Honor Device Co. walked into Mobile World Congress 2026 in Barcelona and did something none of its Chinese smartphone rivals had done before: it put a humanoid robot on stage and let it dance. The unnamed, matte-black biped — blue LED strip blazing across its chest — backflipped, waved at journalists, and shook hands, halting only occasionally enough to confirm it wasn't under remote control. Alongside it was the company's "Robot Phone," a smartphone with a 200-megapixel gimbal arm that nods, shakes its head, and dances to music. Together, the two devices are the clearest signal yet that China's smartphone industry has decided its next battleground is not the screen — it's the body.
Why a Smartphone Maker Is Building Humanoids
Honor isn't an obvious candidate to pioneer humanoid robotics. The Shenzhen-based manufacturer — spun off from Huawei in 2020 and backed by a consortium of Chinese state investors — built its business on affordable, feature-rich smartphones. But the logic behind its pivot is straightforward: the smartphone market is saturated, and artificial intelligence is becoming too powerful to stay inside a glass rectangle.
The company has set up a multibillion-dollar initiative to expand into new industries, with robotics and agentic AI at its core. Like rivals Xiaomi, Oppo, and Vivo, Honor is building its own AI services that will be integrated across its device ecosystem. The difference is that Honor moved first to put that AI into a body — a humanoid capable of delivering items, providing companionship, and assisting in retail and home service environments.
"We've combined cutting-edge robotics with the ultimate mobile experience," Honor said in its MWC teaser. The result on stage at Barcelona was more proof-of-concept than polished product — specs are thin, pricing is unannounced, and commercial availability remains unclear. But the intent is unambiguous: Honor wants to be the company that makes physical AI as common as the smartphone.
The Robot Phone: A Bridge Between Screens and Bodies
Honor's Robot Phone is the more immediately concrete product. At its core, it's a smartphone — but one with a motorized, folding camera arm packed into the rear housing. That arm contains a 200-megapixel sensor inside what Honor claims is the smallest 4DoF (four degrees of freedom) gimbal system in the industry. The result is a device that can track subjects autonomously, stabilize video in three axes, nod in conversation, and move responsively to its environment.
The phone isn't ready for commercial release yet. Honor confirmed a launch window of the second half of 2026, with initial availability in China only. But its significance goes beyond the hardware specs. The Robot Phone represents Honor's design philosophy for the physical AI era: intelligence embedded in objects that interact with the world, not just display information about it.
Journalists who handled a working unit at MWC confirmed the arm unfolds, holds AI-enabled conversations, and folds back in — basic but functional. One journalist noted the robot feature was "asleep" much of the time in the demo area, and the humanoid waved back at people only after being prompted multiple times. Rough edges, yes. But the underlying capability is there.
China's Hardware Advantage: Why This Wave Is Different
To understand why Honor's humanoid debut matters, it helps to understand the broader context: China already dominates the humanoid robot market in ways most Western observers have not yet internalized.
According to research firm Omdia, the global humanoid robot market experienced 500% revenue growth in 2025. Of the approximately 13,317 units shipped that year, Chinese manufacturers accounted for the vast majority. The top five global humanoid robot companies by 2025 shipments — AgiBot (39% market share, 5,168 units), Unitree Robotics (4,200 units), UBTECH, Leju Robotics, and Engine AI — are all Chinese firms. U.S. players like Tesla and Figure AI trail dramatically in unit volume, though they claim technical advantages in AI sophistication.
Unitree alone shipped roughly 36 times more humanoid units in 2025 than U.S. rivals Figure AI and Tesla combined, according to TechCrunch analysis. The reason isn't magic — it's supply chain. China has spent two decades building the world's most capable hardware manufacturing ecosystem, first for consumer electronics, then for electric vehicles, and now for robotics. Sensors, actuators, motors, batteries, and structural components that cost thousands of dollars from Western suppliers can be sourced for hundreds from Chinese manufacturers.
"China has a more robust hardware supply chain — much of it built up through the EV sector, from sensors to batteries — and the world's strongest manufacturing base, allowing companies to iterate far faster than Western competitors," Selina Xu, China and AI policy lead at the office of former Google CEO Eric Schmidt, told TechCrunch.
That supply chain advantage is now available to Honor, Xiaomi, Oppo, and Vivo — all of whom have spent decades optimizing complex device manufacturing. When these companies enter robotics, they don't start from scratch. They start with fully operational logistics networks, world-class manufacturing plants, and the cost structures to undercut anyone else in the market.
The Competitive Landscape: AgiBot, Unitree, and the New Entrants
China's humanoid ecosystem is already crowded at the top. AgiBot leads by shipment volume with its general-purpose humanoid, followed closely by Unitree — whose G1 robot, priced at $16,000, has become a benchmark for cost-effective humanoid capability. Unitree's robots have demonstrated remarkable durability: models have completed 130,000+ steps at −47.4°C in endurance testing, performed kung fu and acrobatics in front of China's 1.3 billion viewers at the Spring Festival Gala, and shown up in viral demos that put Boston Dynamics' expensive Atlas platform to shame on cost-per-capability metrics.
Galbot, whose G1 humanoid also appeared at the Spring Festival Gala, has raised over $300 million in fresh funding, reportedly pushing its valuation to $3 billion. Fourier Intelligence, UBTech, Engine AI, and Leju Robotics round out a competitive field that is iterating at software startup speed with hardware manufacturing at automotive scale.
Into this field now comes a second wave: the smartphone manufacturers. Honor is first to show hardware at a major global conference, but Xiaomi, Oppo, and Vivo are building agentic AI services with robotics integration as an explicit roadmap item. Xiaomi, which began humanoid robot research in 2022 with the CyberOne prototype, is further along than the others — the company demonstrated advanced dexterous manipulation in 2024 lab footage and is expected to announce commercial plans within 2026.
Huawei, though it has been effectively barred from Western markets by U.S. export controls, is developing AI models specifically designed for humanoid robotic control. The company's deep chip design expertise — constrained in mobile GPUs but not in specialized AI inference hardware — positions it as a potential supplier of the inference stack that will run inside China's next generation of service robots.
From Demo-Driven to Operations-Driven
The narrative around humanoid robots has long been dominated by demos — robots performing tasks in controlled environments that look impressive until you ask how they'd handle the messy unpredictability of the real world. That narrative is shifting.
"The biggest shift recently has been from 'demo-driven excitement' to 'operations-driven adoption,'" Yuli Zhao, chief strategy officer at Galbot, told TechCrunch. "More customers are asking: Can the robot run stably in real environments and actually take work off people's plates?"
That transition is visible in deployment data. Chinese humanoid robots are now being piloted in logistics warehouses, retail environments, automotive manufacturing lines, and elder care facilities. The deployments are small — dozens to hundreds of units — but they represent real operational use rather than one-off showcases. The feedback loops from these deployments are feeding rapid iteration: Chinese companies are releasing new robot models quarterly, incorporating lessons from real-world operation at a pace that Western competitors, with longer development cycles and higher labor costs, cannot match.
Honor's service robot is explicitly targeted at this operational market. Shopping assistance, item delivery, companionship, and household help are all applications where the economics of a service robot make sense — particularly in China's aging society, where a shrinking labor force and rapidly growing elderly population create structural demand for automation that doesn't depend on macroeconomic speculation.
The West's Counter-Argument: Intelligence Over Volume
American and European humanoid makers aren't conceding the market. Their argument is that hardware volume is a lagging indicator — what matters is the quality of the AI running the robot.
Tesla's Optimus program, which entered mass production at its Fremont factory in early 2026, leverages the same neural network architecture powering Full Self-Driving. The company argues that a robot trained on the world's largest real-world driving dataset — plus factory floor telemetry from its own internal deployments — will develop generalization capabilities that volume-focused Chinese manufacturers can't replicate by shipping more units of simpler designs.
Figure AI, backed by OpenAI, NVIDIA, Amazon, and Microsoft, is pursuing a similar logic: that the winning humanoid will be the one with the best world model, not the cheapest bill of materials. The company has raised hundreds of millions and is targeting commercial pilots in 2026, with BMW as a disclosed manufacturing partner.
The tension between these two strategies — Chinese hardware scale versus Western AI sophistication — will define the humanoid market for the next five years. History suggests that manufacturing scale tends to win in mass-market hardware categories, while software and ecosystem differentiation can sustain premium positions in enterprise AI. Humanoid robots are both — making the outcome genuinely uncertain.
What Omdia's 2035 Projection Actually Means
Omdia projects annual humanoid shipments reaching 2.6 million units by 2035 — roughly 200 times the 2025 base. That projection deserves scrutiny. The 2025 figures themselves mix confirmed commercial sales with demo units and pilot deployments; the industry is too early-stage for clean data.
What's not uncertain is the directional trajectory. Humanoid robots are moving from research labs into operational deployments. The cost curve is declining — Unitree's G1 at $16,000 was unthinkable two years ago. Multimodal AI is giving robots the sensory processing needed to handle real environments. And the labor economics driving demand — aging populations, shrinking workforces, rising wages — are structural, not cyclical.
Within that trajectory, Honor's MWC debut is both a data point and a signal. A data point: China's smartphone manufacturers are now formally entering the humanoid market, bringing massive manufacturing scale, established supply chains, and hundreds of millions of existing customers. A signal: the era of humanoid robots as curiosities is ending. What comes next is the era of humanoid robots as products — and the companies that win will be the ones that can ship them at scale, at price points the market can absorb.
Honor hasn't proven it can do that yet. Its robot is unnamed, unpriced, and available only in China for the foreseeable future. But the demonstration matters. On a stage in Barcelona, at the world's largest mobile technology conference, a Chinese smartphone company showed a humanoid robot dancing — and the audience mostly wasn't surprised. That normalization is itself the story.