University Newark: https://blog.hardwareclub.co/theres-an-investment-frenzy-over-ai-chip-startups-d9b5ea42b5c4

[2] More info here from Patrick Shafto, Associate Professor @ Rutgers University Newark: https://theconversation.com/why-big-tech-companies-are-open-sourcing-their-ai-systems-54437

[3] More info here from the Integrate.ai blog: https://medium.com/the-official-integrate-ai-blog/transfer-learning-explained-7d275c1e34e2

In early January, I got the chance to attend CES Las Vegas for the very first time. I joined the VC Aster only a few months ago after several years in consulting, and this was an amazing opportunity to get my hands dirty at the world’s hottest tech innovation event.

As you can expect, everything there was huge, there were a tremendous number of startups to meet, events to attend, conferences to join, corporate booths to visit… And funny enough, I felt like large companies have switched places with startups. If you wanted to be inspired and get a taste of what the world could look like in the distant future, then the corporate halls were the place to be. During CES, corporates deploy significant efforts to prove they are innovation leaders in their fields, sometimes at the expense of common sense. On the other hand, startups have a more pragmatic approach (at least the ones I met with). They need to reach markets in the next few years to survive so they are working on more medium-term innovations. Reconciling both parties around common interests is the biggest challenge of all!

As mobility is one of Aster’s core focuses and was heavily show-cased at CES, I spent a lot of time in the LVCC North Hall, which was dedicated to vehicle technology. After driving into the future, I decided to walk back to the startup world at Eureka Park, and I took the chance to wonder: in the era of autonomous driving, what will be the new relationship between riders and their cars? How will new mobility use cases be enabled by 5G? How can cities prepare today for what is coming tomorrow?

The travel experience will change drastically, but a car will stay a car

Two major trends impressed me regarding the travel experience: the “tailored travel experience” and the “augmented travel experience”.

Many car manufacturers and OEMs (Faurecia, Valeo, Honda, Kia) are working on the “tailored travel experience”: when riders enter a car, their physiological and psychological states are assessed (cardiac frequency, skin temperature, clothes, mood/expression, sweat, etc.). The travel experience is customized to best fit this state: light, temperature, sounds/music, odors, itinerary, seat tilt, etc. OEMs and car manufacturers study the human body to better design car parts in order to meet the highest level of comfort. Faurecia is working with ESP Consulting on this topic. This trend started to appear at CES last year but now everyone is working on it.

Regarding the “tailored travel experience”, it seems technology has passed the stage where it was merely relieving people of hard and troublesome chores or providing tools to help/accelerate a decision process. Now technology aims to take the decision in place of people, completely removing the need for a human action. The main goal is to make people “happy and comfortable” … and helpless some might say. This reminded me of a scene at the beginning of Wall-E movie.

A few players explored more precise use cases to offer an “augmented travel experience”. For example, Nissan goes further than just displaying basic navigation information on the car windshield. They are working on a 3D interface connecting the real and the virtual world, i.e., the new windshield, called “invisible to visible”. Using data collected from previous drives, the screen can adapt and change to a sunny blue sky if the weather is cloudy/foggy and rainy to increase visibility. On difficult roads (eg mountains) where the driver takes control, it also gives augmented content like a virtual car ahead, allowing the driver to anticipate the right path to take. If needed, a human assistant can also be projected on the seat next to the driver. This also finds applications in driving lessons. Finally, to find a parking spot, the driver can see through whole buildings and check streets nearby.

Another example came from the Valeo R&D department, which is working on the “augmented phone call experience”. In that case, the “driver” can see the interlocutor as a hologram seated in the back of the car (and hear him as if the voice came from the back of the car). On the other side, the interlocutor can wear VR glasses to talk as if he was indeed on the back seat. Value proposition is quite unclear at that stage, but CES is all about demonstrating new stuff!

All these innovations easily make us forget one thing… even though vehicle software evolves incrementally each year, the vehicle hardware (shape and bodywork at least) in itself does not change so much. Basically, it is still a box on 4 wheels. And yet, there surely is a lot to do to improve maneuverability, parking, traffic, etc. Hyundai is the only one to imagine a “walking car”, which can activate robotic legs to walk at 5km/h over rough terrain and a variety of obstacles. It might be outlandish and unbuildable, but at least Hyundai is thinking “beyond the range of wheels”. Not to mention all flying vehicles, with Bell’s hydrid-electric flying taxi, available via Uber by the mid-2020s. I choose to sort them into “small personal aircrafts” rather than “flying cars”.

In a short-term view, a lot has yet to be done to turn regular cars into connected cars, starting with a new telecom network made available for all vehicles. All vehicles presented at CES 2019 seemed to be connected and autonomous by default! A gap needs to be bridged by all players of the mobility ecosystem.

5G is not far away from enabling connected cars and autonomous driving

Even beyond use cases in mobility, 5G was a major theme at CES 2019, advertised both during the CES conference launch and throughout the whole show on the center stage. Verizon, the first US telco operator involved in 5G, spoke in many conferences and for them, words were not enough to describe the tremendous opportunity of 5G: “the next generation wireless network will be a quantumly wireless and speed connectivity for users”, “ 5G will only be limited by our imagination”, etc. 5G will indeed bring a high speed (until 20 Gbit versus maximum 1 Gbit for 4G and Wifi), high density (1m connected devices /km2 versus 2000/km² for 4G), high security and reliability, low latency (1ms versus at least 100ms for 4G) and low power (10% of current consumption) connectivity to many use cases, including some that are not currently using wireless connectivity.

As a matter of fact, connected cars will be among the first use cases for 5G. For instance, better connectivity on board (in cars, trains, subway, etc.) will enable enhanced infotainment (being able to download a movie in a moving car in a few seconds), parking and navigation services. Augmented and virtual reality applications will be infinite, and players are only beginning to explore the possibilities in the mobility area. Traffic management and safety will be improved, with better communication between vehicles and infrastructure. Verizon reminded that the lag time of the human eye is 150ms to see an obstacle, whereas the information would be transmitted in 1ms with 5G — for a car running at 100km/h, it means a distance of 1.5m VS 2.5cm.

A bit later down the road, 5G will enable autonomous driving, allowing more people to have access to mobility services: people who don’t like driving, the disabled and the elderly, etc. Non-driving tasks will be covered as well: hailing a self-driving taxi, monitoring a self-driving fleet, supporting riders through their journey. The first autonomous vehicles such as Waymo’s did not wait for 5G and are already on the road, for training and educating purposes. Indeed, explaining autonomous driving to the wider public will help legislators pass new bills.

However, in the short-term, first usage scenarios for 5G need to be targeted in a local approach in order to avoid waiting for a broad deployment. For example, a 5G enabled parking garage, a 5G enabled campus, with autonomous vehicle onsite only, or a 5G travelling corridor, alongside highways or subway routes. 2019 will see the first outcomes but 5G will still take years of worldwide effort to come into full effect.

Indeed, 5G is coming for the autonomous vehicle, but several challenges remain: the whole ecosystem is still waiting for frequency bands to be allocated, not just in the US but across the world. The right algorithms need to be developed and tested for different scenarios. Technological and regulatory challenges remain to densify networks and not allow dead notes (higher frequency means less coverage). Last but not least, investment from public authorities is necessary to upgrade the city infrastructure.

Still, a lot has yet to be done to prepare cities for new mobility solutions

Having a limited knowledge of what the mobility landscape will look like, cities are worried about the investment and regulatory decisions they may need to make for the 10 years to come. As a matter of fact, there are critical and pressing issues, since autonomous driving is already reaching the road. For example, Waymo autonomous ride-hailing services was launched in December 2018, but only as a demonstration since public authorities don’t have the policy framework for this yet.

The mayor of Baltimore took part in a panel discussion at the conference called “What’s next for transportation in the Smart City?”. She repeatedly asked about how to prepare Baltimore for autonomous vehicles and how to ensure access to public transportation for all citizens, regardless of their age, social status and neighborhood (what about narrow streets? Historical centers?). Her current challenges are the following: building bike lanes in the city (impact on elderly, street access for emergency services, regulation regarding helmet usage, other vehicle access to the bike lanes such as scooters, wheelchairs, etc.), electrifying public transport (impact on the grid), ensuring safety of the citizens without invading people’s lives (every vehicle could be a surveillance camera), and upgrading city infrastructure (including decaying roads) to install new technologies. These challenges all seem quite basic compared to the future of mobility showcased at CES. The need to upgrade our cities was far less in the spotlight than the need to have a stress-free and decision-free travel experience… And yet, these are real issues that will significantly slow down the deployment of connected and autonomous vehicles. Thankfully, some companies address part of the problem, such as the French startup Citymagine, which provides an automated infrastructure inspection in order to secure infrastructures and speed up the arrival of autonomous vehicles in our cities.

At Aster, we are keen on funding companies that will transform the travel experience and enable connected and autonomous driving. That is why we recently invested in Wluper, which is developing a conversational AI to provide navigational intelligence to mobility service providers. However, upgrading cities in a sustainable way, while simultaneously keeping economic, social and environmental challenges in mind is another one of our priorities.

“STRIKES: CONSIDER CARPOOLING” is a message that motorists in the Paris region could read on overhead signs on the A86 highway during walk-outs by railroad workers in the spring. Most likely, few of them connected the emergence of this new transport solution to the possibilities offered by artificial intelligence (AI). And yet AI is one of the technological building blocks used by Karos, the top court-voiturage™ (short-carpooling) app in the Paris region, to organize the liquidity of their marketplace.

AI — the ambitious goal of which is to create cognitive processes comparable to those of human beings — was developed as a specific field of research in the middle of the 20th century; however, it is only now that its promise is being realized as an everyday reality.

What is behind the recent surge of this technology and what specific prospects does AI offer for mobility?

An alignment of the planets is ushering in a new era for artificial intelligence

After various bouts of hype and disillusionment over the past 60 years, it appears that the time for mainstream, economically viable AI-based applications has come. This myriad of uses is the result of an alignment of the technological planets, including:

1. Algorithms which have constantly improved since 1950 and become increasingly widespread
2. The huge increase in the amount of data available: some of this data is produced by connected devices, the number of which will soar from 8 billion in 2012 to 50 billion by 2020
3. The significant increase in calculation and storage capacities: today’s smartphone has 1,000 times more computing power than NASA had to guide an Apollo mission!

“Data is the new source code”

With the implementation of AI, the days of programming machines appears to be over: let’s hear it for training machines! As input for machine learning, data therefore becomes the key ingredient.

GAFA and other BATX called it right when they made data collection and enhancement the cornerstone of their strategy, as did, even more so, the contributors to the Villani report in insisting on the need to organize the sharing and pooling of sector data to counteract the dominance of web giants.

If the battle over the massive collection of data is being played out among the giants, startups using AI as a technological building block are, for their part, concentrating on data preparation, uses and solutions and drawing the outlines of more sustainable mobility based on a simple equation

The Villani report on artificial intelligence published in March highlights three promising areas of application of AI: “technological developments and the expectations of the ecosystems are concentrated in three key markets: the development of self-driving vehicles, sustainable — particularly electric — forms of transport, and the emergence and interconnection of new mobility solutions to reduce traffic density in certain segments.”

The innovative products and services developed by startups illustrate the way in which these hopes are gradually becoming a reality.

For example, in the area of self-driving vehicles — which we agree must be shared in order to be sustainable — some startups such as RealSynth (D) and Cognata (IL) are developing solutions to virtually train artificial intelligence. The goal is to cover the maximum number of driving scenarios by saving millions of kilometers driven and to bring self-driving vehicles to market more quickly.

While DCbrain (F) and its Israeli counterpart Optiways (IL) implement their AI-based solution to optimize logistical flows, Karos (F) and Padam (F) are developing on-demand carpooling and transport as solutions for inadequate or limited mass transit.

AI can also be a good way to prepare cities and operators for the arrival of electric vehicles (fleets or individually owned). For instance, Cosmo Tech, a startup specializing in augmented intelligence, is gearing up to use its know-how to optimize recharging infrastructures and strategies.

Lastly, to take full advantage of the prospects offered by AI, a data cleanup and structuring policy is necessary. This is precisely the value of the solutions offered by Element Analytics (US) and Tamr (US).

* verbatim statement made by Nvidia’s CEO in March 2018 at the GPU developers conference

Adidas and the Berlin public transport company created a buzz recently by releasing limited-edition sneakers with an integrated fare ticket. In addition to this quirky story, traditional e-ticketing is undergoing a major transformation that could ultimately make the validation and payment process completely transparent for passengers. Following are a few keys to understanding a fast-growing market, particularly since e-ticketing is a channel through which to access passengers and therefore their mobility data.

Mobile devices: the ideal integrated e-ticketing and payment system

After an initial digitization phase in the early 2000’s with the introduction of NFC-based cards such as Navigo and Oyster, e-ticketing has reached a new milestone by allowing public transport users to pay and validate simultaneously, for instance using their bank card as a ticket. Operators see this as an opportunity to both reduce their collection costs (from 15% to 6% of the ticket price according to an Accenture study published in 2015) and improve the passenger experience by saving them a trip to the ticket machine.

Although bank cards and even biometric technologies are sometimes tested, mobile devices are well-positioned to become the ideal integrated e-ticketing and payment system. They offer numerous advantages, such as engaging with the customer by becoming a conduit for real-time information or for sales of additional services.

Traditional players are getting ready and startups are trying their luck

Traditional players are naturally taking numerous measures to adapt their product line. For instance, Siemens is promoting SiMobility JustGo based on Bluetooth Low Energy and beacons technology while moving forward with the acquisition of Hacon. The foursome created by Orange, Gemalto, SNCF and RATP are banking on NFC via their Wizway Solutions joint venture. For its part, Thalès’ e-ticketing division is still searching for its Prince Charming after the missed opportunity with Latour Capital in the spring of 2017.

At the same time, a myriad of startups are working furiously to develop clever solutions that appeal to operatorswho, for a few thousand euros worth of configuration, can test solutions that improve their customers’ seamless experience. For example, MotionTag is conducting a Check-in/Be out proof-of-concept with Deutsche Bahn thanks to the precision of its geolocation algorithm; Israeli startup HopOn is rolling out its ultrasonic beacon technology on buses in Tel Aviv and Jerusalem; and Atsuke has already penetrated a number of Swiss cities with its SMS-based ticketing system and is now targeting the French market…

Gated or not gated? That is the question

NFC, Ultrasonic, WiFi, SMS, QR code, geolocation algorithm, Bluetooth Low Energy…

How to navigate this litany of technologies and which one to choose?

A first market segmentation criterion is the presence or absence of gates in the public transport system. In fact, the presence of gates, a costly but important factor in the fight against fraud, hinders the use of some technologies: increasingly accurate geolocation algorithms (see MotionTag or GeoUniq) could, in theory, completely eliminate the validation process (Be In/Be Out). They are, for all practical purposes, unsuitable in about half of European cities that use gates in their subway system. In this market segment, the NFC protocol seems to be the popular choice, as evidenced by its recent adoption by New York City Transit.

The second segmentation criterion pertains to the fluidity of the passenger experience.

Integrated e-ticketing in route planners, a sure step in a multimodal world

The concept of mobility-as-a-service is becoming more of a reality every day, for example in tests in the Paris region concerning short-distance carpooling. Karos informs us in its September 2017 deployment report that 20% of its court-voiturage® (short-carpooling) rides are intermodal. The ability to pay for one’s ride in multimodal route planners will therefore be the next step.

E-ticketing therefore goes well beyond a quest for a marginal improvement in the profit margin of traditional operators. For them, it is about staying at the center of the choices made by the end customer, who has many options for getting from point A to point B. It is also about maintaining control of data, a firewall against the GAFA offensive, and the startups we met, which collect passenger data like cautious squirrels, fully understand this.