Spirit AI Says Robot Brains May Break Through in 2027 While Homes Wait a Decade
Spirit AI cofounder Gao Yang told Reuters that embodied intelligence may reach a GPT 3 class moment by mid 2027, but data scarcity could keep household humanoids eight years away.
Classy aggregates publicly available material; we did not conduct a private interview.
What changed
On 18 September 2026, Reuters published an on the record conversation with Gao Yang, cofounder and chief scientist of Beijing embodied AI startup Spirit AI. Gao said the "brain" remains the weakest link in the humanoid stack, forecast a possible industry breakthrough around mid 2027 comparable to the leap from GPT 3 to ChatGPT, and estimated at least eight years before robots become useful in homes. Spirit AI has raised about $670 million in two years and is valued near $2.9 billion, according to Reuters.
Gao also said Spirit AI has deployed tens of Moz1 wheeled humanoids on production lines at battery maker CATL and retailer JD.com, with JD.com among its investors. Reuters did not specify exact line tasks or autonomy levels for those units.
Why it matters
Factory automation buyers and humanoid investors often treat hardware scaling and model scaling as the same clock. Gao's timeline splits them: industrial pilots may widen in one to two years, commercial service roles in roughly two years, but unstructured home environments stay far out because training data for dexterity and social interaction remains scarce. That framing shifts capex from demo theater toward data pipelines, teleoperation logs, and structured deployment feedback.
For applied AI teams, the quote also reframes 2027 as a capability threshold for interpreting natural language and attempting broad physical tasks, not a guarantee of reliable fine motor work such as opening bottle caps.
Who is affected
Manufacturing executives evaluating humanoids on battery and logistics lines, robotics R&D leads building vision language action stacks, and venture investors pricing embodied AI rounds against OpenAI style language model timelines.
What to do next
If you are piloting factory humanoids, ask vendors for task level success rates, trial counts, and failure modes on your SKU mix before expanding POs. Treat Reuters deployment counts as presence signals, not proof of autonomous throughput.
What to watch
Spirit AI or CATL publishing audited throughput, mean time between intervention, and per task success rates on live PACK or warehouse lines. A mid 2027 benchmark with open protocol details would test Gao's GPT 3 analogy.
Sources
- Primary. Reuters, Founder of Chinese startup Spirit AI says robot brains set for 2027 breakthrough (18 September 2026). On the record quotes from Gao Yang, funding figures, deployment counts at CATL and JD.com, and timeline estimates.
- Secondary. Humanoid Guide, Spirit AI targets 2027 humanoid brain breakthrough (20 September 2026). Context on the limited task threshold Gao described and Reuters reporting gaps on test protocols.