Thanks to all of our members for supporting us! We hope you enjoy our Didi write-up! We dive deep into their accounting, which may get a bit technical for some, but we promise our takeaways will be clear at the end!
In 2012, Cheng Wei, an 8-year employee of Alibaba struck out on his own to emulate a UK start-up, Hailo, that enabled consumers to order one of London’s notorious black cabs. Three years after Uber was founded in San Francisco, there were already many ride-hailing copy-cats in China, but Beijing-born Didi differentiated itself by being the first to attack the 2mn large yellow taxi market. Anchored by RMB 800k from from his superior at Alibaba, along with several work colleagues, Cheng launched Didi’s services in Shenzhen, giving people the ability to call a taxi cab with the tap of a smartphone app. After facing many local regulatory hurdles, they hit their first stride of luck when a snowstorm hit Beijing, making citizens desperate to get home. For the first time ever, they hit 1,000 orders, which caught the attention of a local VC firm that helped supply them with $2mn of capital to continue their experiment. With their newfound traction, they added a Uber-esque ride-hailing service to better compete against the flurry of burgeoning ride-hailing start-ups. Despite counting Cheng as an alumni, Alibaba invested in one of Didi’s largest competitors, Kuaidi Dache, in 2013. This resulted in Cheng taking sides with Tencent, who helped fund them.
The growing pains were palpable, with a series of outages culminating in the infamous “Seven Days, Seven Nights” where Didi engineers worked nonstop for a week to stabilize the platform, and only succeeded thanks to 50 engineers, 1,000 servers, and more office space, all courtesy of Tencent. Around this time, Kalanick started to spin-up their China operations and met with Cheng Wei about a potential partnership. When Kalanick demanded a 40% stake in Didi, Cheng curtly responded, “Why would I take that?”. With capital gushing into the sector, “subsidy wars” broke out with companies lavishly subsidizing the costs of rides in an attempt to gain market share. By 2015, Didi had inched into the lead against domestic rivals. However, Uber was quickly gaining share, laying claim to ~1/3rd of the private ride-hailing market. Despite being the newcomer, Uber was seen as the ride-hailing goliath benefiting from more reliable tech, a larger engineering team, and far more access to capital and a valuation 10x that of Didi’s. To stem the bleeding of cash flow spent on incentives to fickle riders and focus their attention on killing Uber, a deal was brokered between Didi and Kuaidi Dache to create the largest China-native ride-hailing company.
While in many industries the blithe focus on market-share is uneconomic, in ride-hailing a disproportionate amount of profits will be earned by the most scaled player because of the density effects inherent in any logistics operation: the focus on gaining share was paramount to ever achieving profitability. Cheng Wei was not prepared to be the first China-native company to lose a market to Silicon Valley, and pushing Uber out of China entirely would be a boost to their ultimate unit economics. Three months after merging with Kuaidi Dache, the merged “Didi Chuxing” pledged to give away RMB 1bn in subsidies. They then invested $100mn in Lyft in an attempt to distract Uber and heat things up in their home market. However, Uber refused to back down, and even matched the RMB 1bn in subsidies. But pressure was mounting internally at Uber to stem the losses, which were estimated at ~$2bn by that point, and focus on their other ample growth opportunities. Didi kept raising funds, $1bn from Apple and $3.5bn from Saudi Arabia Public Investment fund, signaling they have no intention of letting up the pressure.
In August of 2016, Wei and Kalanick struck a deal. Didi would take over Uber China’s operations, and Uber would get a 17.7% economic interest in the combined company. With a $35bn valuation and having thoroughly conquered all notable competitors (they had ~90% ride-hailing market share), Didi started focusing on adjacent business opportunities. Over the next few years, they created subsidiaries focused on new initiatives like electric vehicles (EVs), autonomous vehicles (AVs), shared-bikes, intra-city freight and community group buying.
On June 30th, 2021, Didi IPOed on the NYSE, raising $4.4bn in primary capital, the largest amount since Alibaba IPOed in 2014. However, the celebration was short-lived with the Cybersecurity Review Office launching a data-security probe into Didi on July 2nd. They requested Didi’s app be taken down from all app stores, which throttled their growth – but refrained from completely halting the service so many Chinese had come to rely on. In just a few weeks, the stock tanked ~50% from its post-IPO pop.
As Didi waits in regulatory purgatory, we go through their business in-depth, which is unlikely to change regardless of the outcome. That said, we do explore some potential regulatory remedies at the end of this report.
Didi has a fairly straight forward business model, but very complicated accounting! While we are usually hesitant to make direct comparisons to Western counterparts, it is pretty accurate to think of Didi as China’s Uber – with the caveat that around ~20% of rides are fulfilled by yellow taxis. While we all are likely familiar and used to the benefits of ride share today, originally it was a step function improvement versus the traditional taxi experience. Not only would a ride come directly to your location with a few taps on your phone instead of a lengthy call to a cab company, the ride sharer acted as an intermediary to handle payments, allowing a rider to simply walk out when the trip was over. Clearly, this technology enabled an exponentially better experience and resulted in more user utilization compared to traditional taxi.
This is the basic flow calling a ride in the Didi app.
Photo 1 is the home screen. You can choose what kind of service to call (Express, Premier, Taxi, etc) and see your location (blue dot), suggested pickup spot (green marker) with estimated time, locations (orange marker). Within Didi Express, you can see the different types of cars offered, along with different prices for each option. Right above the “call car” button, you can choose whether it’s a personal trip or a business trip.
In Photo 2, you’re able to see the price of the ride, any discounts applied, and choose what payment method you want to use (ie Alipay, WeChat Pay, or something else).
In Photo 3, once you’ve reached your destination, you can rate the driver and provide feedback.
This two-sided marketplace has natural flywheels. The more drivers there are, the shorter the consumer waits. The shorter the wait, the more people are drawn to calling a ride. The more ride demand there is, the more drivers take up driving. The more drivers there are, the more comfortable riders feel relying on ride-hailing, trusting that they won’t be stranded by a lack of rides. All of this is helped by dynamic pricing which balances rider demand and driver supply by raising prices to draw in more drivers when demand surges or there is a driver shortage. Importantly, ride-hailing only enjoys hyper-local network effects with intra-city density being instrumental to providing a good service for all parties. Network density does not extend advantages to a larger proximity, making barriers to entry relatively low in theory as a newcomer can saturate a market with subsidies to spin up a network one city at a time. Riders are notorious for having low loyalty to a ride hailing service, but they also have no incentive to adopt an unheard of one and risk unreliability without heavy subsidies. A newcomer needs critical share to ever reach profitability and a competitor could simply match the subsidies and burn them out of cash, locking the newcomer out of the market. So while the barriers to entry are low, the cost to compete quickly becomes very high. It is this game theory that usually keeps upstarts from trying to disrupt the order. A national brand does have some advantages as consumers value knowing that a ride-hailing service could serve them for long-distance trips if needed, and, at least as far as initial adoption is concerned, having a brand that you could trust to be able to pick you up almost anywhere is a powerful way to win over consumers from private car ownership, rather than having to rely on a hodgepodge of different services that may not be as reliable or could go under. The original thesis of ride sharing was that cars are under-utilized and sit stagnant ~90% of the time, and that sharing ownership would not only increase utilization but also reduce the cost of a personal ride. In their S-1, Didi emphasizes the point that ride-hailing is cheaper on a per km basis than owning a car (although they do not disclose how many miles they assume the private car owner drives annually). This is in contrast to the US, where (according to Didi) car ownership is less than half the cost of ride sharing.
However, the core thesis hasn’t quite played out as they expected, with many people still opting to buy a car rather than solely rely on ride hailing. Thankfully for Didi, many new use cases for ride-hailing quickly proliferated that were not cannibalistic to traditional car usage, such as after a night of drinking, for an elderly parent who can’t drive, or when going somewhere crowded without easy parking. It appears that a lot of usage has been more supplementary than substitutive. However, Didi maintains that their core TAM is all of mobility ($6.7tn TAM in 2020) and that they expect the shift from traditional mobility, specifically, private cars and public transportation, to shared mobility to further accelerate. Given GTV has essentially stayed the same since 2018 while private car ownership and public transit usage have increased, it is hard to support their thesis. But of course, the world could be very different in 10 years. We were not able to get a data series on driver’s licenses issued by age in China, but if the data is anything similar to what we have seen in the US (those under 18 who have driver’s licenses dropped from 80% of teens to 60% since the 80’s), then there is reason to believe that Gen Z will be less likely to drive personal cars. However, translating lower driver’s licenses rates to higher ride hail usage takes a leap of faith as it could also be explained by many other factors like smartphones making teens more comfortable taking other means of transit. It is also worth noting that a far higher portion of Chinese drivers ( >70%) are under 35, likely owing to the relatively new availability of automobiles in China. The fact that cars are often thought of as status symbols will also keep demand strong. Simply said, it is totally possible ride-hailing becomes a main means of transit, but that would require that the future look very different from the present, which an investor may only prudently assume should they be properly compensated or have outsized confidence in what form the future will take.
Didi organizes their operations into three segments: 1) China Mobility, 2) International, 3) Other Initiatives. Putting aside their growth initiatives, they principally make money from ride sharing, which accounts for >95% of their revenues. Across all ride sharing options, the user pays Didi the cost of the ride, based on distance and time, plus a fee for facilitating the transaction and Didi remits payment to the driver. There are four different categories of Ride sharing in China Mobility, touched on below.
Ride Hailing: Connects a user with a driver who uses their personal car for a ride. This is the most popular option at ~78% of GTV (gross transaction value) and essentially is the same as calling an Uber with many different ride formats (shown below, from their S-1).
Taxi Hailing: Connects a user with an official taxicab for a ride.
Chauffeur: Didi dispatches a driver who arrives on a foldable electric bike to meet the rider at their car. The driver puts their bike in the trunk and drives the rider to their location before taking off on their bike.
Hitch: Drivers input their destination and can pick up “hitchhikers” along their route to their destination.
Just looking at revenue by segment, we already start to run into accounting issues. We will dive into the minutia of different accounting treatments later on, but be aware that for most of China Mobility revenue, Didi is considered the principle and thus records revenue on a gross basis before payments to driver earnings. In their international operations, they consider themselves an agent and thus report revenues net of driver payments. There are more accounting ramifications that are dependent on this principal vs agent distinction, which we will pick up on later in this report. As shown below, China Mobility Revenue is 94% of revenue versus just 2% for international. “Platform sales” is a metric of Didi’s creation that tries to adjust for the different revenue recognition methods (more on this later) and on that basis (and not including Other Initiatives), international revenues increases to 7% vs 93% for China Mobility. The takeaway is that clearly China is their most important market, while International can extend their growth runway.
In total, Didi has 493mn active users, 15mn drivers, and 41mn daily transactions across 16 countries. For reference, Uber has ~100mn active users and completes an estimated ~15mn rides per day. So Didi is the global ridesharing leader by activity. Its global leadership largely stems from its ~90% Chinese market share though, so we will first dive into their China Mobility operations.
As mentioned, China Mobility includes Didi’s ride hailing business (79% of GTV), as well as Taxi, Chauffer and Hitch. We show some segment level metrics below. What is striking is how GTV hasn’t grown much since 2018, the earliest data available. Transactions are also at a similar level even after taking into account Covid abnormalities. While they were still having select lockdowns, China was largely back to normal in 1Q21, where Didi completed 2.28bn rides. If you annualize this and compare it to their total 2018 transaction volume, you see that transactions have grown at just a 2% CAGR. We will reserve commenting on other metrics until we dive into the minutiae of Didi’s accounting, which distort any cursory takeaways.
To build a better sense of how the accounting flows, see the exhibit below, which shows how Didi treats different items from GTV to platform sales. Platform Sales is a metric they use so their “revenues” can be compared to their international segment and global peers. There are two key things to note: 1) consumer incentives are added back from revenues to platform sales, so we cannot back into how much driver earnings are historically (we do estimate this in our model though), 2) this accounting walk is only applicable to Didi’s China ride hailing business where they act as the principal and not their China Taxi, Chauffer or Hitch businesses. In the latter, Didi act as an agent and thus revenues are reported net, after the driver earnings are deducted. This means that within the China Mobility Segment, Didi act as both a principal and an agent, and thus their metrics are obfuscated from pooling two different accounting treatments in one segment. We don’t have any evidence to suggest there has been a material mix shift away from Taxi (reported on net basis) to ride hail (reported on gross basis) in the past 3 years we have data for, but it is worth noting that if ride hailing is gaining share within their total rides vs Taxi, then Didi’s revenues would look like they are growing. But this is simply a result of the accounting treatment putting ride hail driver earnings in cost of revenues instead of as a contra revenue item. (Accounting note: contra revenue items are just revenues recorded as “negatives” — or in accounting jargon, as a debit instead of credit — reducing total revenues when summed up.) Similarly, if Taxi, Hitch, and Chauffer were growing rapidly, their impact on revenues would be more muted. Though, our opinion on that is if that was the case they would call it out, as they have every incentive to put their business in the best light possible.
The other thing to be aware of is how the treatment of consumer incentives changes. It will be clearer when we show the treatment side by side on a unit basis below, but note that consumer incentives are not included in revenues reported on a gross basis and accounted for as a reduction to GTV.
They do provide the helpful waterfall below, but since we cannot look at the business on a historical basis with any granularity, it is hard to gauge how the core business is trending. Ideally, we would have wanted to see financials with 1) separate segmentation within China for ride hailing and the three other services reported on a net basis, 2) driver earnings itemized, and 3) consumer incentives broken out.
To help clarify how the different accounting treatments affect a single transaction, see the unit illustration below with Didi-provided figures. As noted, the distinction in accounting treatments is predicated on whether Didi is considered a principal or an agent in the transaction. As alluded to above, the distortions of recording revenues on a gross basis are: 1) revenues are much higher and 2) cost of revenues are much higher. In this regard, it is similar to the 1P vs 3P distinction we call out in our ecommerce pieces (SE, BABA).
You may have noticed how consumer incentives and driver incentives are sometimes deducted from revenues and GTV, and other times expensed in cost of revenues and S&M, respectively. This creates a problem when trying to deduce what their ultimate profitability will be. With many growth companies, S&M is inflated as they spend to take advantage of value creating investment opportunities that do not materialize on the P&L until years later. To illustrate this, consider an investor looking at a SaaS company. A SaaS company spends prodigiously on marketing, knowing that every customer they get is likely to stay for a long time (very low churn and long customer life) and spend more with the platform overtime (spend by cohort increases). Before investors were comfortable with SaaS companies, they complained about how they were always unprofitable and could grow losses on increasing revenues. However, over time, many realized that SaaS companies “invest” through their P&L instead of their cash flow statement. So stripping out the growth investment in S&M (which can by far be the largest expense line item), you would see a much more profitable company. This is fair treatment for a SaaS company because their customers seldom churn, so there is almost no “maintenance” S&M that is needed to incentivize them to stay. Contrast that with many consumer companies that sell products in one-off transactions, where marketing can seldom be fully turned off because you have to “re-acquire” the same customer over and over again. This is to say, the troubles we run into with Didi are twofold. First, “growth expenses” are sprinkled throughout every line item including revenues (via contra revenue entries) and cost of revenues so we cannot simply adjust out the growth expenses. Second, it isn’t actually clear what is a “growth” expense that is driving new user growth or what is a “maintenance” expense that is needed to get a user to return. If you have used ride-hailing or food delivery services, then you likely recall getting a push notifications or emails with a coupon incentivizing you to come back to order again—it is not clear if this marketing spend is to habituate you to the service so you become a customer for life or is simply a reoccurring customer acquisition cost that can never be fully eliminated. These are important distinctions to make as we think about the quality of the businesses and what the mature core business’s profitability could look like. We will pick up on this again in the Revenue and NOPAT build, but the punch line is that a conservative investor would not be able to give Didi much credit for S&M leverage and would assume that usage and frequency would drop with lower S&M spend.
In Photo 1, you can see a reminder for 2 coupons the rider has in the orange card, prominently displayed in the middle of the screen.
In Photo 2, you can see a RMB 118 coupon as displayed within WeChat.
In Photo 3, you can see the terms and conditions of the said WeChat coupon.
The exhibit below shows that take-rates (defined as platform sales over GTV) have increased 10 points from 2018 to 1Q21. However, platform sales do not commensurately increase with revenues: annualized 1Q21 revenues are 17% higher than 2018 despite take-rates doubling and platform sales increasing ~125%. The biggest accounting adjustment is that platform sales adds back consumer incentives, while revenues are reported net of it. As we explained above, we do not believe that Didi has the ability to cut these consumer incentives out and keep usage at the same clip, so we do not put much weight to their platform sales growth. We originally suspected this discrepancy was driven by “raising” prices (which increases GTV) while simultaneously increasing incentives (which washes out the price increase). However, GTV per transaction (or AOV), has not increased enough to support that this is a big factor. Decreasing driver incentives and earnings could be another factor, but it is not entirely clear.
In 1Q21, Didi’s take-rate hovered ~19%, which is lower than Uber’s adjusted 22% and in-line with Gojeks ~20%, but we put little faith in the usefulness of these comparisons. Theoretically, a take-rate is supposed to be the portion of value you capture for facilitating an economic transaction, but it loses its power when the economic value of the transaction is unclear. In Didi’s blog they claim that their take-rate was an average of 19% in mid-2019 (15-25% range), which is defined as Didi’s portion of “actual fare” paid by passengers, but that is higher than our 2019/2020 calculated take-rate. Accounting ambiguities aside, we believe that their take-rates are fairly mature and we do not believe there is an opportunity to raise take-rates. In fact, it seems take-rates are more likely to go down than up with pressure from drivers complaining that their share of earnings is too small (whose complaints have more substance in a less capitalistic market, especially, while under sever government scrutiny). Global peers like Gojek are actually cutting their take-rate in half–ostensibly because of the pandemic–but that could also just be cover as they vie for more market share (they currently have ~40% vs Grab’s ~60%). Didi doesn’t have a strong second competitor, but because of how the average consumer calls a ride (explained at length below), there is more of an opening for upstarts. Additionally, as we will go through below, our consumer survey work leads us to believe that they do not have much of an ability to raise ride prices. In short, we believe growth will have to largely come from new users and ride frequency per user increasing.
In 2019, Didi was able to achieve adjusted EBITA profitability in their China segment. But when you take into account stock comp, they likely didn’t reach profitability until late 2020 or last quarter (it is not clear exactly how much stock comp to attribute to the segment). Interestingly, Adj. EBITA increased from RMB 3.9bn in 2020 to them doing an annualized RMB 14.4bn on last quarter’s numbers. The moderate increase in revenues does not explain this level of improvement and we do not have segment level cost items to figure out the main driver, but it seems likely they cut costs when going through 2020 and the quick growth in rides (GTV +113% y/y in 1Q21) allowed operating leverage to show through. It is less clear if there are costs that will be added back or if this is a permanent improvement in their profitability.
China Mobility: Consumers
Our survey work of 200+ consumers uncovered a few interesting factors for Didi that have negative ramifications for their competitive position, transition to AV, and overall general ride share TAM. Big picture though, Chinese riders use Didi an average of 24 times a year, spending an average of RMB ~24 per ride. This means their average consumer spends RMB ~575 with them annually. At that level of frequency, it implies most people use it as a supplementary transit option and do not rely on it is their primary means of transportation.
This is backed by our survey which showed only ~11% of consumers used ride hailing once or more a day, with once a week being the most common answer.
Not surprisingly, over 80% of our respondents view ride hailing as very substitutable with other means of transportations, opting to always take the cheapest or fastest options. Only ~20% of respondents believe that Didi is the cheapest mode of transit with ~50% believing it is the most expensive and 30% indifferent. Perhaps oversimplifying, this means that if most consumers want the cheapest and fastest means of transportation, and Didi isn’t the cheapest, then they have to win on being the quickest transit option. This is precisely the case as ~75% of respondents believe that Didi is their best transit option if they need to get somewhere quickly. A few other factors that ride hail wins on is the ability to go to places public transit doesn’t service, at times public transit doesn’t operate, and higher cleanliness. Nevertheless, it appears that speed to destination and price of transit are the crucial factors (not a particularly riveting conclusion). Winning the incremental trip for Didi will require them either to deliver rides quicker (~35% of consumer would order more if the wait was shorter) or drop prices (virtually everyone said they would use it more if it was cheaper).
Theoretically, Didi could do both by lowering prices to create more demand, which in turn brings in more drivers whose greater density decreases consumer wait times and increases driver earnings with more trips per hour. However, there are natural equilibria between all of these variables that Didi has likely already reached in their mature markets. Anyhow, their tiny profit per ride (detailed in the unit economics section) would make them reticent to decrease prices. The only natural way for prices to drop would be for demand to organically increase which would initially push up prices but then also draw in more drivers from the higher earnings. The larger number of drivers would mean a higher likelihood one is close to a consumer when they request a ride. The ability to waste less time with an empty car would mean higher driver earnings per hour, some of which would be competed away from more drivers joining the network, who are willing to drive at a moderately lower rate because hourly earnings are slightly higher from the increase in utilization. Didi’s dynamic pricing algorithm would respond to the influx of increased driver supply with lower sticker ride prices. This is what has been happening over the past decade in ride hail, but at a much faster rate with many subsidies thrown at both sides of the network to accelerate the flywheel. The question is whether or not it can continue naturally, which we believe is quite likely with secular demand from ride hail increasing overtime, but it will be a slow grind over many years. In short, we do not believe that price decreases or faster speed to destination are factors we can count own to spawn more demand for ride hail in the interim.
We will pick back up on demand and ride frequency assumptions in the China revenue build and will first go through some other consumer behavior habits our surveys unearthed. The results below show that most users (~70%) have used a 3rd party app that aggregates different ride hail services to request a ride and over half use this as their most common means of ride hailing. A 3rd party aggregation app is an app like Gaode (also known as AMAP – read our BABA piece for more context) that allows a user to order a rideshare in the app. The app aggregates different ride hailing services so when the consumer orders from the app, they are essentially accessing multiple different ride hailing services. This is good for the consumer as it can help them get the cheapest/quickest car. It is bad for Didi because a 3rd party is supplanting their otherwise direct consumer relationship. This effectively limits the benefit of their two-sided network on the consumer side. While many of these rides are likely routed back to Didi drivers, having consumer access through a 3rd party means the barriers are lower for a competitor to gain access to a large pool of consumers with no advertising necessary—they just simply need the 3rd party app to connect into their driver network. While ride hailing competition may not be Didi’s biggest concern given their market share, Meituan, T3, and others have recently become more aggressive in light of their restrictions from the Chinese government, the bigger issue is the implications for an AV transition (we touch on in the AV section). In fact, Meituan relaunched their ride-hailing services ~1 week after the government announced regulatory actions. (T3 is a ride-hailing entity formed in 2019 3 state-owned carmakers with investments from Alibaba and Tencent.
As we will show in our China revenue build, more usage is an important factor to rationalize their valuation, but as we have suggested, Didi has very limited ability to entice users to ride hail more. In order to see how many transactions are needed to support Didi’s valuation we first need to go through the economics of a single transaction.
Unit Economics and Growth Vectors.
Below, we estimate Didi’s profit per ride. Ideally, we would want to get to a contribution profit per ride figure, but in theme with our prior comments, no disclosure of itemized segment expenses and other accounting issues mentioned prevent us from being able to. Contribution margin is helpful because it would allow us to see the profit on an incremental ride before accounting for fixed costs, which could help give us a good idea of operating leverage in their model. However, while they theoretically should have high operating leverage as the cost to serve one more ride is minimal, we do not actually see much leverage come through. The chart below shows expenses as a % of revenues, and every item, has increased from 2018. Covid distortions, international growth initiatives, and other new project investments have elevated expenses, so we are not suggesting there is no operating leverage in their business model, just that the numbers do not reflect it. It’s also worth mentioning the relatively stagnant number of annual transactions also make it hard to see operating leverage come through.
Even though there should be operating leverage inherent in their business model, it is our opinion that it is relatively muted compared to most tech companies due to the higher headcount needed to support drivers and keep consumers safe (today, about half of employees sit in the operations & support and the G&A departments). It is for these reasons we do not try to pencil out the unit economics with a more favorable cost structure and only look at actual figures on a 2020 and LTM basis.
The unit economic build below is for ride hailing and we estimate ride hail transactions by assuming the percent of ride hail transactions is the same as ride hail is a percentage of GTV.
The unit economics show that for every ride that costs a consumer RMB ~21 (after incentives), Didi ends up keeping RMB 2.4 in revenues which is keeping ~11% of the transaction (lower than the mentioned take-rates earlier because it doesn’t include consumer incentives). After accounting for RMB 1.6 in LTM costs, that leaves Didi with RMB 0.8 per ride in profit. This is a razor thin profit margin of just 3.5% (on gross revenue) or 3.3% of GTV. With 7bn rides, that equates to RMB 5.6bn in pretax profits. Layering in another 2bn of Taxi transactions with likely similar economics and taxing at 15%, we get to RMB 6.1bn (or ~$1bn) in total for China ride hailing. We will circle back on valuation, but in our opinion, current profits from their core China operation cannot support their valuation and an investor will need to believe that usage will grow, international will be a material source of value, or their EV/AV initiatives will make their unit economics more profitable.
There are two ways you create a successful business: 1) high frequency and low margins or 2) low frequency and high margins. Didi clearly is on former path. The benefit of a high frequency, low margin businesses is that they tend to be very hard to disrupt and most competitors cannot operate at a profit until they reach gargantuan scale. The negative though, is that it takes a lot of sales velocity to increase profits because of the thin margins. With ~3% profit margins today, an investor may believe that margins can improve. Two plausible sources of margin improvement are electric vehicles with lower operating costs to the driver allowing Didi to keep a piece of the cost savings through a lower driver payout or AV, which could replace driver payments with smaller depreciation or financing expenses. Absent of improved margins, which we do not think an investor can conservatively count on, we need to see higher usage in the future. On one hand, an average of ~24 rides annually does seem low and has plenty of room to grow, however it also speaks to how the cost of the service and other quicker transit options limit use cases. Should the TAM open up more with higher incomes and more user penetration, Didi is clearly in position to gain the most of any ride share player, but as mentioned previously, it seems they have limited ability to drive ride hail usage in an organic way.
In our China revenue build below, we show the assumptions needed to support Didi’s total valuation with just their core China business. We assume a $45bn valuation and apply 15-25x earnings multiples to back into how many trips Didi need to complete annually to support their valuation at a conservative range of valuation. On the right-hand side is the growth rates required to reach that level of rides on a 5 and 10 year basis.
We show that they need at least ~100% more trips to rationalize their valuation with just their core business, which can come from a mix of increasing ride frequency per user or adding new users. This translates to a required 7-15% y/y transaction growth rate, which is much higher than the 3% compounded growth rate they have grown transactions at from 2019 to 1Q21 annualized. Given the anemic growth in the core business though, a lower 15x multiple is more justified in the out years, which show that rides need to increase to 29bn, or triple, requiring a marked acceleration in trip growth.
It seems unlikely that future growth will be driven much by user growth. We consider Didi’s core market to be densely populated, urban areas like the top 400 Chinese cities, which have a collective population of ~370mn (the 400th smallest city has a population of 110k). We believe this market is already highly saturated with virtually every user who is likely to rideshare already on Didi. While an estimated 65% of China, or ~900mn people, live in urban areas, lower cost options and a worse value prop in less dense areas will greatly limit that TAM. Average incomes fall outside of the largest cities and so they are more cost sensitive. For reference, bus tickets are <1/10th the cost of a ride hail and the cost of car ownership is much less without the restrictive permits common in Tier 1 and 2 cities. In their S1 they quote CIC as saying Tier 1 and Tier 2 is 24% penetrated with shared mobility, however given that we know they have 377mn users, that is a 27% penetration rate across all of China. So the CIC data is either flat-out wrong or distorted somehow by limiting who they include in their sample. We think that their TAM is already saturated in larger cities and new user growth will have to come from the tail-end of less populous cities, but for the cost reasons we noted, that seems unlikely to be material. Management today seems more focused on growing outside of China than they are in penetrating more of China, which is indicative of a maturing domestic market. It is also worth mentioning that they do not disclose China user growth, which is rather odd for a consumer internet company (and fairly telling that is not strong). As a side note, they do provide historical global platform user figures, but it seems to include users across other services. Globally, they had 493mn and 429mn users for 1Q21 and 4Q19, and noted 60mn and 23mn, of those users come from outside of China, respectively. This implies 433mn and 406mn users within China versus a reported 377mn China mobility users. If we cannot count on material user growth to support transaction growth then ride frequency per user will have to make up the difference. At an average of ~24 rides annually or once every two weeks, there is clearly room for this to improve but why it would change all of a sudden is unclear.
Hence the shift to emphasizing carpooling, which can solve the frequency issue. The benefit of carpooling to the user is fairly straightforward: share the cost amongst more people. If Didi can increase carpooling adoption, then users benefit from a greatly reduced price, which can spawn more demand and increase Didi’s average ride frequency. This could be a critical swing factor in getting total transactions to grow, but there is a chicken and egg problem. In order for carpooling to work efficiently, there needs to be a large number of people who are requesting rides into their carpooling network to build order density that allows quick matching of riders to drivers without the driver having to go out of their way. With weak local carpooling demand, the driver will have to spend more time to pickup / drop off other riders, which increases the time of the trip. Recall that trip speed was one of the most important factors when choosing ride hailing over the alternatives, so this really debases the product value prop for many users. We can think of carpooling as potentially opening up new use cases, but they will have to grow usage a lot to get batching more effective for it to be more of a consideration for daily use versus public transit. In short, it is possible carpooling becomes a bigger means of transit and more private car ownership is replaced by ride hailing overtime, but these will be slow secular trends that do not materially change Didi’s usage in the interim and longer-term AV can change everything. We will pick back up on AV later and our China revenue build assumptions later.
Didi’s international businesses are a mixture of investment stakes and a several operations they set up themselves. Today they have 60mn users outside of China which is up 40mn from the end of 2019.
Didi’s international strategy is a mix of buying stakes in or outright acquiring different local foreign operations or launching in territories under their own brand. In 2018, Didi started their international expansion through purchasing the ride hail start-up “99” in Brazil for ~$1bn. That same year, they announced a JV in Japan with Softbank, then launched in Mexico and Australia under their own name. The following year, they expanded Latin America operations with Chile, Colombia, and Costa Rica. In 2020, they launched in Russia, New Zealand, and even more LATAM countries: Peru, Panama, Dominican Republic, and Argentina. In 2021, they launched in South Africa. Unlike in China, the food delivery market was immature and Didi saw an opportunity to launch food delivery in Mexico, Brazil, Costa Rica, Columbia and Japan. Over that period, GTV grew from RMB 7.9bn in 2018 to LTM RMB of 26.2bn, a >40% revenue CAGR while transactions went from 283mn to 1.4bn. As shown below, Didi’s international operations are not EBITA profitable.
In areas where ride hailing was already more established Didi invested with existing players including Grab in Indonesia, Ola in India, Lyft in the US, Taxify in Europe, Careem in Dubai, and even Chinese food delivery provider, Ele.me. Since then they have sold most of these stakes, but they still have their Grab and Lyft positions as of the S1 publication (although their ownership stakes in these are not disclosed).
The international segment is hard to value for similar reasons as the domestic business is, but also for the fact that there are multiple different markets and services with different economics that are being combined. As seen above, AOVs dropped from an average of RMB 28 to RMB 19, indicating new markets have very different economics than prior ones. Their market share in each territory varies drastically which will affect their ultimate unit economics. For many of these markets they are still heavily subsidizing both the drivers and consumers, which could go on for an indeterminant amount of time.
In the LATAM markets, Uber is the only big competitor and the clear market leader after other players like Cabify shrunk their operations and retreated from operating in some countries. Didi has claimed a 50% market share in LATAM, but how they are defining market share is unclear and this figure seems misleading. It is worth noting that in their breakout where employees are domiciled, they only itemize China (14,654 employees), and Mexico and Brazil (with 428 and 333 employees, respectively), which is an indicator of their largest markets. While there is limited data available, we believe they are closer to ~30% market share in big countries like Mexico and Brazil, which is still a solid foothold (similar to Lyft in the US). The TAM is arguably still growing with ride-hail penetration in an earlier stage than China, estimated somewhere around 5-10% versus ~25% for China (on total population). However, purchasing power parity differences, less city density, and higher usage of motorbikes are factors that likely limit their ultimate penetration. Market share is unlikely to change much though as Uber has matched all subsides Didi has offered, and Didi seems to be unwilling to go to war with Uber the same way they did in China half a decade ago.
LATAM (combined) is the biggest international growth opportunity for Didi with the most traction, especially after Covid hampered operations elsewhere, like in Japan where they shut down operations in half of their prefectures. Today this is a money losing segment with unclear unit economics and it is tough to size up how many consumers ultimately use their services internationally and at what frequency, but we take a stab at it in the revenue build below. More important than whether the assumptions are “correct”, is your judgement of their plausibility and the return you would receive if the world played out that way. Members Plus of course can change the assumptions as they wish in the downloadable file.
Didi has launched a flurry of other initiatives outside their core ride share businesses including bike sharing, intra-freight, community group buying (CGB), electric vehicles, autonomous vehicles, and auto solutions. We will start with Auto Solutions as it directly interplays with their ride share business.
Auto Solutions. The Auto Solutions operation consists of a partner network of charging and fueling stations, repair shops, and car lease services. While most of these services are provided through 3rd parties, where they secure bulk discounts for directing drivers to them, in some instances they do own the services (when Auto Solutions was first launched, they acquired Hiservice and their 28 physical locations). Oddly, enough they record all of these services within “Other Initiatives” with one exception: their leasing business that is carried out by 3rd parties, which is included in China Mobility. They only mention this in a passing footnote so it is not clear how large this business is, but it should be noted that it could be artificially distorting China Mobility financials, as a 3rd party leasing business could simply be them receiving a fee (high margin) for sourcing deals for them. This could be somewhat material as they have the largest lease partner network in China with over 3,000 lease partners and 600,000 vehicles that have been leased as of 1Q21. Drivers like utilizing their lease services because they can receive lower monthly payments (20% lower than directly leasing from a leasing company) and shorter lease terms, which on average are just 6-12 months. This helps lower the bar for many to enter the driving pool and helps Didi’s recruit drivers.
They have more than 8,000 charging and fueling stations in their partner network, and drivers receive fuel discounts and can book charging stations appointments through their app. The fuel discounts help make driving more affordable, saving more margin for the driver. They also provide a network of approved maintenance shops to direct drivers to so they have some assurance of quality. Strangely, they actually own some of these repair shops and they do not clarify how many they own and operate. While car maintenance services are less built out than other developed countries, it is odd that this was a business they wanted to get into as it is very capital and people intensive versus their core platform business. In 2018 they created Xiaoju Automobile Solutions to supposedly raise independent capital with the speculation that they would spin it off ahead of an IPO. $1bn was raised but it seems like all of that was from Didi. There are several articles that imply this spin-off has happened, but that does not seem to be the case (in articles published in 2020, Xiaoju Automobile Solutions is referred to as a “subsidiary” of Didi) and they have gross RMB 3.4bn of Vehicles (some could be related to their D1 vehicle mentioned later) on their B/S and RMB ~165mn of finance lease receivables today. This is a relatively small amount, but also a very different business with capital requirements and credit risk. We would become uncomfortable if their lease business grew too large, but this does not seem to be the case with the book having peaked in 2019 at RMB ~350mn.
Overall, the advantages of this business to a ride sharing network are to enable more people to become Didi drivers and help keep them more captive to Didi. Drivers have to be registered Didi drivers and use their app to access the services, but there aren’t usage requirements to keep them from splitting driving time elsewhere and still taking advantage of Didi services and discounts. With their current dominant market share though, Didi is probably not too worried about drivers going to other platforms, but it is worth noting they could lock down their drivers better if they wanted to (3mn of their drivers used at least one solution in 2020). While it makes sense for Didi to help seed some of these driver services to help spur more ride hail drivers, longer term as these markets become more developed, it is hard to see why they would need to be in these businesses as they offer limited strategic advantages, are capital intensive, and are unlikely to be very profitable. Unfortunately, the lack of disclosures prevents us from being able to dig in deeper.
Bike sharing. This service allows a consumer to find one of the millions of bikes on city streets on a phone app and unlock it with a QR code. The consumer can then ride the bike wherever they want to go, hop off, and end the ride by locking it. It is a very convenient way to travel for shorter distance trips and for longer distances they also have electric bikes. The consumer is charged a small fee of RMB 1.5 for unlocking it and then another RMB 1.5 every 30 minutes. Didi entered this space initially through an investment in the now defunct Ofo, then acquired some of the now defunct Bluegogo’s assets, and ultimately launched their own service, Qingju. The bike sharing space is incredibly competitive, partly due to how good the unit economics can superficially look. As we show below, many start-ups were attracted the industry because of the incredibly attractive unit economics with payback periods as low a month with moderate frequency. However, no one foresaw how damaged the bikes would become and how common theft would be. In fact, people often intentionally destroy the bikes and many can be found in rivers.
In the unit economic sketch above we estimate most of the costs, but the point stands that it is very easy to show incredibly strong unit economics if you assume a bike can last 3 years (or even 2) and frequency is just moderate. This attractive payback is what brought a glut of competitors and capital into the market. In reality though, the unit economics were much worse with most bikes lasting much less than a year. The increase in competition drove an oversupply of bikes and relative frequency suffered as a result. Tack on increased subsidies to get consumers to try their respective service and many start-ups were unable to bear the losses: Kuqi Bikes, Dingding Bikes, 3Vbikes, Wukong Bikes, Ofo, and many others were all forced to shut down operations. While the the start-ups thought they could get really quick paybacks, perhaps as low as a month, in reality they were usually contribution margin negative. The Actual Unit Economic sketch on the right is informed by Hellobike, which released an IPO prospectus before deciding to not IPO. In it we can extrapolate that average frequency is only ~1.4 rides per bike and contribution margin is close to 0 (they report just 7% gross margins). This is a great example of how a great business in theory can be decimated by idealistic assumptions and a ruthless competitive response.
Today, bike sharing competition has started to rationalize, with most players having been driven out of the market. The big remaining players are Hellobike, which is backed by Ant Financial (36% stake), Mobike which was renamed to Meituan bike after being acquired, and Didi’s Qingju. Both Hellobike and Qingju have claimed a leading market share, but given how often bikes are salvaged and the lack of information Qingju provides, it is hard to scrutinize either claims. While we do know that Qingju has fewer bikes and e-bikes with 7mn vs. Hellobike’s 10mn, depending on frequency Didi could theoretically have a higher share of rides. The market is also undergoing another shift from regular bikes to e-bikes. e-bikes have a small electric motor in them which helps the rider go farther without exerting as much effort. e-bike unit economics are likely similarly attractive, but the big question is how often they have to be replaced from damage. The higher outlay for an e-bike could make losses higher and not be offset by more usage or higher rental prices. Similar to Auto Solutions, we cannot go into more detail due to limited information, but we do know that they raised some outside capital from Softbank at a $1.9bn valuation in 2021 and hold a ~88% stake. We know on a combined segment level “Other Innovations” is loss making, but we would guess most of the gross profits, estimated crudely to be RMB 3bn, belong to the bike share and automotive segment. In the future, if they can make their bikes stronger and more theft proof to prevent damage and loss, then the unit economics will greatly improve. Additionally, if bike share supply further rationalizes, bolstering average frequency, this could become a nice cash generative business, but they appear from that today.
Intra-city freight. This is another early innings business that aims to do what Didi did for passenger rides for shipping goods. Similar to ride share, it is an on-demand network with a large numbers of drivers who source one off jobs. There is a need for on-demand shipping as it can help businesses better manage surge demand and removes the need for excess flex logistics capacity. They have only disclosed that they completed 11mn trips from July 2020 to Dec 2020, but haven’t provided any business operating metrics since. There is reason to believe the margins in this business can be ultimately better than ride-hailing as it is easier to batch orders from multiple delivery requests and there could be less time sensitivity. Didi raised $1.5bn in outside capital to fund their expansion at a $2.8bn valuation and owns ~58% of this initiative. We know the TAM here is large (> RMB 1tn), but it is hard to have much of an opinion given the limited resources to diligence.
CGB. Community group buying was launched in June 2020, fairly late to the rapidly growing CGB segment. Their model was similar to others, where consumer orders were aggregated and then delivered to a single group leader. Consumers would then pick their items up from the single spot, obviating the need for separate deliveries and allowing consumers to aggregate purchasing power (for a fuller analysis on CGB see our BABA piece). Didi quickly incurred very high losses and then deconsolidated their CGB segment prior to IPOing so they wouldn’t be included in the P&L. It was last valued at $1.8bn and Didi has a ~33% stake in it. Given their very limited traction and better positioned competitors, we are not too optimistic about the future of this business and would not be surprised if it was shuttered.
AV and EV. The Autonomous Vehicle section below talks about the implications of AV on the ride hailing business, but first we will talk about Didi’s own AV and EV efforts. Their electric vehicle efforts include designing their own vehicle in partnership with BYD, dubbed the D1, and also building out a network of charging stations. They claim to have 30% market share of total public charging volume today across their partner network with 1mn electric vehicles on their platform. The focus on EV is because AV cars are always electric and so by controlling some of the EV infrastructure they can perhaps have a leg up in AV. Didi claims that their D1, the electric car they designed, is purpose-built for ride sharing, but the features they flaunt (sliding doors, ergonomic seats, and driver assistance system) are uninspiring and common in other vehicles. From our perspective, and with only 1,000 vehicles produced so far, it seems the real reason they ventured into EV production was to be ready for when autonomous technology comes online with a vehicle blueprint that can quickly be integrated with the new technology. Their EV services are not intended to be stand-alone profitable (as far we understand), but rather help foster an ecosystem of EVs that are human driven that will eventually help serve the AV community.
Didi runs their own Autonomous Vehicle research division, in partnership with Nvidia and Volvo. They are definitely a viable candidate to eventually reach fully autonomous cars, but are well behind Baidu and Alibaba in China. They raised money for this division several times, including a $500mn round in 2020, $300mn in February 2021 and another $500mn in March 2021 at a $6bn valuation. They are estimated to have a ~70% stake in their AV unit. Their AV efforts are likely to be a significant burn of cash as research continues to ramp up.
Below we talk about the ramifications of Autonomous Vehicles for their core ridesharing business. Of course, the future is unknowable and all of this is speculative, but we explore different possibilities to get a better sense of the risks and potential upside AV brings. At one extreme of thinking, autonomous driving could make Didi widely more profitable and at the other it is an imminent risk that is likely to end the viability of their business model.
First, we think its important to note that no matter when we get AV or who is first to figure out fully autonomous driving, there will be a transition period where most cars on the road are manually driven. This means that it is very unlikely a ride hailing network would be able to service all consumer demand with just AV cars initially. Didi currently has ~10mn drivers just in China, which if you assume only a quarter are on the road at any time, that is 2.5mn cars that are needed. At just $40k per car, that implies $100bn in capex is needed to build the fleet. Not an insurmountable amount for such an important project, but certainly not easily financed either.
Since the transition to a fully autonomous fleet will be slow, Didi will need to rely on having human drivers alongside AV to handle consumer ride requests. The fact that this transition will not be immediate is a saving grace for Didi, as the ability to displace their network with a new ride hail service that is fully AV seems implausible. However, Didi has an Achille’s Heel that was surfaced in our consumer surveys. If you recall, over half of all rides are sourced through an aggregation app, supplanting Didi’s direct relationship with the consumer. In a world with new AV players trying to enter the ride share market, they do not have to amass their own consumer network for their model to work, they can simply plug into Gaode or Baidu maps and receive consumer requests without having to go through the billions of subsidies Didi spent to gain critical scale.
The ease to which AVs can source ride demand has very negative implications for Didi’s moat longer-term. Today Didi may have a tenuous relationship with the consumers, but at least they have a better lock on the drivers. There are other ride hailing services, but none of them can come close to delivering drivers as many rides consistently and drivers who are not on Didi are likely missing out on rides and thus earning less. For these reasons (not to mention the Auto Solutions group benefits), Didi will fare well versus any potential competition for drivers. Today a driver who wants to pick up passengers may have to plug into Didi’s network, but in the future the autonomous vehicle may be able to circumvent Didi’s network all together and plug directly into the ride aggregator for rides. This isn’t farfetched as the two leading AV researchers, Alibaba and Baidu, both operate their own ride aggregation apps, which garner significant consumer demand. This puts Didi in an extremely precarious position as Baidu and Alibaba can slowly seed their autonomous ride hailing network by filling demand with their AVs and forking over the rest to Didi or other ride hail networks. This would not diminish the aggregator value prop for the consumer and would allow each AV player to enter the market with relatively little friction. This could be a long headwind to Didi’s ride hail business that chips away at GTV indefinitely. However, Didi could respond by pulling the API that allows Gaode and Baidu Maps to talk to Didi, which would likely make consumers go directly to Didi for rides as the aggregator apps cannot fill all demand without them. But pulling the API could also open the door for more competition and result in other ride hailing services quickly trying to fill the gap (Meituan has been known to be very opportunist in this regard, see regulation section).
There are positive outcomes from AV that are on the table for Didi though. Today, Didi has limited pricing power because consumers respond to higher ride prices with less usage and they cannot squeeze drivers’ earnings for fear of regulatory backlash and them shifting to other platforms (or jobs). However, autonomous vehicles eliminate drivers’ earnings, which leaves more profit to be split between the owner of the car and Didi. For illustration, if Didi owned the car, then the unit economics will look something like the exhibit below.
We see that with AV, Didi’s profits per ride increase ~7x from RMB 0.8 to ~6. While Didi has an AV unit, it is unclear if they aim to be the owner of the vehicle. This will give them the rights to a higher portion of profit, but also require more capital (or debt if they finance it). We assume above that they finance 100% of it, but they will likely have to put up some equity capital as well (putting 20% of the purchase price down decreases annual financing cost 20% and profit per transaction 50% to RMB 8.7). Whatever the numbers, clearly there is more money to split, it is questionable how much of it goes to Didi and whether ride price drops eliminate most of the increased profits.
Over time, despite how advanced the technology today seems, it is likely that AV capabilities become commoditized and profits are driven down. Historically, most technologically advanced, but capex-heavy, consumer-oriented industries have seldom turned large profits for long (everything from automobiles to refrigerators and toasters to TVs come to mind). As more AV players enter the market, which will include not just big tech players like Alibaba and Baidu, but also financiers of AV fleets and even individual private car owners who want to “rent” their cars out when they are not in use. AV operators are incentivized to reach for incremental demand by lowering ride price, since marginal costs to operate are minimal. While supply fractionates and consumers are indifferent to which AV provider serves them, Econ 101 would say that price drops until marginal cost equals marginal revenue, implying the cost savings will be competed away. It is also easy to see how private car owners would not calculate the cost of car ownership in their decision to “rent” their car out and thus could also drive prices down. In this scenario, we see prices dropping and profit margins falling, but the flip side is Didi’s value as an aggregator of autonomous vehicles is high as there can still only be a few (but more likely one) service that is best able to find an AV owner the most rides. The other piece worth mentioning is that Didi possibly has a data advantage and their algorithms are better refined to match a rider and consumer. With this advantage, they can help AV owners earn more profit per hour on their cars versus other networks with less optimized algorithms. The caveat to this, as mentioned, is if Baidu and Alibaba, who are also collecting data through their map aggregators, are able to slowly ween off the Didi network, which means an AV player may be equally suited by using any of the options.
Whether AV ends up being 1) a big margin boost to Didi or a negligible impact or 2) the factor that ends their ride hailing dominance or entrenches it, is highly speculative so we cannot take a strong opinion on how any of this will play out. However, we do believe strongly that investing in a company that requires grappling with such ambiguities and disparate outcomes necessitates a higher-than-average return, absent of which an investor would be best suited looking elsewhere. You will have a better sense of where Didi lies after going through the revenue builds and valuation sections below and of course you may have better judgement on what fair assumptions are than we do.
After a rushed road show, pricing was set just 3 days after launching and, on June 30th, Didi IPOed at $14 a share, raising $4.4bn at a $73bn valuation. Just 2 days later, the Cybersecurity Review Office of China announced they were conducting a cybersecurity review of the company and new user registration on Didi would be suspended in the interim. On July 4th, the Cybersecurity Administration of China (CAC) which houses the Cyberspace Review Office, ordered Didi’s apps to be removed from app stores and accused Didi of having seriously violating laws and regulations around collecting and using private info. Didi users can continue to use pre-downloaded apps as usual though. However, while ride hailing market share was pretty steady for the last few years, this presented an opportunity for competitors to turn up the intensity to acquire users while Didi is handicapped and distracted. Many sub-scale ride share platforms are taking this opportunity to bring back high subsidies aimed at winning over drivers and consumers. Perhaps most concerning though, Meituan revived their ride sharing, stand-alone app that was discontinued in 2017, in under 10 days (!), and is aggressively trying to suck up riders and drivers with promotions. On July 5th the CAC’s investigation was expanded to Zhipin, Huochebang and Full Truck Alliance, which makes it look like this new push is a broader issue than just misconduct from Didi, though the CAC could still want to make an example out of them. In response to all of this a flurry of speculation proliferated around why Didi IPOed at all, whether they knew about and whether there are broader implications for all Chinese stocks. Below we will walk through some common questions that come to mind.
Why did Didi IPO when it did?
Many investors including Softbank (>20% stake) have been pushing for an IPO so they can exit the investment for some time. The very receptive IPO market, especially in the US, meant that capital market conditions were about as good as they can get, so it made sense to IPO. Additionally, many of their “Other Innovations” are burning prodigious amounts of cash and if Didi is serious about all of these new areas of development, in addition to their international expansion, they will need to raise money because their core operations are barely profitable. There is another more pernicious reason, which is that Didi knew these regulatory changes were coming and waiting to IPO to address these issues could take too long and they would have to do another private raise or have to pull back on growth investments, so they quickly rushed the process before the new rules hit (the quick roadshow seems to support this).
Exactly what Didi knew is unclear, but it was known that a data security law would go into effect 9/1/2021 that covered collection and usage of data within China and regulated cross-border transfer of data, specifically for national security reasons. This is important in light of the proposed PCAOB (Public Company Accounting Oversight Board in the US) proposed a draft rule that would speed up the implantation of a Trump-era law–The Holding Foreign Companies Accountability Act–that requires all US listed companies to be subject to PCAOB oversight. This essentially means the PCAOB, a US entity, would need to be able to independently confirm all financial data for Chinese companies listed in the US, like Didi. Confirming the veracity of financial data though, requires full access to operational data (to ensure the numbers are not fabricated), which means a US institution would need full access to Didi’s data to ensure financial figures match operational data. This of course, is a nonstarter for China, who, we hypothesize, worries that access to this data could allow the US to know travel patterns of Chinese government officials because Didi will show routed traffic around the routes they are traveling. It is speculated that rather than deal with how all of this will be resolved, Didi pushed for their IPO right before the 7/1/21 Communist Chinese Party Centennial celebration, hoping the government would be distracted. It’s worth mentioning that circumventing laws and regulations seems to be a feature of building ride-sharing businesses.
Why didn’t Didi just list in Hong Kong?
The Hong Kong Exchange (HKEX) has stricter listing requirements than the NYSE or NASDAQ with strict profit requirements. They have been loosening them recently and it is likely Didi could have been granted relief from those restrictions as a “growth stage company,” but it still would have created tension as they have never reported positive operating income. Additionally, and perhaps more importantly, it turns out most of Didi’s drivers do not have both of the licenses required to drive (you need both a driver’s license and what is basically a “car license”) and this lack of regulatory compliance would have been a nonstarter to listing. Lastly, there was also possibly some currency-related issues for investors, and it is possible some funding rounds stipulated an exit in USD.
Did Didi know about the investigation prior?
It is unclear, but Didi denies that they did. Many pundits think that the pattern of facts seems to suggest that they did. Either way, there is a class action shareholder lawsuit underway for withholding information and misleading investors that will adjudicate that claim.
What are the broader implications for Chinese firms?
It seems that Chinese firms with any kind of “sensitive” data will have to undergo data reviews and it is possible they will be swayed to not IPO in the US, even though the Chinese government has explicitly stated they are not restricting Chinese firms from IPOing in the US. China still needs access to international capital markets and the US is the best way to get that, but we could see companies in certain businesses being privately pushed to list in Hong Kong instead until the issues with the PCAOB get closer to being resolved. In the meantime, the SEC has frozen processing registration of Chinese companies while it creates new guidance.
Are delistings imminent?
As it stands today, the PCAOB’s rules have not been finalized and, once they are, Chinese firms will have 3 years to comply. It seems unlikely to us that total delistings of Chinese firms are an optimal outcome to either China or the US, as there is plenty of US interest in wanting to keep Chinese firms listed in the US and China still wants easy access to US capital markets. Of course, we do not know anything you don’t, but our base assumption is that China and the US will figure something out. It does seem Chinese companies are now taking considerations of a Hong Kong listing more seriously.
How will this be resolved for Didi?
The Cybersecurity Review Office has 30 days to complete a preliminary review, which can be extended by 15 days, and offer recommendations to regulators. Regulators then have 15 days to respond. In theme with prior regulatory actions, we believe a remedy will be reached fairly quickly (months as opposed to years in the US). Possibilities include 1) greater regulatory oversight of their data collection and usage, 2) restricting data collection, 3) a large fine, and 4) even possibly going private again. It is worth remembering that over 10mn Chinese rely on Didi for their incomes and hundreds of millions of Chinese consumers depend on them for transportation, so we do not believe Chinese Authorities will want to cripple them completely. A large fine could be fairly disruptive to them given they do not generate much cash and needed IPO proceeds to continue their growth initiatives (which is unlike BABA who was able to easily handle their record-setting $2.8bn fine). A fine larger than what BABA was served could eat up most of their IPO proceeds, especially if there is also a successful shareholder lawsuit, which means they will need to curtail growth ambitions or raise again.
Worse than a fine though, is the damage the Chinese Administration is doing to Didi by restricting their apps and emboldening the competition. As mentioned, many competitors have kicked back up subsidies to grab market share from Didi, and Meituan even revived the stand-alone ride hail ambitions that they abandoned in 2017. We’ve largely ignored competition throughout this write-up because it hasn’t been a big factor since beating out Uber (~90% of market share) and there currently is not a clear focal competitor. But this opens the door to change. Meituan can not only go after new users, but capitalize on the bad taste Didi’s actions have left with consumers. It’s only been a few weeks, so there isn’t much we can say at this point, but we have seen Meituan be ruthless in the past and it is possible they build a toehold in ride hail that allows them to build off of, even after Didi is out of the regulatory dog house. However, while we want to be aware of Meituan and the recent uptick in competitive intensity, we think it does not present a new era of increased competition and it is most likely that Meituan’s endeavors are largely experimental and rolled back when Didi is fully back online.
One last thought on regulatory ramifications is that to fully roll out AV, Didi needs to work closely with regulators to win their approval and this kerfuffle does not bode well that they will be able to easily get it. While they have permission to test AVs in certain cities today, future permissions could be harder to get. Generally speaking though, our base assumption is that in a year or two there won’t be any material adverse outcomes they still are affected by. We could be wrong.
Revenue and NOPAT Build.
Below is our revenue and NOPAT build for Didi’s China and International businesses, which does not include any of the businesses that sit in “Other Innovations.” We wanted to present it on a “Net” basis for clarity so keep that in mind when comparing it to other exhibits. Our estimates are based off of triangulating limited segment disclosures and the disclosed 2020 ride sharing economics breakdown. We wanted to show what the business would look like with what we consider fairly optimistic assumption to steel man (inverse of straw man) the investment case. We have transaction frequency doubling from 23.8x (coincidentally the same for both China and International) to 50x. We grow the international business considerably to 300mn users as well. The only other big profit lever we are not including in the below build is the impact of AV. As we discussed, it is very tricky to know how that will play out for Didi, but in a very bullish scenario they could earning ~7x per ride what they usually do or, in a bearish scenario, lose their relevance and bleed market share indefinitely. Rolling up these assumptions and applying significant S&M leverage, we can get Didi to earn RMB ~20bn of NOPAT in 7 years from just their China Mobility and International businesses. If you recall our previous commentary on S&M leverage, we were unconvinced that they will be able to lower their S&M spend and keep frequency flat. In the model below, we are both dropping S&M and increasing frequency, which would necessitate a marked shift in how consumers use ride hailing. It should be noted that most of this profit will be plowed back into their Other Innovations, so an investor must believe those are value creating opportunities and they are not burning cash while building lousy businesses with dubious profitability. While we are only mentioning this in passing, this is a critical point to consider: the majority of their internally generated cash flow will be spent on these initiatives and not reinvested into the only business that is currently generating a profit (China ride hailing), so to believe Didi merits an investment you must think positively of where they are allocating all of their cash. As always, Members Plus can adjust the assumptions as they see fit.
Our valuation below builds off of the NOPAT Build above. We apply a range of EBIT margins so an investor can better see the distribution of outcomes. With a 35-40% margin (on a net basis), and a 15-20x multiple on their ride share businesses (which is fair if they have similar growth in 7 years as they did last year), an investor would earn a 0-6% annual return.
1) Regulation. Outcomes of regulation are unknown and something worse than we can currently conceive is possible. A large fine may be hard for Didi to withstand without raising additional capital, diluting shareholders. It is also possible they face local regulation and have their services restricted on a city-by-city basis.
2) Competition. Competition is now emboldened and perhaps a new era of competitive intensity is beginning with Meituan relaunching their ride hailing app. It has been 4 years since Meituan abandoned their ride hailing ambitions, but now that they are a much bigger company with more dominance in local food delivery, this could be a battle they are more willing to fight today.
3) Autonomous Vehicles. It is entirely unclear if AV will be a boon to profits and help entrench Didi’s position in the ride share ecosystem or undermine their functionality by opening up new avenues for consumers to access ride share. If Didi didn’t exist or was drastically smaller in 20 years, today the best guess on why that would be is AV.
4) New Innovations destroy value. All of their new innovations are losing money today without clear market dominance or a path to profitability. Almost all of their internally generated cash is going to unprofitable businesses that could prove to be poor uses of capital.
5) Burn through cash. Today they an estimated ~6.4bn of cash after their IPO, but large fines and a successful shareholder lawsuit could mean they have to raise more to continue to sustain their money-losing endeavors. If the fine and shareholder lawsuit result in outsized monetary compensation, then Didi could be crunched for cash. In such a scenario it is possible their stock drops considerably and Didi must raise money at an unfavorable rate.
6) Accounting Gimmicks. Didi has very complicated accounting with different revenue recognition methods and expense treatment consolidated within a single segment. The lack of disclosures makes a lot of analysis tricky to do and present ample opportunity to manage the financial statements.
7) Growth never comes. The viability of Didi’s investment thesis is predicated on an acceleration of growth in their core business or an incredibly high growth in their international business to make up for China not growing, but it is very possible that it never materializes.
8) Delisted. We do not think this is high probability, but it is possible that China forces Didi to delist, whereby it is unclear how investors will be able to trade shares or how much of a discount that will result in.
Below we show our model summary for Didi. We model starting with their disclosed revenue segments, but that gross vs. net issue creates a problem for estimated cost of revenues. We try to back into payments from drivers by taking the difference between platform sales and revenues, but if you recall this isn’t totally correct since platform sales adds back consumer incentives (and we do not know what consumer incentives are on an annual basis). Nevertheless we thought this was still a better way to model the cost of revenue item than grouping driver payments with other cost of revenue items, but we acknowledge it is far from perfect. Our expense line items are driven on gross profit too to get around the net vs gross issue, but be aware if you are increasing gross margin in the model you will need to offset that by lowering expense % assumptions, otherwise they will increase. As you can see we are showing an acceleration in growth in the China business, which is considered bullish. Nevertheless, we do not see Didi making any operating profit on a consolidated basis in the period through 2025.
A special thank you to all of our subscribers! We hope you enjoyed the write-up on Didi and we kept our promise on making the accounting clear enough! Please join our discord group for further talk on Didi and if you have any questions on the piece—we will be making a new chat just for them. Thanks again!!!
*Disclaimer: People who have helped craft this analysis may individually (or through funds they work at) have a position in the company. Absolutely nothing in this report is investment advice nor should it be construed as such. We make no claims to the veracity of all facts and figures presented in this report. Certain figures could be stated erroneously and certain analysis could be incorrect. Please see the full disclaimer linked above.
DJY Research, its author(s), any of its contributors, and any funds they may advise at may own securities in the companies, or related companies, written about on this website.
All rights reserved. No content on the DJY Research website can be redistributed without the express written consent of DJY Research.