The richer you look, the harder they come for you

Today, I'm digging into Meituan (3690.HK).

In my last profile, I looked at JD.com. They are the $190B Chinese retailer that voluntarily torched 90% of its operating profit to start a food delivery business from scratch (in an already competitive industry).

Today is the other side of that trade. Meituan is the incumbent JD attacked, and the damage to the defender appears to be worse than the damage to the aggressor.

Meituan runs China's dominant local services super-app. Around 700 million annual users open it for food in 30 minutes, hotels, hair appointments, instant grocery, and Dianping restaurant reviews, with about 60 million transactions flowing through it daily.

(Quick note on currency: Meituan reports in RMB, so dollar figures below are translated at roughly 6.79 RMB per USD (HKD pegged near 7.78). For simplicity, I’ve held this constant across years).

A few things that stood out to me:

• Meituan was a printing press in 2024, then JD showed up. Operating profit jumped from about $1.2B in 2023 to roughly $5.0B in 2024. JD formally launched food delivery in April 2025 and started subsidizing aggressively. Meituan ended 2025 with revenue of about $53.7B (up 8%) and an operating loss of about $4.4B. Roughly $9B of operating profit gone in a year!

• The Core Local Commerce segment, the cash engine, swung from about a $7B operating profit in 2024 to a $1B operating loss in 2025. The moat reacted exactly the way JD's CFO seems to have hoped it would.

• They have rock-solid balance sheet to fight from. Meituan holds roughly $29B in net cash, almost 5x JD's $6.3B.

• JD chose to hire 120,000 full-time food couriers. Meituan still leans on gig riders. If the regulator ever pushes employment classification, it coiuld get ugly Meituan.

A couple more bets they are making:

• Keeta, the international food delivery brand, is live in Hong Kong, Saudi Arabia, the UAE, Qatar, Kuwait, and Brazil, with a $1B commitment to Brazil alone.

• On top of that, they are pouring billions of RMB into GPU clusters and a homegrown trillion-parameter LLM called LongCat. The focus of their LLM efforts appears to be on enabling its employees and improving their products. Not selling the LLM to others (like Anthropic).

For a decade Meituan looked like the textbook answer to "what does a dominant local services platform earn at scale." Operating margins finally cleared double digits in 2024 and the company was buying back stock. One credible entrant with deep pockets erased the entire profit pool in twelve months.

My key takeaway: Scale can turn a moat into a target. Meituan is a reminder that huge markets attract huge attackers. That makes me more interested in smaller vertical businesses, like Fullstack Ag (the AI/analytics for agriculture course and community venture I’m working on) , where the opportunity may be big enough for the company but not big enough to tempt a JD-scale assault.

This will be a fun market share battle to watch play out.

With that, I'll see you on Tuesday!

Nick

TL;DR

• Meituan is China's dominant local services super-app, anchored by food delivery, in-store coupons, hotel and travel bookings, and a fast-growing instant grocery and international delivery arm.

• About 71% of FY2025 revenue came from Core Local Commerce, with 29% from New Initiatives (grocery, Keeta international, B2B food distribution).

• 2025 swung from a record profit to an operating loss as JD entered food delivery and a brutal subsidy war drove Core Local Commerce from a roughly $7B operating profit to a $1B loss.

• Capex is climbing fast, driven by GPU buildout for the LongCat LLM and overseas expansion through Keeta.

• The stock returned roughly -45% over the trailing year and about -71% over five years. JD returned about -1% and Alibaba about +6% over the same one year window. The defender of the food-delivery fight took the bigger hit than the aggressor.

• Operator takeaway: dominant share in a network effects business is real, but it is not the same as pricing power once a deep-pocketed entrant decides the category is strategic.

The 30,000-Foot View

Meituan runs a consumer super-app on top of an enormous physical delivery network. The app aggregates demand across dozens of high-frequency local categories (a lunch, a haircut, a hotel room, a bottle of shampoo in 30 minutes), and the operations side moves the food, goods, and riders to fulfill it. Merchants pay commissions and marketing fees. The Dianping review layer (the closest Chinese analogue to Yelp + OpenTable + Google Maps reviews in one) sits on top and feeds discovery and ranking.

The density flywheel is what makes this hard to copy. More merchants attract more consumers, who justify more riders per square kilometer, who compress delivery times. JD's strategic insight was that this flywheel could be neutralized by a rival willing to hire its own riders and pay subsidies long enough to seed the other side. That is exactly what 2025 played out.

The response is offense in two directions at once. Meituan is leaning into international (Keeta) and AI (LongCat). Both are real businesses with real spend, and both will look smart or foolish in two years depending on whether the China truce holds.

Revenue mix (FY2025):

• Core Local Commerce (food delivery, in-store, hotel, travel): ~$37.7B, ~71.5%

• New Initiatives & Others (grocery, Keeta, B2B food distribution, instashopping): ~$15.1B, ~28.5%

Key Stats

• Market cap: ~$63B

• TTM revenue: ~$53.7B

• TTM gross margin: ~30.4%

• TTM operating margin: ~-8.2%

• Employees: ~111,000

• Net cash position: ~$29B

• 1Y total return: ~-45%

• Industry: Specialty retail / consumer cyclical, China local services

• Listing: Hong Kong Stock Exchange, ticker 3690

Company History

• 2010: Meituan launches as a group-buying site, modeled on the Groupon playbook for China.

• 2013–2014: Pivots aggressively into mobile and food delivery, going head-to-head with Ele.me (later acquired by Alibaba) and Baidu Waimai.

• 2015: Merges with Dianping, locking in the most important review and discovery layer in Chinese local services.

• 2018: IPO on the Hong Kong Stock Exchange (September 20) at HK$69 per share.

• February 2021: Shares peak at HK$460. Antitrust regulators later fine Meituan RMB 3.4B for merchant exclusivity practices.

• 2023: Launches Keeta in Hong Kong (May). Within ten months Keeta becomes the largest food delivery platform in Hong Kong with about 44% share.

• September 2024: Keeta launches in Saudi Arabia, the first major international expansion outside Greater China.

• 2024: Record year. Revenue of about $49.7B, gross margin of 38.4%, operating profit of roughly $5.0B.

• April 2025: JD.com formally enters food delivery. Within months, Ele.me, JD, and Meituan are in an open subsidy war.

• 2025: Keeta expands into Qatar, Kuwait, the UAE, and Brazil ($1B Brazil commitment). LongCat-Flash, a 560B-parameter open-source LLM, is released, and an AI ordering agent called Xiaomei goes live in the consumer app.

• Full year 2025: Revenue grows 8% to about $53.7B. The company swings to an operating loss of roughly $4.4B. CEO Wang Xing publicly vows to "firmly oppose inward competition" and pivots messaging toward AI and overseas growth.

• June 1, 2026 (upcoming): Q1 2026 results scheduled, the first clean read on whether the August 2025 industry truce held.

Show Me the Money

Standout financial features:

• Operating margin compressed from roughly 10% to roughly -8% in twelve months, driven by Core Local Commerce alone swinging from a $7B profit to a $1B loss.

• Gross margin compressed from 38.4% to 30.4%, absorbing rider subsidies, merchant rebates, and near-free orders. For comparison, JD's gross margin sits closer to 9 to 10% by structure (asset-heavy 1P retail), so Meituan still produces three times the gross margin per dollar of revenue. That difference is what makes the 2025 swing into losses so notable.

• CapEx more than doubled from $1.0B in 2023 to $2.2B in 2025, driven by GPU buildout for LongCat and physical infrastructure for instashopping and overseas.

• Rock solid balance sheet. At FY2024 Meituan held about $24.8B cash and $13.3B short-term investments against $9B of debt, a net cash position of roughly $29B. Almost five times JD's $6.3B, which matters a lot if the war drags into 2027.

Financial Data

All figures translated at the current spot of about 6.79 RMB per USD, held constant across years.

(Net debt is negative because Meituan holds far more cash than debt. The 2023 and 2025 figures are estimates anchored to the reported 2024 position and adjusted for 2025 cash burn.)

Stock Performance

(HKD returns treated as USD-equivalent given the peg. 5-year baseline is approximate based on the May 2021 trading range. Meituan does not pay a dividend.)

1-year peer comparison (price + dividend return):

Alibaba is the most natural comp because it owns Ele.me. JD is the right second pick because it caused the 2025 swing. The takeaway from the table is that this is not a sector-wide drawdown. The aggressor (JD) was roughly flat over the trailing year because investors gave the move the benefit of the doubt. The defender (Meituan) lost about 45% because investors priced in lost profits and an uncertain runway back to 10% margins. The trailing three months are even sharper: JD is up 22.5% as the food-delivery thesis gets re-rated, while Meituan is down another 11%. The market currently believes the attacker has more upside from this war than the defender does.

The N.O.O.B. Nine: Competitive Powers

The Nerd Out on Business Nine is made up of Hamliton Helmer's famous "7 Powers" of competitive advantage (Scale Economies, Network Economies, Counter-Positioning, Switching Costs, Branding, Cornered Resource, and Process Power) combined with two of my own (Data Flywheel and Distribution Advantage).

Average Score: 3.6/5 - A genuine moat built on density, data, and network economies, with two visible cracks: low counter-positioning (the model is copyable) and only moderate switching costs (multi-homing is easy when a rival pays for it).

Memorable Marketing

Meituan's playbook is mostly not advertising. The product itself does the heavy lifting (push notifications, coupons in the wallet, five-star reviews nudging a new place), and the brand spend tends to go to viral activations, holidays, and IP tie-ins rather than mass-market campaigns. The Dianping side runs more like a content business with KOL partnerships and curated city guides. JD's contrasting move with JoyBuy and JD Dianping is explicitly trying to attack that content moat.

Notable campaigns and tactics:

• Bonus Fandom Group merchant chats: Group chats that let high-frequency merchants run flash-sale and exclusive-coupon drops directly to followers. Functionally an owned-audience tool inside Meituan, similar to a private WeChat group.

• "Super Power" 2024 instant retail and gaming IP fusion: Co-branded campaigns with popular Chinese gaming IPs drove a Gen Z engagement surge for instashopping. Made a logistics service feel like a content drop.

• Green Packaging and Healthy Eating filters (2024): A 22% engagement uptick among health-conscious urban users after surfacing these filters in the app. Filters are marketing in a long-tail marketplace.

• Dianping food festivals and city guides: Curated "must-eat" lists and pop-up events that turn the review corpus into a destination, not just a search result.

• Keeta launch playbook (Hong Kong 2023, Saudi 2024, Brazil 2025): Aggressive zero-fee promotions and rider recruitment bonuses at launch, then a measured tightening once liquidity is established. Hong Kong, which got to 44% share in ten months, is the template.

Tactical takeaways:

1. Give merchants owned-audience tools on your platform, not just listings. It turns a transaction relationship into a tenancy relationship.

2. Co-brand around IP and pop culture on high-frequency surfaces. People scroll your app many times a day, and "yet another coupon" is invisible.

3. Filters and discovery flows are marketing, not just product. Surface a new value (healthy, sustainable, vegan) and the data will tell you what to do next.

4. When you enter a new geography, buy your way to liquidity fast and then taper. Half-spending the launch is worse than not launching.

AI Uses & Opportunities

Current exposure:

• LongCat-Flash is a 560B-parameter mixture-of-experts model open-sourced to attract developer attention. LongCat 2.0 was reportedly trained on 50,000+ domestic compute cards (Nvidia was not invited).

• Xiaomei, a consumer-facing voice ordering agent, runs on LongCat inside the Meituan app.

• AI is also used internally for routing, demand forecasting, fraud detection, Dianping moderation, and merchant ad placement.

• Worth noting against yesterday's JD profile: both are building proprietary LLMs (Meituan's LongCat, JD's JoyAI/Yanxi). JD's stack is already in production at scale with 150M+ users on the shopping assistant. Meituan's bet is bigger on infrastructure (training) and smaller on consumer surface area so far.

Future opportunities:

• An AI concierge for travel and in-store services that books across the full Meituan inventory by conversation rather than search.

• Merchant copilots that draft menus, optimize ad spend, respond to Dianping reviews, generate photos, and run promotions on autopilot.

• Automated dispatch and dynamic pricing that closes the gap between order surge and rider supply at peak hours.

• Localization at scale for Keeta, with on-the-fly translation of menus and reviews across Arabic, Portuguese, English, and Cantonese.

Bumps in the Road

• The price war is the headline risk. Even with the August 2025 truce, three well-capitalized players have shown they will subsidize aggressively. Returning to 10% margins requires industry discipline. That has not historically lasted in Chinese internet.

• JD is not going home. JD has its own logistics, its own consumer base, 120,000 full-time couriers, and a stated commitment to local services. Meituan now competes on what used to be its “owned” turf.

• The rider model asymmetry could flip against Meituan. JD hired its couriers as full-time employees. Meituan still uses a gig model. If regulators reclassify riders, Meituan absorbs a bigger one-time cost shock.

• Overseas expansion is expensive and unproven at scale. Keeta is winning early in smaller markets, but Brazil alone is a $1B commitment and local incumbents (iFood, Talabat, Careem) will not give up share quietly.

• AI capex is meaningful now. Billions of RMB into GPUs and LongCat is real spend while the core business is losing money. If LongCat does not produce a clear moat or revenue stream, the spend will look like vanity in hindsight.

Your Swipe File

• Operating profit in a marketplace is partly a truce, not just a moat. If a credible competitor decides your category is strategic, headline margins will compress fast no matter how strong your network effects are. Underwrite the durability of the truce, not just the strength of the moat.

• Density is the real cost advantage in physical-world platforms, but density only matters if your rival cares about cost. JD chose not to care for at least a year.

• A fortress balance sheet is what lets the defender survive a war the attacker also cannot easily afford. Meituan's $29B net cash is roughly 5x JD's $6.3B. The defender has a longer runway than the headlines suggest.

• Build merchant tools that look like tenancy, not just transactions. Group chats, fan tools, and owned-audience products turn marketplace listings into long-term locations on your platform.

• If you choose to expand to new geographies, use your launch geography to write a playbook that can be refined over time. Meituan's Hong Kong rollout in 2023 became the template for Saudi Arabia in 2024 and Brazil in 2025. Treat the first new market as a manufacturing line for more to come.