“In the next decade, the humanoid robot industry will be larger than the automotive industry, and it will never be monopolized by one or two giants.” At the Global Sustainable Leaders Conference in Shanghai on October 16, Wang Chuang, a partner at Zhiyuan Robotics, electrified the audience with this statement.
As soon as he finished speaking, the capital circle was already flipping through contacts: Who will benefit from this wave of “fragmented hundreds of billions”? Is it NVIDIA? Is it Qualcomm? Could it be Rockchip, the “price butcher” that sells tablet chips to global white labels?
Zhiyuan’s latest mass-produced Lingxi X2 prominently features two RK3588 chips on its main control board, replacing the previously dominant Jetson Xavier. Even more exciting is that Rockchip has established a “Robotics Division” internally, planning to launch three generations of dedicated SoCs by 2026, aiming to become the “Qualcomm Snapdragon of the robotics world.”
1. Zhiyuan’s “Prototyping”: Two RK3588 Chips Cut Costs to 1/5
The BOM list for Lingxi X2 has been leaked:
– NVIDIA Jetson Orin NX (70TOPS) version: total machine cost 14,500 yuan, with the main control board costing 4,500 yuan;
– Rockchip RK3588 dual-chip version: total machine cost 9,800 yuan, with the main control board costing less than 900 yuan.
“Running motion control + perception fusion, the 6TOPS NPU of RK3588 is more than sufficient, with a power consumption of only 7W,” revealed the head of Zhiyuan’s supply chain. This year, the RK solution has already accounted for over 60% of the market share, and it is expected to rise to 80% next year.
2. Wang Chuang’s “Fragmentation” Prediction is Rockchip’s Home Turf
Wang Chuang outlined three major industry characteristics at the conference:
1. Extremely fragmented scenarios—elderly care, shopping guidance, patrolling, off-road, explosion-proof, each segment has sales in the hundreds of thousands;
2. Huge regional differences—Africa needs dust protection, Northern Europe needs cold resistance, the Middle East needs high-temperature resistance;
3. Extremely high cost sensitivity—machines priced at tens of thousands are needed to achieve scale.
These three points perfectly align with Rockchip’s “white label gene” comfort zone:
– Directly transferring the supply chain that achieved millions of shipments for tablets/OTT boxes to robotics, diluting R&D costs;
– Turn-key delivery with public boards + SDK, allowing customers to produce complete machines in just six weeks;
– Covering price ranges from $30 to $80, making “thousand-yuan robots” possible.
3. Technical Roadmap: Three Generations of Chips, Each More Powerful than the Last
Li Kai, General Manager of Rockchip’s Robotics Division, disclosed the roadmap for the first time:
– Q2 2025: Release RK3668, integrating MCU island + CAN-FD + EtherCAT, a single-chip solution for motion control, with power consumption < 3W;
– Q1 2026: Launch RK3599, with NPU increased to 20TOPS, supporting Transformer edge-side inference, and collaborating with universities to open-source a 7B multimodal robot model;
– Q1 2027: Trial production of RK3900, using 4nm + Chiplet technology, allowing CPU/GPU/NPU to be combined like Lego, with customers paying to unlock computing power based on scenarios, completely breaking the “one chip rules them all” model.
4. Ecological Strategy:
With a $1 model store + white label alliance, Rockchip replicates the “turnkey” model from the mobile phone era, announcing:
– RKNN Toolkit 3.0 is fully open-sourced, with a quantization toolchain that compresses PyTorch models to INT4 with one click;
– Launching the “RockRobot Store,” with edge small models licensed for $1/month, featuring over 200 algorithms including offline voice, VLA commands, and fall detection;
– Forming a “White Label Robot Alliance” with 50 solution providers in Shenzhen, promising to help customers go from idea to mass production within 12 weeks, aiming to ship 1 million complete machines by 2026.
5. How to Estimate Valuation
The global annual output value of the automotive sector is about $3 trillion, and Wang Chuang asserts that robotics will be “larger.” Even with a conservative estimate of $1 trillion, the core chip’s share is 5%, equating to $50 billion. If Rockchip captures just 10% of that market, it corresponds to $5 billion in revenue, giving a 10x PS ratio, resulting in a market value of $50 billion—20 times its current valuation. Some institutions have already begun to reassess using the model of “the MediaTek of the robotics world.”
Conclusion:
While everyone is looking up at NVIDIA’s 1000TOPS, Rockchip has taken the lead in bringing robots into the “thousand-yuan era” with two $30 RK3588 chips. Wang Chuang’s fragmentation prediction has become the sharpest moat for this “white label king.” The next trillion-dollar industry may not belong to the chip with the highest computing power, but to the player who minimizes costs, breaks down scenarios into the smallest parts, and delivers the fastest. Rockchip has already secured a ticket for Zhiyuan’s latest mass-produced Lingxi X2.
Special Statement: Much of the information above is sourced from the internet, and the author only cites it, which does not represent the correctness and accuracy of the information. We also published “Thoughts on the Potential Risks of Rockchip’s Sustainable Development” on October 12.