JD.com leverages AI to boost supply chain efficiency, driving user growth and profit gains, while making heavy asset investments in embodied intelligence to gather data and build its merchant ecosystem.
Over the past 10 quarters, industry reports have been dominated by talk of "traffic peaking." Major e-commerce platforms long ago stopped expanding aggressively, and everyone is focused on squeezing profits internally. Across the entire online retail sector, no one dares to let their guard down.
In this zero-sum game, JD.com's earnings stand out as an outlier. Its quarterly active user base has maintained double-digit growth for consecutive quarters. Looking at these 10 quarters as a whole, annual active users grew by more than 200 million during this period, pushing the overall user base quietly past the 740 million mark. Buyers are multiplying, and their shopping frequency is climbing alongside.
Where did these 200 million users come from? Relying purely on burning cash for one-off transactions could never generate growth on this scale. The era of front-end marketing-driven growth is long over. Downloads forced through algorithmic recommendations and pop-up coupons vanish the moment subsidies stop.
JD.com's ability to secure this growth hinges on its strategy of forging a triangular loop out of product, price, and service. The math works like this: the platform compresses back-end fulfillment costs to the limit, freeing up profit margins; that margin is passed on to merchants; merchants, with their baseline business at break-even, can afford to list lower prices on the front end; users get quality goods at fair prices plus reliable delivery, so traffic naturally sticks; the scale of users and high-frequency repeat purchases give merchants predictable expectations; merchants then reinvest their earnings into new product development, giving back to users.
This triangular loop has replaced the old "impossible trinity," where you could only pick two of the three: product, price, or service. In the old e-commerce model, rock-bottom prices meant tolerating inferior goods and slow shipping, while top-tier service came with steep price tags. JD.com, through sheer supply-chain efficiency at the foundational level, has forced all three onto the same table. E-commerce platforms and ecosystem players no longer bicker daily over a few cents of commission—they now bank on efficiency dividends.
Revenue of 315.7 Billion Yuan: A Triangular Loop Replaces the Impossible Trinity
In the first quarter, JD.com posted total revenue of 315.7 billion yuan, with Non-GAAP net profit attributable to shareholders of 7.4 billion yuan. Outside observers are quick to slap on the "cost-cutting and efficiency" label. But look at the R&D line in the earnings report: JD.com's R&D spending in Q1 rose 59% year-over-year. While everyone else trims budgets and slashes tech initiatives, AI has become one of the heaviest line items on JD.com's books.
That spending didn't drag down profitability—it actually propped up the 7.4 billion yuan profit pool. The operational numbers are right there on the table: the electronics and home appliances category held its ground; the daily necessities category posted double-digit year-over-year growth; service revenue hit 70.9 billion yuan, up 20.6% year-over-year.
There's a physical ceiling to supply-chain and back-end cost savings. No matter how hard you push delivery drivers, a day only has 24 hours; no matter how much you squeeze warehousing, the aisle spacing can't be less than a forklift's turning radius. Relying on that alone could never yield this 7.4 billion yuan in net profit, let alone service revenue growth of over 20%. Profit on this scale demands a full-scale, systemic overhaul.
Service revenue rising means third-party merchants are actually making money in this ecosystem and are willing to pay real cash for the platform's ads, warehousing services, and data tools. Double-digit growth in daily necessities means users are coming back more often for low-ticket, high-frequency items like toilet paper and soy sauce. The platform's user stickiness is now beyond doubt.
In this ledger, AI is a purely engineering-driven tool. It runs back-end smart warehousing and sorting, dynamic production scheduling, and autonomous vehicle dispatch. Computing power squeezes out redundant time in the fulfillment chain, second by second. A truck idling five fewer minutes at a sorting hub, a forklift taking ten fewer meters of detour in a warehouse—these seem like tiny gains, but multiplied by JD.com's massive order volume, they add up to a hidden fortune in profit.
The savings are passed on to merchants. With their profit floor secured, they can afford to cut prices on the front end. The price war has shed its vicious cycle of forcing merchants to bleed for participation. JD.com uses back-end computing investment to buy the license to open fire in front-end pricing.
Building Hardware and Harvesting Data: The Heavy-Asset Gamble of Embodied Intelligence
The software race is over. The technical barrier for text-based large language models has been flattened by the open-source community. The next key battleground is all about how robots get deployed to do real work.
The embodied intelligence track is currently bottlenecked on data. There's too little of it, and its quality is a mess. The tens of millions of hours of low-res video scraped off the internet can teach a large model to write poetry or paint, but they can't teach an industrial robotic arm how to precisely torque a screw in a messy factory floor environment. The three-dimensional physical world involves incredibly complex lighting, friction, and spatial depth—two-dimensional flat video data can't produce a robot that can actually get its hands dirty.
Faced with this critical bottleneck, JD.com has gone straight to building physical data infrastructure. Its self-developed capture terminal, the JoyEgoCam, weighs just 220 grams—lighter than an average smartphone. Workers hauling goods in warehouses or assembling parts on production lines usually operate in environments full of high noise and tricky lighting interference. A worker can wear this device for a full eight-hour shift without wrecking their neck. The machine precisely translates the wearer's real-world perspective and hand muscle movements into digital coordinates. The first-person video it records holds reprojection error tightly under 0.2 pixels. Only when a robot sees this clean, precise, firsthand data can it understand where to reach and how much force to apply when making contact.
Once the hardware is in place, the next phase is pure grunt work. JD.com plans to have up to 600,000 people wear the devices, amassing 10 million hours of real human video within two years. This is a heavy-asset gamble—a human-wave strategy to fill the data void, using real people to blaze the trail for machines. The financial risk is staggering: if the mass-collected data goes sideways during cleaning or alignment, or fails to match what the final model training actually needs, the massive investment in hundreds of thousands of man-hours, hardware, and time will rot on the vine. That's the brutal reality of the physical economy—pouring real money in without hearing a splash.
A 35 Billion Yuan Ecosystem Bill: Upfront Investment for Long-Term Lock-In
To judge a platform's ecosystem, don't read its press releases—check the survival rate and profit accounts of new merchants. In Q1, JD.com's third-party merchant count grew 57% year-over-year, with nearly 1,000 new merchants topping 10 million yuan in first-year sales.
The Jingmai marketplace has opened its monthly active traffic channels, in the hundreds of millions, to third-party sellers. At the same time, per JD.com's disclosures, a fund exceeding 35 billion yuan is on the table. JD.com offers new merchants up to one year of free tools, including various AI design tools. Previously, a small or mid-sized merchant setting up shop would shell out thousands to tens of thousands of yuan for a full visual package—hiring models, renting studios, retouching photos. Now, with AI tools, a few clicks and a few cents of electricity generate product images that meet platform standards.
Take a small robot-vacuum maker: its machines sell into third-tier cities, but once one breaks down, the company can't afford to build a local repair center. Without after-sales support, its reputation collapses. Facing this heavy-asset dead end, JD.com has opened up its own repair network, rolling out "robot ambulance" doorstep service, with plans to expand its robot repair engineer team to over 10,000 people.
This is a shrewd calculation. The old model of collecting rent and commissions on paid traffic squeezes small merchants to death in a downturn. JD.com's current play is to bundle upfront subsidies, tech infrastructure, and after-sales support into one package, absorbing the prohibitive trial-and-error costs for small merchants.
Front-loading costs will inevitably drag on the platform's own margins and eat into significant cash flow. But what it buys is permanent lock-in for everyone in the ecosystem. Once hardware makers rely on JD.com's after-sales network, and merchants get used to free AI image-generation tools, they survive and thrive within this chain. The logistics and data flows of the entire industry can no longer bypass JD.com. Enlarge the pond, fatten the fish, and JD.com will have absolute leverage and pricing power when collecting service fees down the road.
Where Technology Ultimately Lands: Purchase Reasons from Kids to Seniors
In the end, technology only matters if someone is willing to pay for it. Without a purchase reason, it's meaningless. No matter how elegant the code, if it can't monetize on the consumer side, it's just sunk cost sitting on the books.
JD.com's AI deployment is all baked into the mundane details of daily life. Anker's AI+UV printer has flattened the learning curve—kids and retirees can voice-command a one-touch print. With the barrier gone, ordinary people become new consumers. Seniors who buy smart appliances don't need to flip through manuals or hunt for remotes; a spoken command in the living room powers the device on. This straightforward approach wipes out the tedious app-configuration process entirely.
In traditional e-commerce pharmacy, a user with a headache buys headache pills and leaves, forcing the platform to spend money acquiring new customers through external traffic. This initiative aims to overhaul that one-and-done transactional model. JD Health's AI doctor "Dawei" has completed hundreds of millions of interactions with a satisfaction rate above 98%, pivoting toward long-cycle chronic disease management. Over the next year, it plans to connect 1 million senior-friendly devices—meaning the AI doctor doesn't wait for users to get a headache before selling pills; it proactively monitors blood pressure and blood sugar, converting one-off sales into long-term recurring engagement.
For new product launches, the hard logic at the consumer end is simple: algorithms sitting in the cloud eventually show up as pricey electricity bills. Turn AI into a seamless everyday tool, eliminate the learning curve, and only then will children and the elderly engage with it frequently. The more it's used, the more the platform and merchants achieve their goal: getting users to buy more, return more, and spend more.
Final Thoughts
Twenty-two years ago, Richard Liu shuttered his brick-and-mortar stores in Zhongguancun and pivoted fully online. Twenty-two years later, JD.com is carrying AI-equipped hardware devices back into the physical world.
The old offline model was about opening stores and selling goods, profiting from buy-sell spreads. Today's offline footprint is about embedding efficiency standards and interaction gateways into production lines, warehouses, and seniors' living rooms.
JD.com has run the numbers on a formula: "The value of AI = Model × Experience × (Industry Depth)²." Under this framework, the ledgers of retail, logistics, and manufacturing can't be tallied in isolation. Fulfillment efficiency drives retail experience—if delivery is fast enough, users won't defect. Retail data steers manufacturing schedules—knowing what users buy tells factories what to produce. These three physical industries are interlocked, forming a community of shared interests.
No matter how grand the technology narrative, if it doesn't show up in the earnings report—if it doesn't translate into one fewer box damaged in the warehouse or one fewer machine down in the factory—it's just an extremely expensive spin of the wheels. In the physical world, the ledger rules. When the numbers balance, the business stands.