September 18, 2026, (Inside AI) — Humanoid robot intelligence could reach a defining breakthrough by mid-2027, but the machines will not enter ordinary homes for at least eight more years. That timeline comes from Gao Yang, co-founder and chief scientist of Chinese embodied AI company Spirit AI, who spoke with reporters at the firm's Beijing offices this week.
The prediction places a specific date on a race that has mostly produced viral videos of machines sprinting, dancing, and performing backflips. Those hardware feats have impressed audiences. They have not yet produced robots that can reliably work in messy, unpredictable settings. Spirit AI argues the bottleneck now sits in software, not metal.
"The brain is indeed the weakest link in the complete robotics stack," Gao Yang, co-founder and chief scientist of Spirit AI, told reporters at its Beijing offices on Thursday.
Gao, who also serves as an assistant professor of robotics at Tsinghua University, compared the expected 2027 moment to OpenAI's GPT-3.0, the model that powered ChatGPT and triggered the current generative AI boom. He did not claim Spirit AI would produce that breakthrough itself. He described it as an industry-level shift that could arrive in roughly nine months.
Read: Unitree CEO Says Robots Are Nearing a ‘ChatGPT Moment’
Spirit AI's own robots have reached a 90% success rate on simple tasks inside structured living-room environments. That figure sounds strong until the setting is considered. A living room arranged for a demo is not the same as a home where furniture moves, pets interrupt, and objects appear in unexpected places.
"The next one to two years mark the initial window for industrial applications. Two years from now, we'll see robots deployed in commercial service settings doing simpler tasks. Entering homes is far harder than both," Gao said.
Data Bottleneck Slows Robot Brains
The company's approach to training separates it from many competitors. Spirit AI relies overwhelmingly on real-world data rather than virtual simulations, which are cheaper and faster to generate but imperfect for certain physical interactions.
"Simulators handle rigid bodies well, but flexible objects like deformable electric cables remain a problem," Gao said.
That limitation matters because homes and factories are full of flexible objects. Cables, cloth, food, and packaging do not behave like rigid blocks in a physics engine. Training a robot brain on simulated cable manipulation can produce a model that fails when the real cable bends in an unexpected way.
To collect real data, Spirit AI employs around 1,000 contractors across China. These workers wear sensor-equipped devices in households and on production lines, repeating motions such as opening refrigerators, unlocking safes, and cutting vegetables with knives. A reporter who visited the Beijing offices saw dozens of young people performing these repetitions inside a robot data training centre.
Other Chinese robot-training facilities often require operators to repeat a movement more than 50 times to capture one clean motion with the needed precision. Spirit AI found a different path. Gao said using what he called dirty data, with a wider variety of motions, helped its models improve faster than chasing perfect repetitions.
"Progress is extremely fast. When Spirit AI was founded, a robot could perform only one isolated task well, like pouring water or folding a piece of clothing," Gao said. "Today, robots operate across large spatial areas and execute continuous complex workflows."
Fine-motor skills remain difficult. Unscrewing a bottle cap and handling tasks the robot has never seen before still challenge the system, Gao said. Those are exactly the abilities that separate a useful household helper from an industrial tool.
Spirit AI has deployed tens of its own Moz1 wheeled humanoid robots on production lines at battery maker CATL and retailer JD.com, which is also an investor. The deployments are limited and structured, which aligns with Gao's view that industry comes first, commercial services second, and homes last.
The 300-person startup has raised over $670 million since its 2024 founding, making it one of China's most rapidly capitalised embodied intelligence firms. It is currently valued at 20 billion yuan ($2.9 billion). Gao declined to comment on any plans for an initial public offering.
Spirit AI's bet on real-world data carries financial weight. Human data collection is slower and more expensive than simulation. It also creates a workforce dependency that software-only competitors avoid. If the approach works, it could produce robot brains that transfer more reliably to unpredictable environments. If it stalls, rivals using hybrid simulation methods may catch up with lower costs.
China's humanoid robot sector has drawn heavy investment and government attention as part of a broader push into frontier manufacturing. Hardware supply chains have matured quickly, allowing firms to build impressive prototypes. The software layer, however, has not followed at the same pace. Gao's eight-year estimate for home deployment suggests the industry's most visible demos remain far from its most valuable market.
For now, Spirit AI's robots work in factories and warehouses, not living rooms. The company's timeline gives the sector a measurable target: a GPT-3.0-scale moment for robot brains by mid-2027, followed by years of work before those brains can safely handle a kitchen.
