AGIBOT Hits 10,000 Humanoid Robots — Who Controls the Software?

AGIBOT hit 10,000 humanoid robots last week. Somewhere in Germany, a factory robot built by a Shanghai startup is moving packages across a warehouse floor. Somewhere in Singapore, another is navigating a hotel lobby. And somewhere in a corporate procurement office in Ohio, nobody asked the question that policymakers are only now starting to lose sleep over: who decides what software runs on it tomorrow?

The production milestone is real. AGIBOT, founded in February 2023 by former Huawei engineers Deng Taihua and Peng Zhihui, announced its 10,000th humanoid robot on March 30, 2026, becoming what the company called one of the first to reach that scale, according to PRNewswire. The ramp is genuinely striking. It took nearly two years to produce the first 1,000 units. One more year to reach 5,000. Three months to add the next 5,000. That compression is not a press release talking point. It reflects actual manufacturing momentum, and it is happening almost entirely in China.

The global numbers tell the same story. The humanoid robot market shipped approximately 13,318 units worldwide in 2025, a nearly 480 percent jump year over year, according to Omdia data cited by the South China Morning Post. AGIBOT shipped 5,168 of those, capturing roughly 39 percent of the global market. Its nearest competitor, Unitree Robotics of Hangzhou, shipped 5,500, good for 32 percent. Combined, the two Chinese companies account for roughly 70 percent of all humanoid shipments globally. Nearly 90 percent of every humanoid robot sold worldwide in 2025 was made in China. Tesla, Figure AI, and Agility Robotics each sold approximately 150 units, a rounding error by comparison.

AGIBOT's robots are no longer a Chinese domestic story. The company says deployments span Europe and North America, Japan, South Korea, Southeast Asia, and the Middle East, according to PRNewswire. Specific partnerships include Minth Group, a German automotive components supplier, and Singtel Enterprise in Singapore. AGIBOT's robots are operating in logistics, retail, hospitality, education, and industrial manufacturing workflows. In May 2025, the company's A2 humanoid obtained certification from China, the United States, and the European Union, clearing a regulatory hurdle that opened the door to Western market deployments.

The certification, the partnerships, and the 10,000-unit milestone combine into a sentence that should be getting more attention than it is: Chinese-made humanoid robots with remote firmware update capability are now deployed in warehouses, factories, and hospitality venues across the United States and Europe. AGIBOT's CTO Peng Zhihui framed the transition as one from seeking technical viability to delivering scalable commercial value. That framing is almost certainly accurate. It is also, depending on your perspective, either a compelling commercial story or a supply chain security problem that no government has figured out how to solve.

The parallel that security researchers are beginning to draw, and that some U.S. lawmakers have already said out loud, is TikTok. TikTok collects data on American users, and ByteDance, its Chinese parent company, can push updates to the app at any time. The proposed solution, the TikTok ban or forced divestiture, has spent two years tangled in courts, geopolitics, and constitutional questions. It is not close to resolved.

Now apply the same logic to a robot. AGIBOT's humanoids are networked devices. They navigate physical spaces using real-time sensor data. They operate alongside human workers. They are managed through software that can be updated remotely. A firmware push from Shanghai can change what they do, what they see, how they behave, and what data they transmit, without a customer in Ohio necessarily knowing it happened. Unlike a social media app, a misbehaving warehouse robot can cause physical harm.

U.S. Senators have noticed. A bipartisan group proposed the American Security Robotics Act to prohibit federal agencies from purchasing Chinese-made robots, citing supply chain and espionage risks, according to The Robot Report. The proposal is in early stages. The regulatory framework it would create does not yet exist at the federal procurement level for commercial buyers, the companies and warehouses that are actually deploying these systems at scale.

This is not a theoretical risk. It is a feature of how AGIBOT's robots are built and managed. The question is not whether remote update capability exists. It does. The question is whether there is any mechanism, technical, contractual, or regulatory, that gives a German factory manager or a U.S. warehouse operator meaningful assurance about what code is running on their floor next month, next year, or after a geopolitical incident changes the calculus in Shanghai.

Omdia's forecast suggests this problem will not stay small. According to Omdia, the analyst firm projects global humanoid robot shipments will reach 2.6 million units annually by 2035. Goldman Sachs estimates the market will be worth $38 billion by then. Morgan Stanley's mass adoption timeline points to the late 2030s. AGIBOT's production ramp and its acceleration curve suggest those forecasts may need to be revised upward. The question of who controls what those machines do, and how, is not a future problem. It is a present one.

AGIBOT's IPO plans add financial complexity to the geopolitical one. According to Reuters, the company is targeting a Hong Kong listing at a valuation of HK$40 billion to HK$50 billion (roughly $5.1 billion to $6.4 billion), with CICC and CITIC Securities as lead managers. Investors include Tencent, HongShan Capital (Sequoia China), BYD, LG Electronics, and Hillhouse Investment. A public listing would broaden the shareholder base and create new disclosure obligations. It would not, by itself, resolve the firmware governance question that procurement frameworks have not yet begun to address.

The warehouse robot moving packages in front of you is not thinking about any of this. It is optimized for throughput. Its manufacturer is moving fast, building real commercial deployments across real industries in real countries, and reporting strong demand signals from customers who are not running pilots. They are placing repeat orders. That part of the story is healthy and值得关注. The part that deserves more scrutiny is the one nobody in the procurement chain is currently equipped to answer: what exactly is going to change inside this machine, and who gets to say no when it does.