This year's 618 shopping festival, JD's best-selling category wasn't phones or home appliances -- it was humanoid robots.

Sales grew over 10x year-over-year. While the base is still small, this growth curve looks exactly like smartphones in 2012, new energy vehicles in 2018 -- the eve of explosion is often quiet, but numbers don't lie.

Giants All Place Their Bets

Huawei launched the CloudRobo embodied intelligence platform at its summer launch event. Unlike others, Huawei's approach is "cloud-edge-device integration": cloud brain does complex decisions, edge does real-time perception, robot body does execution. This approach is very Huawei -- not single-point breakthroughs, but full-stack infrastructure.

Nvidia released the Jetson AGX Thor robotics chip, boosting local compute 7.5x. What does that mean? In the past, robots needed to connect to cloud brains via network for complex decisions, with high latency, instability, and fear of disconnection. Thor lets robots run LLMs locally; response speed drops from seconds to milliseconds -- a qualitative leap for robots that need real-time interaction with the physical world.

Tesla Optimus V3 mass production preparation is complete. Musk says small-scale deliveries begin end of 2026, with large-scale mass production in 2027. If this timeline holds -- note the "if" -- humanoid robot cost curves will drop rapidly like Tesla EVs, from millions to hundreds of thousands or even hundreds of thousands of RMB.

Why 2026?

Embodied AI isn't a new concept, so why is it exploding this year? Three conditions are simultaneously maturing:

First is the brain. LLM reasoning, multimodal understanding, and long-horizon planning capabilities are finally "good enough." Robots in 2023 couldn't even do "hand me the red cup on the table" well; robots in 2026 can complete simple factory assembly, restaurant food delivery, and warehouse sorting tasks.

Second is the cerebellum. Motor control was the biggest bottleneck for past robots -- ask a robot to grab a cup, it might crush it or grab air. But advances in reinforcement learning + simulation training are astonishing; robot hand manipulation dexterity now approaches human levels. The domestically produced AI dexterous robot hand demonstrated at Tianjin Davos can replicate the vast majority of human fine hand movements -- threading needles, folding paper cranes, no problem.

Third is cost. Sensor, motor, and compute unit prices have dropped over 70% in the past three years. A humanoid robot's BOM cost fell from 1.5 million RMB in 2023 to 300-500K in 2026, and continues to drop rapidly.

The First Scenarios to Be Changed

Don't think of robots and immediately picture sci-fi movie humanoid butlers. The first embodied AI scenarios to land will definitely be the "3D jobs" -- Dull, Dirty, Dangerous.

Repetitive assembly in factories, cargo handling in warehouses, inspection in dangerous environments, post-disaster rescue, agricultural harvesting, medical assistance -- these scenarios don't need robots to look human; they need robots that can act autonomously in the physical world. In fact, today's most mature embodied AI applications aren't humanoid at all, but various robotic arms, AGV carts, and drones.

Humanoid robots are the ultimate form, but not the first to land. Just as cars don't need to look like horses, robots don't need to look human to be useful. But the humanoid advantage is versatility -- a robot that can use tools and move in human environments like a person could theoretically replace any manual labor; the upside is enormous.

Where Are the Opportunities for Ordinary People?

The embodied AI explosion will, like the PC, internet, and smartphone before it, create entirely new professions and industry ecosystems.

Robot trainers -- teaching robots specific tasks, like teaching AI to write code today. Robot maintenance -- robots break and need people to fix them, like IT operations today. Scenario designers -- designing robot workflows in specific scenarios, like product managers today. Not to mention the countless jobs in robot R&D, manufacturing, and sales across the supply chain.

But more importantly: embodied AI will free humans from large amounts of repetitive manual labor. This isn't an "unemployment" story; it's an "upgrade" story -- just as excavators freed coal miners, washing machines freed housewives; technology never eliminates work, it moves people to more creative work.

Of course, this process won't be painless. The transition period will have growing pains, anxiety, fear of being obsoleted. But history tells us: every technology revolution ultimately creates more jobs than it eliminates.

AI has already changed our work on screens; next it walks into reality to change our lives. Summer 2026 may be the starting point of that process.

See you tomorrow.