However, beyond the polished showrooms of Beijing, another very real scene—this one in Hangzhou—still lingers in my mind. On the rooftop of Unitree Robotics’ office building, employees had built a robot testing ground filled with various terrain obstacles. There, I saw the very same robot dogs and humanoid robots that nimbly vaulted over barriers in promotional videos struggling to navigate complex, real-world environments without an engineer holding a remote control. Compared to the flawless performances of humanoid robots in staged settings, this stark contrast with reality left me with a sense of the embodied intelligence industry’s awkwardness on the eve of full commercialisation.

Fast forward to this year. In just one year, Unitree Robotics successfully went public, surrounded by a dazzling aura. Yet, at the recently concluded 2026 World Robot Conference in Beijing, Unitree’s founder, Wang Xingxing, poured cold water on an enthusiastic audience of practitioners and investors under the spotlight of the main forum. He bluntly pointed out that humanoid robots have not yet truly entered the stage of large-scale application, and their biggest shortcoming remains a lack of general-purpose capability. In his view, the true "ChatGPT moment" for embodied intelligence—when a robot can enter a completely unfamiliar home environment and autonomously complete about 80% of daily tasks using only voice commands—is still some distance away: 2 to 3 years in an ideal scenario, or perhaps 5 to 10 years if one is more pessimistic.

Wang’s remarks stand in sharp contrast to the numerous " social media trending topics" generated by two recent robotics conferences in Beijing. After the hype surrounding these events, I can’t help but wonder: how far are we really from the "ChatGPT moment" for embodied intelligence?

The 'Robot Olympics'

In late August, the 2026 World Robot Conference and the Second World Humanoid Robot Games were held back-to-back in Beijing. Thousands of domestic and foreign robots competed on the same stage, comprehensively showcasing the latest development achievements of the global robotics industry. Whether you were at the venue or just scrolling through short videos on your phone, you couldn't help but be infected by the feverish atmosphere, which felt like a sci-fi movie coming to life.

At the Second World Humanoid Robot Games, the humanoid robot "Tiangong Ultra" ran the 100-meter dash final in 8.64 seconds, nearly one second faster than the human world record held by Jamaican athlete Usain Bolt. In the standing high jump, a humanoid robot cleared a height of 2.88 meters, far exceeding the 0.96-meter best mark set at the inaugural games last year, and shattering the human high jump world record of 2.45 meters set by Cuban athlete Javier Sotomayor in 1993. In addition, humanoid robots made a strong impression and dominated trending topics with outstanding performances in the 400 meters, gymnastics, boxing, and other events. It is worth mentioning the globally viral "shy face-covering run" robot, which topped social media trending lists worldwide for its unique posture. This running style was not deliberately mimicked or designed by humans; rather, it was the optimal solution the robot autonomously evolved through data iteration in a virtual simulation environment.

The World Robot Conference held in Yizhuang also drew significant attention. This year’s conference unveiled 44 new technologies and solutions, attracting 557,000 attendees and inviting over 10,000 foreign guests to view the exhibition. The exhibition area exceeded 60,000 square meters, with 373 participating companies, a substantial year-on-year increase of 69%. In this feast of steel and code, over 3,000 exhibits were displayed, and an unprecedented 311 new products made their debut, setting records for the largest scale and the most new product launches in the conference's history. The topic garnered 730 million views on short video platforms, up 62% from the previous session.

Red-hot.

That was the visceral feeling surrounding the robotics industry at the end of August.

The 2026 Humanoid Robot Industry Development Report noted that in the first half of the year, China's humanoid robot shipments exceeded 40,000 units, capturing an astonishing 97% global market share.

Amidst these boiling statistics and social media revelry, it is easy to succumb to the illusion that we have already stepped half a foot into a new era of universal robot popularisation. However, when the tide of traffic gradually recedes, beyond recognising the rapid development of the robotics industry, we also need to have a more sober understanding and reflection on the hopes we pin on it: China's robotics industry is striving to move past mere flashy demonstrations and accelerating toward real production and daily life scenarios. But when will we truly be able to bring robots into our everyday lives?

From Leg-Focused Flash to Hand-Focused Precision

At robotics exhibitions over the past few years, companies were most eager to show off the capabilities of their robots’ lower limbs: backflips, resisting kicks, and parkour. These highly visually impactful actions were the "hard currency" for attracting venture capital. At this year’s conference, while combat and boxing remained popular crowd-pullers, the narrative logic of core companies has shifted; "operational metrics" have become the new standard of competition. Industry media even dubbed 2026 as the "Year One of Embodied Intelligence Tactile Sensation." In this inaugural year, electronic skin and tactile sensors debuted on a massive scale, and the speed of cost reduction in the supply chain was even more shocking. Tactile sensors, which once relied heavily on imports and sold for over 100,000 RMB (or 15,000 USD) per unit, now have a starting price of just 199 RMB (or 30 USD) for domestically produced alternatives. This cliff-like drop in hardware costs has made it possible to endow robots with "human-level touch."

Simultaneously, the operational capabilities of robots in real-world scenarios are being quantified. Xiaomi’s new generation of humanoid robots has quietly begun an "internship" at an actual part-loading station in Xiaomi's car factory. Data shows its success rate in bilateral operations has reached 98%, just 1 percentage point shy of a skilled human worker. Galbot's home service demonstration was equally impressive; when faced with interference deliberately created by staff moving items randomly on site, the robot could adjust dynamically in real-time by re-perceiving its environment, without relying on a pre-set script. In the logistics sector, Robotera's logistics sorting robot has achieved a practical sorting efficiency of 1,200 pieces per hour and has partnered with giants like China Post and SF Express to realise regular operation in over a dozen logistics centres across five provinces.

An official from the industry department of the Beijing Economic-Technological Development Area stated that as the focus of the embodied intelligence robotics industry shifts from technological breakthroughs to large-scale applications, Beijing Yizhuang will accelerate the construction of a national innovation hub for the embodied intelligence industry.

The invisible bottleneck

Although China's robotics industry is developing rapidly, industry leaders remain clear-eyed about the current challenges. Zhao Tongyang, founder of EngineAI, highlighted a cruel reality at the conference: in industrial and commercial scenarios, a 99% task success rate is still far from sufficient; it must be raised to 99.99%.

If hardware precision can be compensated for over time through supply chain iteration, then the true Achilles' heel restricting embodied intelligence from reaching its "ChatGPT moment" is invisible: "data." Zhao Qinping, an academician of the Chinese Academy of Engineering, along with several experts, clearly pointed out that the current "chokepoint" problem facing embodied intelligence technology, while existing at the model level, is a far deeper crisis concerning data.

Experts state that the core pain point hindering the continuous iteration of model systems is a severe scarcity of high-quality, real-world physical data. In virtual simulation software, the deformation process of objects, complex tactile feedback, and unpredictable environmental disturbances are extremely difficult to replicate with complete fidelity. Without a massive amount of high-quality data from the real physical world to support it, virtual simulation can never truly approximate reality, leading to a massive "reality gap" between simulation training and physical deployment.

Even more headache-inducing for companies is the severe lack of cross-platform data standards and reusability. Xie Shaofeng, Chairman of the OpenAtom Foundation and Chairman of the Ministry of Industry and Information Technology's Humanoid Robot and Embodied Intelligence Standardisation Technical Committee, noted that robots currently on the market operate in silos regarding sensor configurations, action space dimensions, task definition methods, and data structures. Consequently, the data collected by companies at a high cost is often limited to a single model of equipment and a single type of task. The entire industry is engaging in duplicated collection, duplicated labelling, and duplicated adaptation, causing massive resource waste.

It is precisely this industry-wide thirst for high-quality physical data that has birthed an entirely new business model. At this year’s conference, for the first time, data shed its status as an internal means of production and debuted en masse as a "commodity." According to a report by the Financial Times, at least 90 data collection and training centres for humanoid robots are currently in operation or under construction in China. Capital flows confirm this trend: the collection, cleaning, open-sourcing, and evaluation of data are being stripped from the internal processes of complete machine enterprises and turning into an independent business. For the first time in the industry chain, a specialised "data mining segment" has emerged.

Panning for technology gold

Today, the shakeup of the robotics industry has quietly begun. According to statistics, approximately 38 companies in the robotics industry chain are sprinting toward IPOs on the A-share STAR Market and ChiNext board. Yet, simultaneously, the clearing out of tail-end enterprises is also occurring; the significant drop in exhibition and rental prices is a clear signal that the market is returning to rationality. In such a phase where heat and cold coexist, it is easy for unrealistic bubbles to form, but it is also the easiest time to sift out the true gold that can withstand the test.

The development of China's robotics industry has transitioned from technological "exhibits" to market "commodities," and from isolated breakthroughs to a new stage of systematic advancement. Policy support, a manufacturing foundation, and a complete industrial chain are collectively driving technological iteration. According to Counterpoint's latest statistics, Agibot and Unitree together account for about 75% of global shipments, having already secured an international first-mover advantage. I have also interacted with robotics companies like UBTECH and Galaxea AI, and their products left a deep impression on me. From an international horizontal perspective, China's robotics industry is booming; however, from a vertical development perspective, it still requires calm, rational observation and continued technological breakthroughs.

From the "manual remote control" on that Hangzhou rooftop a year ago to the "autonomous operations" striving to break the 99% success rate on today’s exhibition stands, China's robotics industry is entering a sober phase after the initial excitement, returning to the most hardcore technological breakthrough stage of manufacturing. Before the true "ChatGPT moment" for embodied intelligence arrives, the Chinese robotics industry will inevitably undergo brutal technological and commercial trials. And who has the last laugh will be closely tied to the sprint in these crucial next few years.

Du Yubin is a CGTN journalist, producer, and chief editor. Having been stationed in Washington, D.C., and London for six years, he focuses on Sino-US and Sino-European relations, with over 16 years of experience in international communication and new media. This article solely represents the author's personal views.

UAE President Sheikh Mohamed bin Zayed Al Nahyan arrived in Berlin on September 9, 2026, beginning what UAE officials described as the first state visit to Germany by a president of the United Arab Emirates. His aircraft was escorted by German military jets after entering German airspace, while his arrival was marked by a 21-gun salute and guard of honour. His delegation included Deputy Prime Minister and Foreign Minister Sheikh Abdullah bin Zayed Al Nahyan, Industry and Advanced Technology Minister Dr Sultan Al Jaber, Foreign Trade Minister Dr Thani Al Zeyoudi, Minister of State for Defence Affairs Mohamed bin Mubarak Al Mazrouei, and other senior officials. On September 10, Mohamed bin Zayed was formally welcomed at Villa Borsig in Berlin by German President Frank-Walter Steinmeier. The UAE president signed the guestbook and referred to the longstanding relationship between the two countries, while expressing hope that bilateral ties would continue to deepen. The visit includes talks with Steinmeier and German Chancellor Friedrich Merz, with discussions focusing on economic and development cooperation as well as regional and international issues. The UAE and Germany have maintained diplomatic relations since 1972, while a strategic partnership agreement was signed in 2004.

UAE condemns Iranian missile attack on Jordan and backs Amman’s security response

The United Arab Emirates strongly condemned an Iranian missile attack on Jordan on September 9, describing it as a violation of Jordanian sovereignty and a threat to regional security. In an official statement, the UAE Ministry of Foreign Affairs said the attack represented a “flagrant violation of the sovereignty of the Hashemite Kingdom of Jordan” and threatened the country’s security and stability. The ministry reiterated what it called the UAE’s full solidarity with Jordan and said Abu Dhabi supported measures taken by Amman to safeguard its security and stability. The condemnation comes amid a broader period of heightened tension between Iran and Jordan. Jordan’s close security relationship with the United States has made it a target of Iranian missile and drone attacks during the recent regional conflict. The United States operates from multiple military installations in Jordan, including Muwaffaq Salti Air Base, which has played a role in US regional operations. The September 9 statement was not the first such condemnation from Abu Dhabi in 2026. The UAE also condemned Iranian attacks on Jordan on July 9 and again on July 29, repeatedly describing them as unacceptable violations of Jordanian sovereignty.

Mohammed bin Rashid tells government teams to accelerate delivery in new 2026–27 season

UAE Vice President, Prime Minister and Ruler of Dubai Sheikh Mohammed bin Rashid Al Maktoum has opened the government’s new 2026–27 work season with a call for faster implementation, stronger coordination and greater use of emerging technologies. At the annual New Government Season Meeting 2026–2027 on September 9, Mohammed bin Rashid met senior officials and government teams to review priorities, major projects and policy directions for the coming period. The meeting was attended by Sheikh Hamdan bin Mohammed bin Rashid Al Maktoum, Crown Prince of Dubai, Deputy Prime Minister and Minister of Defence, as well as Sheikh Ahmed bin Mohammed bin Rashid Al Maktoum, Second Deputy Ruler of Dubai. Mohammed bin Rashid praised government teams for improving operational efficiency, accelerating national priorities and developing public services, but told them to increase the pace of delivery. “With the start of every government season, I begin my work by meeting my team. We set out on a new season, carrying renewed energy, strong resolve and ambitions that match our country,” he said.

UAE moves agentic AI programme into second phase with FedAI government platform

The UAE is moving into the second phase of its ambitious Agentic AI Project , part of a wider plan to automate a large share of federal government operations using autonomous artificial-intelligence systems. The Ministerial Council for Artificial Intelligence and Development, chaired by Vice President and Deputy Prime Minister Sheikh Mansour bin Zayed Al Nahyan, reviewed the programme at a meeting at Qasr Al Watan in Abu Dhabi on September 9. The second phase includes the creation of FedAI, a national technical ecosystem designed to allow federal government entities to develop, adopt and operate agentic AI systems through a secure and flexible environment. The programme will also establish a customer experience laboratory to test agentic AI in public services and launch internal government competitions encouraging employees to develop AI-based solutions. The first phase focused on assessing government services, operations and tasks, identifying priority areas and building institutional capabilities. Ministries and federal agencies have since submitted services and processes that are expected to move into implementation during the next stage.

Labour violations fall 15% but illegal recruitment and unpaid wages remain major offences

Labour violations detected at private-sector companies in the UAE fell 15% in the first half of 2026, according to the Ministry of Human Resources and Emiratisation, though illegal employment and failures to pay salaries remain among the most common offences. The ministry said its inspection teams conducted about 212,000 visits during the first six months of the year using what it described as a “smart monitoring system” combining artificial intelligence, data analysis and risk indicators to identify establishments most likely to breach labour rules. The number of serious cases referred to the Public Prosecution fell 49%, from 962 during the first half of 2025 to 490 in the same period of 2026. The ministry said the decline indicated fewer severe violations requiring judicial intervention. However, inspections continued to identify several recurring breaches. These included employing or recruiting workers without the required permits, failing to pay salaries through the Wage Protection System, and submitting inaccurate information or documents to the ministry. Employing someone without the appropriate permit can result in fines of between Dh100,000 and Dh1 million under UAE labour legislation. Authorities can also refuse to issue or renew work permits, or cancel existing permits, when companies provide incorrect documentation or fail to comply with the Wage Protection System.

Artificial intelligence is moving beyond simple chatbots. The latest models are being built to reason through difficult problems, write and debug code, analyse information and carry out longer, more complicated tasks with less human guidance.

That rapid progress has also turned AI into a global competition.

An Artificial Analysis ranking shows that the United States and China are now the two main centres of advanced AI development. Every company in the top 10 shown on the graphic above comes from one of two countries: five American companies and five Chinese companies.

At the top is US-based Anthropic, whose Claude Fable 5 model recorded the highest score in the ranking. OpenAI's GPT-5.6 Sol ranks second, while China's Moonshot AI, with its Kimi K3 model, ranks third.

SpaceXAI's Grok 4.5 comes fourth, followed by China's Z AI with GLM-5.2. Meta, Alibaba, Google, MiniMax and DeepSeek make up the rest of the top 10.

Artificial Analysis uses its Intelligence Index to compare models across a range of tests covering areas such as reasoning, coding, science , knowledge and complex problem-solving. The results are combined to give an overall score.

That means one model may perform particularly well at coding while another is stronger at reasoning or other tasks. The ranking is therefore best understood as a comparison of overall performance rather than a definitive answer to which AI model is the best.

The leaderboard can also change quickly. AI companies are releasing new and updated models at a rapid pace, and a new release can change the rankings within days or weeks.