LYFT and UBER: Welcoming the New Robo-Cab Overlords

sagawa
Print Friendly, PDF & Email

SEE LAST PAGE OF THIS REPORT Paul Sagawa / Tejas Raut Dessai

FOR IMPORTANT DISCLOSURES 203.901.1633/.901.1634

psagawa@ / trdessai@ssrllc.com

twitter.jpg @PaulSagawaSSR

April 8, 2019

LYFT and UBER: Welcoming the New Robo-Cab Overlords

LYFT and UBER are market makers, brokering connections between riders and drivers and tapping into $4T+ in global spending on personal transportation with a service superior to previous alternatives. We believe the market is underestimating two major risk factors that may preclude either company from ever reaching profitability. First, we believe on-line consumer platforms, like GOOGL, AMZN, and AAPL are in position to disintermediate ride hailing apps from their customers, removing the friction of “shopping” for rides via AI assistants and raising pressure on pricing. Second, we believe autonomous robo-cab services – with significant cost advantage over traditional ride hailing – will spread to 20 or more US cities by 2025, with clear trajectory to nearly comprehensive coverage of urban/suburban markets by 2030. Neither LYFT or UBER is well positioned to compete in the robo-cab market, which we expect to exacerbate the competitive pressure from the on-line platforms. In this scenario, growth and progress toward profit might appear on track for several quarters, but we see the positive narrative likely to collapse further out. Investors may choose to trade the stock for the time being, but we fear the investment window will be short.

  • Transportation-as-a-Service addresses a huge market. LYFT’s roadshow pegged US household transportation spending at more than $1.2T. On a global basis, the spending is several times that, and while some of it cannot be addressed by TaaS and while TaaS is likely to prove somewhat deflationary, the TAM is obviously huge. The average US household owns 1.97 vehicles, with an all-in cost of $0.59 per mile traveled (but with significant variance depending on the type and usage of the vehicle). 85% of US car trips are less than 15 miles and 95% are less than 30. LYFT and UBER bookings were nearly $60B in 2018, on pace to grow nearly 50% in 2019, but are still a tiny fraction of total household transportation spending.
  • LYFT and UBER hope scale, network effects, self-driving will lead to profitability. While the ride hailers are growing strongly, neither is close to profitable or cash positive. Getting there depends on a few things: 1. Stable pricing (customer stickiness, marketing effectiveness); 2. Increasing commission rates (greater utilization for drivers, robo-cabs); 3. OPEX growth lags sales (wide consumer awareness, self-driving R&D manageable); 4. Auxiliary businesses successful (delivery, bikes/scooters); 5. No disintermediation (independent robo-cab services >5 yrs. away). IF the favorable scenario plays out, UBER could reach profitability in 5 years and LYFT in 8, but we are skeptical.
  • Direct consumer relationships could be disintermediated. LYFT and UBER act as brokers, connecting consumers in need of a ride with an inventory of drivers willing to pick them up. Having built critical masses of both riders and drivers, the ride hailing apps can make an efficient market, and importantly, erect barriers for would-be new entrants. However, the extraordinary consumer reach of on-line platforms, such as GOOGL, AMZN and AAPL, could let them establish their own market for transportation as a generic feature of their platforms. AI assistants, which could simultaneously check multiple suppliers and arbitrage by rate and availability once a user requests a “ride”, could disintermediate UBER and LYFT’s customer bases and pressure pricing.
  • Clearly superior cost structure for robo-cab model a huge threat to LYFT/UBER. Drivers take more than 70% of every dollar generated by ride hailing, absorbing the operating costs of the cars in their share. Considering much higher utilization, efficient servicing and maintenance, and the elimination of the driver, robo-cabs could cut the “driver take” by more than 60%. Moreover, a company with a fleet of robo-cabs would have substantial leverage to pressure broker fees paid to ride hailing networks lower and could sell rides directly, leveraging high-reach consumer platforms. Combined with a consumer on-line platform, like GOOGL, AMZN, or AAPL, robo-cabs could circumvent ride hailing networks entirely.
  • Ride hailers are poorly positioned for robo-cab opportunity. Both UBER and LYFT are investing in self-driving technology, but testing is far behind market leader Waymo and followers like GM/Cruise and Ford/Argo.AI. UBER suffered enormous setbacks in its once promising internal program – a consuming lawsuit with Waymo and a fatal accident spurred significant attrition of talent and disrupted development and testing. LYFT has chosen to collaborate with Delphi’s Aptiv unit (supposed partnerships with GM and GOOGL were never active). Neither have posted significant testing results and both have had user reports suggesting very significant limitations. Importantly, we believe that successful robo-cabs will depend on extremely detailed digital 3D maps of their service area. We do not believe that either LYFT or UBER can generate these maps on their own. LYFT partner Aptiv has shifted strategies, open sourcing its self-driving software to all comers for free with hopes of monetizing via 3D map subscriptions.
  • Early-to-market robo-cab leaders (Waymo, maybe GM) will win. We believe Waymo is 2-3 years ahead of GM/Cruise in reaching commercial quality service for its robo-cabs. We expect expansion to full driverless service for the metro Phoenix market before year end with the announcement of 2-3 new markets for 2020. 3D mapping of new geographies will become the gating factor for market expansion and we believe Waymo will set the standard. By 2025, we believe Waymo will have robo-cab service, sold directly to consumers, in 15-20 major US markets, with GM/Cruise commercial in a handful. This is a very difficult scenario for LYFT and UBER.
  • Near term ride hailing upside will give way to disappointment. Unicorns like to IPO with the wind at their backs. We see little indication that our longer-term concerns would affect the immediate growth and progress toward profitability of either LYFT or UBER in 2019. LYFT is a share gainer in the US in the wake of bad publicity for UBER, but the ride hailing pioneer should see plenty of growth outside the US. Still, the few good quarters ahead look like a bit of a chicken game with plenty of skepticism to keep investors hovering over the sell button and potential bad news from robo-cabs and AI assistants awaits. Current valuations depend on a narrative of likely profitability. We believe that an alternative scenario, where competitive pressures and the rise of robo-cabs slow growth and keep both LYFT and UBER unprofitable, is far more likely. As that narrative emerges, support for the stocks will similarly deteriorate.

A Bad Ending for the Ride Hailing Narrative

The LYFT story is simple. American’s spend $1.2T per year on personal transportation. Ride hailing services can eliminate the cost and bother of owning a car (or second or third car) and the hassle of driving (responsibility, parking, etc.) while greatly improving the experience (short waits, door-2-door service, vehicle type choice, etc.) and/or reducing the costs (vehicle utilization) associated with traditional alternatives (e.g. car rental, taxi, public transport). In this context, LYFT’s revenues – just the commissions earned after paying the drivers – are growing at a 103.5% (FY18) YoY pace, while rival UBER (which also competes in global markets) is growing 45.3%. So far, so good.

Neither LYFT or UBER is profitable or cash flow positive. In its S-1, LYFT outlined its plan to get there. First, scale will see sales growing faster than OPEX, particularly once the early phase brand building marketing spend is behind the company. Second, the split of booking fare between LYFT and its drivers will shift toward the company as rising utilization reduces downtime for cars. Third, autonomous robo-cabs will reduce ride costs considerably, enabling lower prices and higher percentage platform commissions while stimulating significant new demand. This also depends on competition, now almost exclusively with UBER, not getting more intense. We see two major risks that are significantly understated by LYFT.

The first the potential for new competition. The ride hailers are brokers between consumers looking for transportation and car owners willing to drive them. Success depends on reaching critical masses of riders and drivers sufficient to make a liquid market. This has been a formidable entry barrier, but one that we believe will be less formidable in the future. While UBER and LYFT have excellent penetration with their apps, their reach is small vs. the big consumer internet platforms – GOOGL, AMZN, AAPL and FB – which dominate user engagement. With the rise of AI assistants, like Alexa, Google Assistant and Siri, these platforms will be able to “front” the ride hailing apps, taking a request for a ride and arbitraging different options for price and availability. Indeed, ~75% of drivers work for more than one ride hailing platform and could offer rides directly for a new entrant with consumer reach. We believe that the technology of assigning rides and managing customer relationships is well within the reach of the big on-line platforms.

Second, LYFT presumes that it can profitably act as an agent providing riders for robo-cab companies. UBER presumes that it can develop its own viable self-driving platform to compete directly. Neither is likely. GOOGL’s Waymo, likely 3 years ahead of all comers in offering commercial quality robo-cab service, has no need of the ride hailers to reach customers. We believe GM/Cruise, number 2 by a good margin, would be best served partnering with AMZN or AAPL and would drive a very hard commission bargain to list rides on LYFT. Delphi’s Aptiv, LYFT’s current best partner has made the decision to open source its self-driving software, hoping to monetize via hardware systems and digital maps. Meanwhile, UBER’s self-driving program was set back years by scandals, losing much of its talent and many months of testing. We are not optimistic that it can make up the lost ground. Well-funded robo-cab operators will emerge in 20 markets or more by 2025 and pose an enormous threat to traditional ride hailing economics.

We believe that the narrative for LYFT and UBER, already skeptical, will get worse with the strong possibility that neither company will every reach profitability. Of the two, UBER, which has strong growth in international markets that may be much longer to see autonomous competition, seems the better bet.

Transportation-as-a-Service

TaaS is a huge potential market, addressing more than $1.2T in annual spending on personal transportation in the US and nearly $5T on a global basis (Exhibit 1). Much of that is spent on owning and maintaining private vehicles that are inefficient and lightly utilized, opening huge opportunity for services that might eliminate the costs and hassles of owning and operating a car for the right price (Exhibit 2). Ride hailing, pioneered by Uber and mimicked by Lyft, Didi Chuxing, and others, was a substantial disruption. Early critics dismissed the service – summoning a car to your exact location for an agreed price within a transparent arrival time with payments managed automatically – as merely a substitute for taxis. However, the significant improvement in the experience and price relative to taxis expanded the market beyond traditional livery, while the flexibility of working to one’s own schedule drew drivers to the platforms.

In the US, both Uber and Lyft have been able to reach critical mass in most markets with enough consumers and drivers on their platforms to make a liquid market in car rides (Exhibit 3). At the same time, that critical mass and their established brands are formidable market barriers to further new entry by competitors looking to ply a similar strategy. The market duopoly is growing the market rapidly – up 24% in the US for 2018 – and making progress toward profitability. A strong competitive moat for a small number of players competing for a massive potential market. What could go wrong?

Exh 1: US Historical Annual Household Transportation Spending, 2000 – 2018

Exh 2: Personal Vehicular Transportation Spending Breakdown in Billions, 2017

Exh 3: US Ride Hailing Platforms Revenue Growth and Forecast, 2016 – 2025E

Exh 4: US Ride Hailing Market Penetration, 2014 – 2018

When is a Moat Not a Moat?

Uber and Lyft are market makers, brokering connections between consumers who want rides and car owners who’d be happy to provide them for a price. As long as new rivals cannot quickly build a large inventory of drivers on the road and a way to reach those many thousands of would-be customers, rivalry should remain benign, but how good of an assumption is that? (Exhibit 5).

We are concerned that investors may be a bit too optimistic. First, the drivers are not employees of the ride hailing platforms, but independent contractors free to seek business from any source. Indeed, a full three-quarters of all US ride hail drivers already work on more than one network, usually both Uber and Lyft, who must compete for their attention. A new platform that was able to funnel ride requests to the drivers would be readily accommodated, and a network willing to offer a better rate would quickly gain their rapt attention.

Second, we believe that “apps” will face an increasing threat of disintermediation by smart AI assistants offered by consumer on-line platforms. Amazon’s Alexa, Google’s Assistant or Apple’s Siri would be able to arbitrage multiple ride hailing platforms for price and availability upon being asked by a user to find a ride, demand a share of the fare as a commission for booking the service, or even solicit rides directly from those independent drivers. These platforms will have considerably greater penetration and engagement with consumers than specialized ride hailing apps and will offer considerably more convenience as well. An assistant would field a request “get me a car to the airport” directly via voice command, a single tap on a calendar or map, or even without being asked at all, by assessing the users schedule, preferences and context

Exh 5: Functions governing ride-hailing platform’s value are very limited

– “Mary has a flight at 6:00PM, she appears to be leaving her 3:30PM meeting and heading to the lobby, it will take 45 minutes to get to the airport, so I will find her a car now”. While this functionality is not in play today, it is easily possible, and as the use of AI assistants spreads, likely. Ride hailing apps do not have access to the information needed to compete with that (Exhibit 6, 7, 8).

Exh 6: Digital Assistant governed TaaS model weakens ride-hailing platform positioning

Exh 7: AI Assistants are more suited to broker rides on behalf of riders

Exh 8: Ride Hailing platform reach pales in comparison with AI Assistants, 4Q18

This will pressure the ride hailing franchises to compete more vigorously for rides, as users easily shift their business with changing prices and availability, as on-line platforms demand a share of the fare, and as familiar brands like Amazon, Apple and Google step in front of their direct relationships with the consumer. This is a logical outcome of the increasingly aggressive competition between these behemoths – Tencent’s subjugation of apps like ride hailer Didi Chuxing into “mini-programs” beneath its WeChat platform is vision of our future. The most effective defense would be anti-consumer government regulation, uncharacteristic in the US but possible, even likely, in other global markets.

I, Robo-Cab

Robo-Cabs – fleets of autonomous vehicles able to pick up and drop off riders on demand within clearly defined and meticulously mapped geofenced territories – are coming. Although there are skeptics, Google’s Waymo has already begun a limited commercial launch in Phoenix with plans to deploy more broadly within months. GM/Cruise is working to offer its own service in San Francisco by 2020 (we think it will take a bit longer). From a ride hailing perspective, this is worrisome. More than 70% of current ride hail fares go toward the driver, who must also cover the operating and capital costs of his/her vehicle. With a robo-cab, there is no need to compensate a driver and the vehicle operating costs (e.g. fuel, insurance, maintenance, etc.) could be nearly 50% lower on a per mile basis. This will be a game changer for TaaS.

Lyft’s S-1 had this to say about Robo-cabs:

We are investing in autonomous technology and employ a two-pronged strategy to bring autonomous vehicles to market. Our Open Platform provides market-leading developers of autonomous vehicle technology access to our network to enable their vehicles to fulfill rides on our platform. Simultaneously, we are building our own world-class autonomous vehicle system at our Level 5 Engineering Center, with the goal of ensuring access to affordable and reliable autonomous technology. We believe that the strength of our brand, our trusted relationships with riders and our expertise in operating a ridesharing network at scale, as well as our two-pronged strategy to bring autonomous vehicles to market, will be competitive advantages that will enable us to capture value in the emerging autonomous vehicle ecosystem.

Each of the last three sentences of this paragraph carries a different desperate hope. Yes, Lyft is willing to let market-leading autonomous fleet owners fulfill rides summoned on its platform, but why would they want to do that? The clear market leader, Google’s Waymo, has already begun an independent strategy with limited commercial service in the Phoenix metro area (Exhibit 9). It has ample reach to the customer via Google’s armada of consumer apps (search, assistant, maps, etc.), the obvious expertise to build its own ride scheduling and customer management systems, world-class mapping capabilities, a nascent infrastructure ecosystem, and a 2-3-year head start on everyone. GM/Cruise is number two on the leaderboard and perhaps has a deficit in its consumer reach, but it is also the best partner for a consumer internet platform, like Amazon’s Alexa or Apple, and would wield substantial bargaining power over a ride broker like Lyft (Exhibit 10, 11).

The obvious technical partners for Lyft are not the market-leading developers, but rather the many ambitious start-ups. Because of the enormous cost of developing hyper-detailed 3D maps of broader service territories, these companies might be interested in cherry-picking very common trip itineraries – say

Exh 9: LYFT autonomous compared with other leading projects on crucial factors

Exh 10: Snapshot of Self-driving partnerships and potential match-ups

Exh 11: Per Mile Cost Economics of Traditional Ride Hailing and Robo-cabs

between big hotels and the airport, or around convention centers – which would limit their necessary investment. Delphi’s Aptiv, thus far Lyft’s most active partner, seems to have picked up on the value of 3D maps and has announced a strategy of making its driving software open source and free to all would-be licensees, but charging subscription fees for the maps it hopes to develop and selling hardware systems to make it happen. This mirrors the strategy of Baidu in China. Conceivably this could speed broader competition amongst new self-driving fleets, some of whom might want to use Lyft’s network, but we are skeptical. It also reveals Lyft’s lack of vision – the company could have invested to build its own mapping expertise but has not done so.

Make vs. Buy

Developing its own self-driving system would seem an extraordinary stretch for Lyft, which, despite its “Level 5 Engineering Center”, is short on world-class AI expertise and has not filed testing results for its work in California. With just over $500M in R&D spending (Exhibit 12) for its entire life, much of which must have been spent on other things than its nascent self-driving program, it is, unsurprisingly, far behind.

In comparison, rival Uber has been working on its own self-driving program for at least 5 years, making headlines in early 2015, when it gutted the Carnegie Mellon University robotics department and set up a major research center in Pittsburgh. However, much of that early momentum was lost in two major scandals. First, in 2016, Uber acquired Otto, a self-driving truck start-up that had been founded by Anthony Levandowski, who had been one of the most decorated scientists behind Google’s early self-driving efforts. However, Levandowski had allegedly absconded from Google with several hard-drives of proprietary information which he was accused of conveying to Uber. The resulting lawsuit was eventually settled in early 2018 with a $245M equity payout to Google and a promise to refrain from using Google’s IPR, but damage was done over the course of the 18-month legal battle. Top Uber scientists fled, and the Pittsburgh lab closed.

The talent flight accelerated shortly after the settlement, when news broke that an Uber autonomous test vehicle had hit and killed a pedestrian in Phoenix. Reporting suggested that the self-driving initiative had been cutting corners to be able to demonstrate progress to the new CEO Dara Khosrowshahi at an upcoming review, and that important sensors (LiDAR) that might have identified the pedestrian on the poorly lit road had been turned off so that the demo might run more smoothly. Uber promptly discontinued its testing program for more than 9 months, returning in a much less ambitious scale at the end of 2018. In the interim, Uber had set up a new autonomous driving development center, this time in Toronto with the cooperation of robotics professors at the UofT. At one point, Uber’s program had been considered second only to Google in self driving – now it is, at best, third tier, behind its erstwhile rival Google and the GM/Cruise initiative. Map building, which might have been a considerable strength given the millions of point-to-point rides tracked by the company’s software, appears to be a sorry afterthought, with many of its own drivers choosing Google’s public maps over the Uber platform (Exhibit 13).

Exh 12: LYFT’s lower than peers R&D budget raises concerns about autonomous capabilities

Exh 13: Autonomy will shift value from Ride-Hailers to Internet platforms that lead in all 3 crucial elements of Self-driving Technology

Who Wins?

This brings us to the final sentence of Lyft’s S-1 blurb on self-driving. It is asserting that its brand strength, relationships with riders and its expertise in operating a ride sharing network at scale are significant competitive advantages that will allow it to capture value from robo-cabs as they emerge. We believe that neither Lyft or Uber have the brand or consumer reach to beat the likes of Google, Amazon, Apple, or even, Facebook, should they choose to play as autonomy disrupts the ride hailing market. Without drivers to worry about, the challenges of operating a ride sharing network at scale would be trivial. The value will flow to those that actually control robo-cab fleet, those that have up-to date hyper-detailed 3D maps, and those platforms with inside position to discover consumer needs and handle consumer requests.

That sounds an awful lot like Google’s Waymo. Despite the challenges of preparing for the long tail of potential situations that a driver might face, with perhaps 15 million miles of training on public roads, meticulous testing for edge cases at its private testing environment in California’s central valley, and more than 10 billion miles of simulation, Waymo has begun a cautious roll out of commercial services. By cautious, it has returned safety drivers to its cars for the time being, although it already operated without them for a few months last year, so it can work out some remaining hiccups in executing smooth pick-ups at the rider’s exact location. The company is building out a second depot that will more than double its capacity for servicing its fleet, a strong indicator that it feels ready to expand its service to the whole of the Phoenix metro area before year end, an expansion that will likely coincide with removing the safety drivers for good. It has the beginnings of a strong ecosystem, with Avis and AutoNation on board to service and repair the fleet. It is beginning to ink deals with partners like Walmart to pay for customers to visit their stores. We expect Waymo to announce new markets for service in 2020, likely from amongst the cities (Austin, Atlanta, San Francisco/Silicon Valley, Kirkland, Detroit, etc.) where it is currently testing and, importantly, building those detailed 3D maps.

GM’s Cruise Automation business is conducting its testing in downtown San Francisco, a location that, it likes to point out, presents unusual difficulties for self-driving. Hardened on that city’s famous hills, narrow one-way streets, rampant double parking, and bold pedestrians, GM believes its system will be ready for most any city once it masters its first. Of course, the scuttlebutt is still mixed. First hand reports suggest the cars are flummoxed by unprotected left-hand turns, overhanging branches, and those maddening double-parkers. The statistics collected by the California Department of Motor Vehicles put GM’s rate of human intervention about where Google had been two years ago, and there are still no signs of a realistic business system to reach customers, book rides and maintain a fleet. We expect Cruise to miss its target of commercial service by 2020, but nonetheless, GM’s progress looks far ahead of the rest of the pack. We see Amazon as a natural partner. While GM made a VC investment in Lyft, we emphasize that there has been NO commercial relationship between Cruise and the ride hailer.

The Sword of Damocles

There is a reason why LYFT and UBER are bringing their IPOs to market now, and we don’t expect either of them to fall down right out of the gate. Demand for ride hailing is robust, both in the US and abroad. Consumer internet platforms have not turned their attention to TaaS and the penetration and engagement

Exh 14: SSR Commentary on Parameters in Modeling Lyft Financials

Exh 15: SSR Estimate of Booking volume on Lyft’s platform, 2017 – 2030E

Exh 16: LYFT will have to keep driving commissions upwards to sustain revenue growth – increasingly hard without ancillary business and driver pushback

Exh 17: SSR Estimate of LYFT Revenue after driver commissions, 2017 – 2030E

Exh 18: Facing pressure from Uber and then Robo-cabs in a few years, we do not expect LYFT to be able to turn a profit for a long time

for AI assistants is still modest. Robo-cabs will be limited to a few markets for the next couple of years. Auxiliary businesses – bike and scooter rentals for LYFT, deliveries for UBER – appear to have legs. The narrative of fast growth and progress toward profitability may well take hold and the stocks may appreciate in response.

Investors are implicitly betting on an optimistic outcome, where ride sharing platforms remain dominant in a vibrant Transportation-as-a-Service market and where they can leverage their market strength and better utilization for drivers to wrest an ever higher commission rate from each trip (Exhibit 14, 15, 16, 17, 18). This narrative also imagines that robo-cabs are still years from viability, and that as they emerge, the operators will be fragmented and lack independent reach to the consumer. In this world, Lyft can turn a profit by 2027 and Uber before that. In this world, both are good investments. However, the scenario also could be quite ugly. Success for Waymo would mean branded robo-cab service in 15-20 markets by 2025, reaching consumers via search, maps and assistant and offering significant discounts and subscription

Exh 19: Summary of Top Autonomous Programs on Technology and Testing

packages (Exhibit 19). GM/Cruise could be in a handful or more markets of its own, perhaps partnered up with Amazon for tight integration with Alexa and discounts for Prime members. Maybe Amazon has already started to tap the Uber/Lyft driver pool for deliveries, or, shudder, has launched its own ride hailing service. Perhaps Ford in partnership with VW and autonomous technology start-up Argo.AI will be ready to come to market, with a raft of options for reaching the consumer, they might drive a hard bargain in negotiations with Lyft or Uber. Imagine commission rates go down rather than up, the robo-cab competitors carve out market share from the ride hailing apps, and Lyft and Uber must double down on investment to play catch-up. In this world, neither Lyft or Uber will ever see profits.

We think this second dark scenario is more likely than the sunny world laid out in Lyft’s S-1. Uber has a better shot at profitability – it starts with much greater scale and international markets will take much longer to see Robo-cab service – but struggles in the US market will weigh heavily. We see the two as trading stocks, with near term upside to expectations but inflated valuations and understated and potentially existential risks (Exhibit 20).

To quote former Uber CEO Travis Kalanick from 2016:

If we weren’t part of the autonomy thing? Then the future passes us by basically, in a very expeditious and efficient way … If we are not tied for first, then the person who is in first, or the entity that’s in first, then rolls out a ride-sharing network that is far cheaper or far higher-quality than Uber’s, then Uber is no longer a thing”

Neither Uber or Lyft are going to be first, or even tied for first.

Exh 20: SSR Summary of Ride-hailing Market