- Baidu's Apollo Go is quickly emerging as a leader in the Chinese robotaxi industry, with operations spanning 10 cities and over 500,000 rides across three quarters.
- The recent release of the sixth-gen RT6 model shows an impressive reduction in production costs that can allow Apollo to quickly scale fleet size.
- Scaling to "tens of thousands" of vehicles across dozens of cities by mid-decade offers tremendous revenue generation potential.
- By 2030, a 100,000 fleet size driving ~150km per day (40% less than current taxi rates) could generate $1.6B in annual revenue.
An unexpected candidate for most investors when it comes to autonomous vehicles and robotaxi operations, Chinese tech firm Baidu ( BIDU ) is quickly solidifying a pole position in the robotaxi revolution in China, with developments far ahead of key rivals Tesla ( TSLA ), GM's ( GM ) Cruise, Alibaba-backed ( BABA ) AutoX, Pony.ai, and others.
Beijing just granted Baidu's Apollo Go and Pony.ai's service the first deployments in the city, with the two set to operate 30 robotaxis across a 60 sqkm area -- this marks another expansion for Baidu's robotaxi service. Baidu also unveiled a new robotaxi with a removable steering wheel that can greatly enhance the unit economics of robotaxi services due to a much cheaper production cost. Apollo Go is far along the commercialization curve, operating in multiple cities with hundreds of vehicles, far ahead of Tesla's yet-to-be-developed robotaxi and preliminary launches from other competitors.
Baidu's Apollo Go has been on a fast track to expansion , recording over 500,000 robotaxi rides from Q3 2021 and Q1 2022, with plans to service 65 cities by 2025. A massive opportunity for growth awaits for Baidu and peers, and although it's not a clear shot to mass commercialization, the long-term opportunity for Apollo Go remains substantial.
A Massive Market Opportunity
Robotaxis are still in the early innings, with deployments at the moment limited in size and scope. Apollo Go has an ambitious plan to scale, growing fivefold from nearly a dozen cities currently to 65 cities by 2025, and 100 cities by 2030. Thus, there's the potential to expand both geographically, to service more and more cities, as well as within cities, with additional vehicles on the road as initial deployments are primarily limited to 30 vehicles at the moment.
The overall robotaxi market in China is expected to show tremendous growth over the course of the next decades as the technology explodes into a much larger scale. BloombergNEF's 2022 EV outlook projected that China "will operate the world's largest robotaxi fleet with about 12 million units by 2040, followed by the US which operates around 7 million autonomous vehicles." IHS Markit projected that the robotaxi "sector is poised to witness robust growth in the coming years," as the "market size of China's self-driving taxi services is expected to surpass 1.3 trillion yuan by 2030, accounting for 60 percent of the country's ride-hailing market by then." IHS added that the "robotaxi market will eventually be dominated by two to three major service providers, with the top providers occupying more than 40 percent of total market share."
Baidu is shaping up to be a key player, and even a significant leader in China's robotaxi market. At the moment, Baidu's core developments with Apollo Go, its current geographic scope and scale, combined with its expansion plan and latest robotaxi unlocking favorable unit economics are positioning it to be one of the, if not the, most dominant players in China. However, all it takes in one slip-up in safety for Baidu to be overtaken, as the deployment of driverless robotaxis hinges critically on safety.
Baidu Unveils Latest Vehicle
Baidu released details of its latest EV to be utilized for robotaxi services, the Apollo RT6. The RT6 is a crossover SUV that features a detachable steering wheel, seating for up to four passengers, and comes outfitted with 38 different sensors -- including 8 lidar units and 12 cameras -- to enhance navigational ability and safety with its Level 4 autonomous driving ability. For a quick description, both Level 4 and Level 5 autonomy do not need a human safety driver; however, Level 4 will operate only under certain conditions and is best suited for localized (geofenced) operations within cities.
The sixth-gen robotaxi for Baidu, the RT6 comes with an impressive unit cost, which can greatly accelerate scaling up of commercial operations and fleet size. According to Baidu, the RT6 only costs ~ ¥250,000 ($37,030 USD) to produce, which is nearly half of the prior generation's ¥480,000 cost. Such a decrease in production cost can allow Baidu to significantly scale fleet size for a fraction of the cost of other services, such as those that retrofit EVs with sensors and autonomous software -- DeepRoute.ai sells Level 4 autonomous software for ~$10,000, and rival WeRide is working to decrease that price to facilitate mass production. Baidu CEO Robin Li said the "massive cost reduction will enable us to deploy tens of thousands of AVs across China."
However, the 250,000 yuan price tag includes the full sensor suite and software kit, increasing the challenge for rivals to compete on price. Tesla, for example, does not even have a robotaxi in development, and is attempting to develop a robotaxi for launch in 2024. Given Tesla's struggles to produce an affordable EV at the $25,000 price point, its refusal to use lidar and the fact that FSD/Autopilot is unlikely to eclipse Level 2 over the next couple of years (Tesla lawyers argued that Autopilot is not meant for city driving in a case in Germany), Tesla's progress to a robotaxi deployment at scale is very far behind Baidu.
A Service At Scale Worth Billions
Baidu's commercialization progress for the Apollo Go service is still in the early innings, but the ridership levels in current operations point to a tremendous revenue generating opportunity at scale as the decade progresses. While it is extremely difficult to predict the exact timeline, scope and scaling of commercialization, projections using initial ridership data and Baidu's internal targets reflects quite positively for long-term potential.
From Q3 2021 to Q1 2022, Apollo Go recorded ~524,000 rides; 115,000 in Q3, 213,000 in Q4, and 196,000 in Q1. With a commercial fleet size of 300 robotaxis, Apollo Go completed about 7 to 9 rides per day per vehicle during Q4 and Q1 -- that's a ride every hour, sometimes longer for a 9-6 operation. Across 10 cities, 7 to 9 rides per day is an extremely small scale. From late 2020, WSJ notes that Apollo Go's 40 vehicle fleet in Beijing saw orders " exceeding 2,000 some days," translating to over 50 orders per vehicle, a substantially higher rate than current average, suggesting significant room to the upside in rides per day as fleet sizes increases.
A commercial fleet size of 300 robotaxis is a drop in the bucket compared to the scale Baidu is targeting with the release of the RT6. Li has said that Baidu aims to deploy "tens of thousands" of the robotaxi; overall, Apollo Go's "fleet is scheduled to scale from 10,000 to 100,000 vehicles at a steady rate, Baidu said, without disclosing when it will reach that size." The production price reduction, bringing the price to one-tenth the industry average, according to Baidu, will be a significant driver of scale.
Reaching a fleet size of 25,000 robotaxis by 2025 across 65 cities, in accordance with Baidu's internal targets, would equal about 400 robotaxis per city. At this scale, ridership and revenue would dramatically increase from current levels.
Apollo Go has not disclosed exact pricing, but testing in 2021 cost a flat ¥30 (~$4.60) per ride. Assuming pricing in a full-fledged commercial operation would be at similar levels, Apollo Go's 524,000 rides in the past three quarters have generated approx. ¥15.7 million ($2.3 million) in revenue.
Quarterly Ride, Revenue Projections
A brief scenario exploration for quarterly ride projections demonstrates how quickly revenues could scale if fleets can scale to Baidu's targeted size:
1) At a 25,000 robotaxi fleet, and ~8 rides per day per robotaxi, Baidu could notch 18 million rides per quarter.
2) At a higher rate of 12 rides per day per robotaxi (still half of Beijing's reported 50 order rate per day), Baidu could notch 27 million rides per quarter.
While this is purely an initial estimate, it reflects the massive growth potential for Apollo Go at a fleet of "tens of thousands."
3) At 100,000 robotaxis across 100 or more cities (a fleet size of approx. 1,000 per city), a ridership level of 12 rides per robotaxi per day would equate to 108 million rides per quarter; however, at such a scale, it's likely that ridership levels would be much higher due to expanded operations and a much larger population within serviceable areas.
4) A ridership rate of 20 rides per day per robotaxi equates to 180 million rides quarterly.
Translating projected quarterly rides at scale is more difficult, given that exact pricing has not been announced. Working off of Apollo's test pricing rate of 30 yuan per ride, Baidu's revenues under the first scenario would equal ~¥540 million (~$80 million). It's hard to see Baidu keeping a flat-rate pricing scheme as rivals and traditional taxis charge per km at lower rates.
Rival Pony.ai has disclosed that pricing of its robotaxi rides cost ¥2.6 per km in off-peak hours, and ¥3 per km during peak hours; this is comparable to a traditional taxi pricing scheme in Beijing. Li has expressed intentions to bring robotaxi pricing to half of that of a taxi, suggesting Apollo Go in the future could charge a fare such a ¥1.6 to ¥2.2 per km.
Should Baidu reach a 100,000 robotaxi scale, by 2030 or 2035 or longer, with ridership levels about equal to the fourth scenario above (20 per day per robotaxi), revenues would project to ¥2.52 billion ($374 million) at an average ¥2 per km pricing scheme. This also assumes an average trip distance of ~7km, what historical data shows for Shanghai; other data also points to an average of 25 trips per vehicle per day for Shanghai taxi fleets, suggesting an average of 20 trips per robotaxi per day is a feasible rate at scale.
Data from Shenzhen shows an average of ~272km of paid rides per day in the city; at an average rate of ¥2 per km, revenues would project to ¥4.9 billion ($725 million) per quarter. Now, obviously not all cities will reach that 272km daily paid km rate, due to different population sizes, serviceable areas, and other factors such as fleet size and robotaxi demand. But the two scenarios with Shanghai and Shenzhen show the raw revenue potential that a robotaxi service can hold in the long-term.
A $1.5B+ Revenue Opportunity?
Data has shown that in a large city such as Shenzhen, taxis can average ~272km of paid distance daily, while in Shanghai, data from the 50,000 taxi fleet points to an average of ~384 km of total distance daily; assuming a 60% paid utilization rate, Shanghai would average ~205km of paid distance daily across its taxi fleet.
Working off a lower bound of ~150km paid distance driven per day (~20 rides with an average distance of ~7.5km) and an average of ¥2 per km, a robotaxi fleet could generate ¥30 million ($4.4 million) per day, or ¥2.7 billion ($400 million) per quarter. Assuming higher paid km driver per day only increases the revenue potential for the service.
Annually, that fleet would generate approximately ¥10.8 billion in revenue, or $1.6 billion. That means a 100,000 robotaxi fleet could generate the equivalent of ~9% of Baidu's current ¥124.5 billion ($19.6 billion) in total revenues. Having a single unit within Baidu's field of autonomous driving generating revenues in excess of ¥5 billion to ¥10 billion per year can supercharge Baidu's valuation, due to the high-growth nature as well as the massive revenue contribution potential from Apollo Go. The service also is likely to be highly profitable -- Baidu is pointing to an operating lifecycle of over 5 years , meaning that a 100,000 robotaxi fleet at its current $37,000 production cost per unit (likely to decrease with new manufacturing and technological advances) would cost just ¥3.7 billion (an average of ¥740 million per year over 5 years).
Tesla Falling Behind
Tesla is likely struggle to obtain a lead in China in autonomy, namely in robotaxis even as Musk plans to bring a robotaxi to market by 2024 for a handful of factors:
A lack of robotaxi experience - Tesla has no experience in robotaxi development, and there are significant difficulties in applying fully autonomous robotaxis at scale with more challenging driving environments and stricter regulation.
Regulatory hindrances - China is quite strict in terms of ADAS development and deployment, and Tesla will not simply be able to unveil a fully autonomous vehicle and deploy at mass scale without running pilot tests across cities first, which can take years to validate.
Favorable policies towards domestic players - domestic players have been testing robotaxis for years, and China's quest to lead in autonomous vehicles will see it promote the domestic cohort ahead of Tesla, such as Apollo Go and Pony.ai.
Competition - again, multiple domestic players have been running robotaxi tests for years and have moved on to full-scale deployment, such as Apollo Go; with Tesla years behind, Apollo Go holds a substantial first mover advantage against Tesla.
Difficulty scaling - even if Tesla does begin producing a robotaxi model by 2024, and assuming it can achieve a safe, validated fully autonomous software, it's likely to require years of road pilots and testing before potentially moving on to mass deployment, which may take until 2030 to get initial small scale fleets on the road across China.
Tesla also faces difficulties in the US and Germany, as Tesla is facing enhanced regulatory scrutiny around Autopilot/FSD due to crashes. A recent case in Munich saw Tesla ordered to refund the full purchase price of a Model X due to Autopilot's inability to work properly on city streets, which saw Tesla's lawyers arguing the function was not meant to do so -- this alone points to a challenging path to full autonomy (combined with Tesla's L2 status) as Tesla does not support city driving, the environment in which a robotaxi is built for.
Outlook
Overall, Baidu has a tremendous opportunity in robotaxis ahead for the long-run if it can successfully reach internal targets, such as operations in 100+ cities and achieving a 100,000 robotaxi fleet size. In such a scenario, annual revenues from a robotaxi fleet could easily top $1 billion per year given modest paid driving rates, and if robotaxis can match traditional taxi paid rates, Baidu could see a 100,000 robotaxi fleet generating $2 billion or more in revenue per year.
Profitability of the robotaxi fleet also is likely to be quite high as production costs halved with the recent RT6 release, with a five-year-plus lifespan allowing Baidu to quickly recapture the bulk of these costs early on in the fleet lifespan.
Competition and safety are two key threats to Apollo Go's emerging status as a robotaxi pioneer, with peers such as AutoX, Pony.ai, and others quickly pushing along with driverless tests and commercial operations. Safety poses an industry-wide risk, as incidents and accidents (of which Apollo Go has none) could threaten a regulatory slowdown and hinder expansion of commercialization efforts.
However, the future is bright for Apollo Go in the long run, assuming that it can successfully scale to a 100,000 fleet size across 100 cities or higher, and command similar driving metrics to traditional taxis, which it aims to displace.
For further details see:
Baidu: Beating Tesla In Potential $200 Billion Robotaxi Opportunity