Unitree Robotics founder Wang Xingxing now predicts that, as soon as 2028, humanoid robots can complete 80% of assigned tasks in the factory or home after receiving verbal orders, Bloomberg reported on 23 October 2026. Bloomberg wrote that the figure is “an arbitrary threshold he defines as the “aha” moment for practical robotic applications.”
In plain words, the 80% is a success rate: of the jobs a person asks a robot to do out loud, it would finish eight in ten. It is Wang’s own yardstick, not an industry benchmark. Bloomberg described the forecast as “a reversal for a Chinese entrepreneur who once predicted intelligent machines would take as long as decade to go mainstream.”
Business Pill · PHYSICAL AI
Unitree founder Wang Xingxing’s forecast: as long as a decade before, as soon as 2028 now, for humanoids completing 80% of assigned tasks after verbal orders, his own threshold. Source: Bloomberg, 23 October 2026.
A short Business Engineer explainer on physical AI, AI that works in the real world. It teaches the general idea only and says nothing about Unitree or the interview in this piece.
The key insight: As we read it, the forecast comes with its own yardstick: Bloomberg reported that Wang defines the 80% himself, as the “aha” moment for practical robotic applications, so the date and the measure come from the same person.
The forecast. As soon as 2028, the machines can complete 80% of assigned tasks in the factory or home after receiving verbal orders, according to Bloomberg’s account of the interview.
The change. Bloomberg reported that Wang has “grown noticeably more bullish on his own industry since his company went public in the summer.” Bloomberg’s AI-generated summary of its article says he had predicted it could take 10 years for humanoid robots to be deployed on a massive scale, and changed his belief due to advances in AI models like OpenAI’s GPT-6.
The examples. A summary posted on X by AI commentator Rohan Paul says Wang credits GPT-6 with letting Unitree humanoids “paint, rinse a palette between colors and place Go stones precisely.” We did not find those examples in the parts of Bloomberg’s article we could read.
Who he is. Bloomberg described Wang, 36, as “one of the country’s most celebrated tech leaders”, who created the nation’s biggest humanoid manufacturer a decade ago.
Who
Forecast or finding
Where we reported it
Wang Xingxing, Unitree founder, now
Humanoids complete 80% of assigned tasks in the factory or home after verbal orders, as soon as 2028
Bloomberg, 23 October 2026 (this piece)
Wang Xingxing, earlier view
Intelligent machines would take as long as a decade to go mainstream
Bloomberg, 23 October 2026 (this piece)
Elon Musk, Tesla
Optimus at a run rate of 20,000 units a week in a year or “maybe a little longer”
Our Optimus piece, citing Axios, 21 October 2026
SenseNova research team
One long simulated robot task with GPT-6 Astra used about nine million tokens and nearly two hours
Sources: Bloomberg (23 Oct 2026, free opening read); our own earlier pieces as linked. The forecasts measure different things and are not comparable with each other.
Unitree founder Wang Xingxing’s forecast, as Bloomberg reported: he once predicted intelligent machines would take as long as a decade to go mainstream; he now predicts humanoids can complete 80% of assigned tasks after verbal orders as soon as 2028. The 80% is his own threshold. Source: Bloomberg, 23 October 2026.
What Was Posted on X
Unitree founder Wang Xingxing expects that by 2028, humanoids told to do something will succeed 80% of the time.
He credits recent models such as OpenAIโs GPT-6, which he says lets Unitree humanoids paint, rinse a palette between colors and place Go stones precisely.
AI commentator Rohan Paul wrote on X that Wang expects that by 2028 humanoids told to do something will succeed 80% of the time, and that Wang credits recent models such as <a href=”https://fourweekmba.com/how-does-openai-make-money/”>OpenAI</a>’s GPT-6 with letting Unitree humanoids paint, rinse a palette between colors and place Go stones precisely. The painting and Go examples in this piece come from his post.
Robots building robots is the next big loop
Asked what infrastructure OpenAI is betting on now, Sam Altman says its robots that build more robots, and data centers that help build more data centers, chip factories and power plants, so compute can grow far faster.
says this loop is almost as big a deal as RSI, AI improving its own research.
—-
Full video on “Stanford Graduate School of Business” YouTubechannel, (link in comment)
Rohan Paul summarised on X a clip from Sam Altman’s Stanford Graduate School of Business interview, in which, per the post, Altman says OpenAI is betting on robots that build more robots and data centers that help build more data centers, chip factories and power plants, and calls that loop almost as big a deal as AI improving its own research.
Why the Forecast Matters
A humanoid maker’s founder setting a date and a success rate gives investors and customers a number to hold the industry to. Bloomberg reported that Wang chose the 80% threshold himself, and the parts of the interview we read do not say how it would be measured or across which tasks.
The change of view is the second number in the story. Bloomberg’s AI summary puts his earlier estimate at 10 years for deployment on a massive scale; Bloomberg’s own article frames the new view as expecting a breakthrough in humanoids “in as little as two years”, with 2028 as the soonest date.
Connecting the Coverage
Our Unitree IPO piece reported in August that, according to Caixin, Unitree was pricing a Shanghai STAR Market IPO, with $235 million of 2025 revenue at about 60% gross margins in that piece’s key numbers. What’s new: the company is public, and its founder has put a date on what he defines as the “aha” moment for practical robotic applications.
Our GPT-6 Astra piece reported that a SenseNova research paper found one long simulated robot task used about nine million tokens and nearly two hours, and that the paper’s own answer to whether the model has achieved embodied intelligence was no, not yet. Our Optimus piece reported that Elon Musk said on 21 October he expects Tesla’s humanoid to reach a run rate of 20,000 units a week in a year or “maybe a little longer”, according to Axios.
Our physical AI funding piece reported that four physical AI startups raised $240 million in rounds announced on 22 October, by our arithmetic.
The Structural Read
As we read it, the change is in the founder’s timeline, not in a reported product result: Bloomberg wrote that he once predicted intelligent machines would take as long as a decade to go mainstream, and now predicts as soon as 2028.
As we read it, the credit Wang gives to AI models is the link to the rest of our coverage: Bloomberg’s AI-generated summary says he changed his belief due to advances in AI models like OpenAI’s GPT-6, while our GPT-6 Astra piece reported a paper that answered no, not yet, on embodied intelligence.
As we read it, the setting matters too: Bloomberg reported that Wang grew more bullish since Unitree went public in the summer, so the forecast comes from the founder of a listed company.
Bloomberg, 23 October 2026
“an arbitrary threshold he defines as the โahaโ moment for practical robotic applications”
Three Implications
FOR ROBOT BUYERS Bloomberg reported that the forecast covers assigned tasks in the factory or home after verbal orders, which is the setting buyers can test against.
FOR INVESTORS Bloomberg described Wang as more bullish since the company went public, and we did not find a revenue or shipment figure tied to the forecast in what we read.
FOR AI MODEL MAKERS Bloomberg’s AI-generated summary says Wang credits advances in AI models like OpenAI’s GPT-6 for his change of view.
The framework says “The Product Overhang Doctrine is the bet that you can build into the gap before it closes”, shipping features for capability that does not yet exist.
As we read it, Wang’s new date is a version of that bet for hardware: Bloomberg’s AI summary ties his change of view to newer AI models. The framework was published in June 2026 and is about software; it does not discuss Unitree or humanoids.
What Is Not Established
We read the free opening of Bloomberg’s article, its AI-generated takeaways as shown in a screenshot posted on X, and Rohan Paul’s summary. The painting, palette and Go examples come from Rohan Paul’s summary, not from the parts of Bloomberg we read. Wang’s 80% is his own definition; we did not find a named benchmark for it in what we read. We did not contact Unitree. Nothing here is investment advice.
Business Engineer Framework
The Product Overhang Doctrine
A Business Engineer framework on building products for model capability that has not arrived yet.
Unitree founder Wang Xingxing has moved his forecast from as long as a decade to as soon as 2028 for humanoids that complete 80% of assigned tasks after verbal orders, a threshold he defines himself, Bloomberg reported. Bloomberg’s AI summary ties the change to newer AI models such as OpenAI’s GPT-6.
A note on sourcing. This piece is based on the free opening of Bloomberg’s 23 October 2026 article, Bloomberg’s AI-generated takeaways as shown in a screenshot posted on X, a post on X by Rohan Paul and our own earlier pieces. We did not contact Unitree. Nothing here is investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
