50,000 units.

That's Tesla's production target for the third-generation Optimus in 2026. In late July, the California factory will officially start mass production. With over 10,000 components, alliance partners have already supplied harmonic reducers and joint modules; Asia Optical supplies high-resolution lenses (each robot equipped with 7-8 lenses for visual recognition and hand perception).

Musk says Optimus will contribute 80% of Tesla's future value. In 2025, Optimus produced 5,000 units; the 2026 target is 50,000 -- a 10x increase. Future annual capacity is planned at 10 million units.

This isn't a PPT, a demo video, or stage performance -- this is mass production. 2026 has been defined across the industry as the "year of 100K-unit mass production" for humanoid robots.

Capital Votes With Its Feet: 46 Billion Pours In During H1

Money doesn't lie. According to IT Orange statistics, in H1 2026, domestic embodied AI and robotics saw 288 funding events involving 226 companies, with disclosed funding exceeding 46 billion RMB.

Qianxun Intelligence completed four rounds totaling 4.5 billion RMB within three months, valuation breaking 20 billion; its self-developed Spirit v1.6 model overtook Nvidia Cosmos 3 to claim #1 globally on the RoboArena leaderboard. Founder Han Fengtao's judgment is clear: "The future ceiling of robot capabilities will be determined by models; the core competitive resource isn't compute but real-world data."

Wujie Dynamics is even more aggressive -- just one year after founding, it completed a $200M+ angel round, with a nearly $200M Pre-A close, nearly $100M in signed orders, and second-generation robot K15 already entering batch production. Founder Zhang Yufeng, formerly a Horizon Robotics executive, is pursuing a "latent-space world model" approach. German humanoid robot company Neura Robotics secured a C round of up to $1.4B, with Nvidia and Amazon participating.

HKU professor Li Hongyang's Yuance Future raised hundreds of millions in seed funding from ZhenFund, Gaorong, IDG, and 5Y Capital just two months after launch. Unitree Robotics went from STAR Market acceptance to approval in just 73 days, planning to raise 4.202 billion RMB with an expected Q3 listing.

Price Collapse: From Millions to Five Figures

The most direct change from mass production is price.

Just two years ago, the cost of a humanoid robot was still in the million-yuan range. In 2026, humanoid robot unit prices have dropped from millions to the 100K or even 10K yuan range. UBTECH's consumer-grade humanoid robot "U-World" surpassed 5,000 pre-orders; Agibot launched a 6-day livestream of "robots going to work in factories," demonstrating material handling/sorting/assembly on 3C electronics production lines.

Siemens showcased something more exciting: embodied AI robots sorting miscellaneous items without programming, declaring that "factories are moving from automation to self-adaptation." This means robots no longer need precise pre-programming to execute fixed tasks -- they can adapt to new work environments through observation and learning, just like human workers.

World models are becoming the key technical path. Nvidia launched the open-source full-modal physical AI world model Cosmos 3; Tencent released Hunyuan 3D World Model 2.0; Alibaba launched HappyOyster; Geely released WAM World Action Model; Xpeng launched X-World. Huawei Cloud released CloudRobo, the world's first end-to-end embodied AI development platform, in public beta on June 5. World models are seen as the key path to embodied AGI -- letting robots learn in virtual worlds, then transfer to physical execution.

The Last Mile From Showroom to Workshop

But a clear-eyed view is essential: mass production doesn't equal widespread adoption.

Currently, the main application scenarios for humanoid robots remain B2B industrial scenarios -- 3C manufacturing, automotive production lines, logistics and warehousing. The C-end consumer market (home helpers, companion robots) is at least 3-5 years from maturity. Hardware costs have come down, but reliability, safety, and versatility issues are far from fully resolved.

The Ministry of Industry and Information Technology and SASAC jointly launched a real-scenario training initiative, requiring the formation of over 100 high-value embodied AI application scenarios by end of 2026, driving 10,000-unit scale deployment. Beijing's embodied AI robot super-factory targets 500,000 units annual capacity by 2030.

From lab to factory, from factory to home -- the industrialization path for humanoid robots is becoming clearer step by step. 2026 is the inflection point from "can we build it" to "can we mass-produce it." The next question is: after mass production, can they actually solve problems?