分类： technology

The global artificial intelligence boom has delivered historic financial results for the world’s two largest memory chip manufacturers, both based in South Korea — but growing investor anxiety over massive capital expansion plans and rising competition from China has sent their share prices sharply downward in recent trading.

Samsung Electronics, South Korea’s largest technology conglomerate, announced Thursday that it notched a record 89.5 trillion won (equal to $62 billion) in operating profit for the second quarter of the year, spanning April through June. This milestone result marks a more than 19-fold jump from the same period one year prior, and nearly all of the profit can be traced back to Samsung’s semiconductor division. The segment’s explosive growth has been fueled by skyrocketing demand for AI infrastructure, which has pushed up prices for memory chips and driven sharp increases in shipments of advanced high-bandwidth memory (HBM) chips — the core component that powers modern generative AI systems. This strong performance from semiconductors more than offset an operating loss reported in the company’s consumer division, which includes mobile devices, televisions, and home appliances, where higher component costs weighed on margins.

Samsung also hit an all-time high for quarterly revenue, which reached 171.5 trillion won ($119 billion). In a statement accompanying its earnings release, the company projected that robust demand for its memory products will continue through the second half of 2024, supported by ongoing global expansion of AI infrastructure and the growing mainstream adoption of agentic artificial intelligence. The firm added that demand for server-specific memory chips is expected to accelerate further, leaving the global market in a state of persistent undersupply.

Samsung’s blockbuster earnings report comes just one day after crosstown rival SK Hynix, the world’s second-largest memory chipmaker, released its own record-breaking second-quarter results. SK Hynix reported 60.5 trillion won ($42 billion) in quarterly revenue, a new all-time high, but its bottom-line profit fell short of the high expectations set by market analysts. The miss sent SK Hynix shares tumbling more than 9% on Monday, and Samsung’s stock has also declined this week amid broader market jitters.

The share price pullback comes as retail investors, who are the main driver of sharp volatility in South Korea’s equity market, have grown increasingly concerned about two key risk factors. First, both Samsung and SK Hynix have launched massive capital expenditure plans to expand semiconductor manufacturing capacity and build new data centers, part of South Korea’s broader national push to solidify its lead in the global AI arms race. But a growing number of industry analysts are questioning whether these multi-billion-dollar investments will ultimately deliver enough returns to justify the huge spending. Second, investors are growing wary of intensifying competition from Chinese chip manufacturers. Recent reports confirm that a Chinese state-owned enterprise has begun mass production of domestically developed immersion deep-ultraviolet (DUV) lithography machines, a critical piece of technology required for advanced chip manufacturing. Investor jitters were further amplified by the blockbuster, highly successful initial public offering of ChangXin Memory Technologies (CXMT), China’s leading homegrown memory chipmaker, which saw its shares surge on the Shanghai stock exchange in its debut.

The pullback in South Korean AI chip stocks has also aligned with a broader global trend: AI-sensitive equities have declined in recent sessions across major international markets, dragging down broader index values worldwide.

When customers hit the ‘purchase’ button for an online order, their goods are rarely processed by human hands alone these days. Increasingly, the first step in getting a purchase out the door begins with a compact silver autonomous mobile robot that glides under warehouse storage racks, lifts the entire shelf, and transports it directly to a worker waiting to prepare the order for shipping.

That seamless automation, developed by Chinese robotics firm Geek+, is already becoming ubiquitous in the UK logistics sector. Major British retail giants including Tesco, Asda, and Next have integrated Geek+’s robotic systems into their supply chains, with 2,000 of the company’s robots now operating across 10 UK warehouse sites through local partner MotionTech. This solidifies the UK as Geek+’s largest European market, at a moment when the country is grappling with persistent labour shortages and stagnant productivity growth.

Unlike traditional warehouse automation, which relies on rigid fixed conveyor belts and permanent infrastructure overhauls, Geek+’s flexible autonomous systems can be installed quickly using simple QR code floor markers and standard safety fencing. The company says automation of these repetitive inventory transport tasks boosts order picking speeds, maximizes storage capacity in hard-to-access warehouse spaces that are uneconomical for human workers to use, and drastically reduces order picking errors. For UK businesses already struggling to fill open frontline logistics roles, the technology offers an immediate solution to strained operations. ‘We have a very big shortage of labour in the UK at the moment. These technologies are really important to keep businesses thriving and meet demand,’ explained Barry Pemberton, MotionTech’s Account Director, noting that clients prioritize space-efficient, fast-deploying systems that work with existing warehouse footprints.

The growing adoption of Geek+’s technology in the UK lines up with broader economic analysis: for more than a decade, the UK has recorded anaemic productivity growth, and the OECD’s 2026 *SME Technology Adoption in the United Kingdom* report identifies wider adoption of robotics and automation as a critical step to reverse that trend. The organization notes that while UK firms have readily adopted mature digital technologies, the country lags behind other advanced economies in robotics uptake, a gap that has created major opportunities for global robotics suppliers. With one of Europe’s largest e-commerce and logistics sectors still in the early stages of automation, the UK market is primed for rapid growth in robotic deployment.

The expansion of Geek+ into Europe is part of a much larger, state-backed push by China to become a global leader in robotics, framed as a core component of developing what Chinese President Xi Jinping calls ‘new quality productive forces’. With China’s working-age population shrinking, policymakers view large-scale automation as essential to sustaining the country’s manufacturing competitiveness and boosting domestic productivity. Researchers point out that China’s robotics sector benefits from overlapping industrial ecosystems built for its booming electric vehicle industry: key components including batteries, electric motors, sensors, cameras, and semiconductors developed for EVs are now being repurposed for robotic systems, cutting production costs and accelerating development. Major Chinese EV manufacturers, including XPeng, have already launched humanoid robot development programs, with XPeng founder He Xiaopeng noting that ‘in the future, any car company will transform into a car and robotics company.’

Analysts say China’s robotics push could mirror the success of its global EV industry, leveraging domestic manufacturing expertise, dense local supply chains, and rapid iterative product development to capture an early global lead. But the sector’s global expansion faced a new setback recently, after the US banned imports of new advanced foreign-made robots over unsubstantiated national security claims. China’s embassy in Washington rejected the move, condemning the US for politicizing trade issues and imposing sanctions based on unfounded accusations.

Beyond its current wheeled warehouse robot offerings, Geek+ and other major Chinese robotics firms are investing heavily in next-generation automation, including humanoid robots designed to handle picking, packing, and other warehouse tasks. Geek+ says its long-term goal is to deliver fully end-to-end unmanned end-to-end warehouse solutions that automate every step of order processing, from inventory storage to final packing. Dozens of Chinese companies, including AgiBot and Unitree, have poured billions of dollars into humanoid robot development, and Chinese firms already account for the vast majority of global humanoid robot deployments to date.

Still, industry experts remain divided on the near-term impact of humanoid robots in warehouses. Developers argue that humanoids can fit into existing workplaces designed for human workers without costly infrastructure overhauls, but most agree that current humanoid technology lacks the dexterity to handle many complex repetitive tasks. Many industry observers, including Pemberton, frame humanoids as a complementary technology rather than a replacement for existing autonomous mobile robotic systems, noting it will likely be years before humanoids become commonplace in commercial logistics settings.

For the UK, where labour shortages and productivity pressures have created urgent demand for automation, the country’s warehouses have emerged as an early testing ground for what could be China’s next major global technology export. The Trades Union Congress, which represents nearly 6 million UK workers, has called for guardrails to ensure automation is used to improve working conditions rather than just cut labour costs, urging the government to require worker input on automation rollouts and mandate employer investment in worker retraining to avoid job displacement.

In a policy announcement delivered Tuesday, the Trump administration formally enacted a ban on new imports of foreign-produced advanced humanoid robots into the United States, citing what it called “unacceptable risks” to American national security. The restriction does not only target humanoid machines—it also extends to four-legged advanced robotic systems, a category of next-generation robotics where Chinese manufacturers hold the lion’s share of global production.

This trade action comes amid a intensifying global technology race between the United States and China, the world’s largest producer of humanoid robots, focused on advancing cutting-edge innovation in both robotics and artificial intelligence.

Alongside the robotics ban, the U.S. Federal Communications Commission (FCC) added another trade restriction: a ban on imports of specific foreign-manufactured power inverters. These components are critical to the operation of data centers and solar energy infrastructure, and the FCC asserted that they carry similar risks that could harm the U.S. economy. FCC Chair Brendan Carr framed the dual actions as part of the agency’s broader mandate to protect critical U.S. supply chains from potential threats.

Chinese technology firms have moved aggressively in recent years to scale development of humanoid robots, designing models that can be deployed across a wide range of use cases, from industrial manufacturing lines to domestic residential settings. Unlike many U.S. competitors—including high-profile developers such as Elon Musk’s Tesla robotics division and Boston Dynamics—Chinese companies have accelerated their go-to-market strategies, rolling out commercial products for both business clients and general consumers earlier than many of their American rivals.

As of Wednesday, the BBC has confirmed it has reached out to the Chinese Embassy in Washington, D.C. to request comment on the new restrictions, and no official response has been published yet.

As the September 24 meeting between U.S. President Donald Trump and Chinese President Xi Jinping in Washington approaches, discussions around mitigating risks posed by cutting-edge frontier artificial intelligence models and resolving long-running intellectual property disputes have emerged as the core items on the bilateral agenda.

This planned first official U.S.-China AI dialogue under the Trump administration, first disclosed by Reuters on July 21, will be led on the American side by Treasury Secretary Scott Bessent. Trump later confirmed the report on July 23, noting that the two leaders first broached the topic of AI collaboration and competition during his May visit to Beijing, where the pair met on May 14, 2026.

Chinese state observers and media outlets point out that the Trump administration’s push for the summit stems from growing anxiety over China’s rapid progress in the global AI race. This concern intensified after China secured new AI development partnerships with 28 countries, mostly from the Global South, at the 2026 World Artificial Intelligence Conference held in Shanghai this past July. Within Chinese policy and tech circles, there is a widespread consensus that further U.S. export restrictions and regulatory curbs on China’s AI sector will only accelerate domestic efforts to achieve technological self-sufficiency and build an independent, sustainable AI ecosystem.

The upcoming talks will center on regulating high-capacity AI models that carry far-reaching implications for global security and economics: these systems could rebalance global military power, enable devastating cyberattacks on critical national infrastructure, and cause widespread disruption to global labor markets. Trump has repeatedly framed AI as one of the most transformative technologies in human history, warning that the nation that leads the global AI race will hold unmatched strategic and economic advantage. “Whoever wins that race is probably going to win,” he stated in his confirmation of the September talks.

Washington’s push for dialogue comes against a backdrop of dramatic breakthroughs by Chinese AI developers over the past 12 months. Leading domestic firms including DeepSeek, Zhipu AI and Moonshot AI have launched a wave of low-cost, high-performance frontier models that match or even outperform U.S. industry leaders OpenAI’s ChatGPT and Anthropic’s Claude on key industry benchmarks.

A key strategy that has allowed Chinese firms to deliver strong results at a fraction of the cost of Western competitors is knowledge distillation, a process where new models learn pattern recognition and reasoning by studying outputs from existing Western AI systems, rather than building foundational capabilities from scratch. This approach eliminates the need for massive purchases of expensive cutting-edge AI chips. Chinese developers have also paired distillation with two efficiency-focused architectural innovations: mixture of experts (MoE), which routes individual queries only to the most relevant subset of the model’s parameters rather than activating the entire system, and sparse attention, which lets models focus only on the most contextually relevant sections of input text. Both techniques drastically cut computing requirements without sacrificing output accuracy.

In a July 26 commentary, the Global Times, a publication under China’s People’s Daily, noted that China’s open-source AI ecosystem has expanded dramatically in power and influence over the past two years, enough to trigger alarm among leading U.S. AI research labs. “Two years ago, names like Zhipu AI and Moonshot AI barely registered in the American tech press. Now their models are going toe-to-toe with those from Anthropic and OpenAI,” the commentary read. “The United States’ panic makes sense.”

The publication added that Washington has adjusted its strategy in the AI contest with China: rather than seeking to rapidly displace China from the global AI market, the U.S. is now focused on extending its current lead for as long as possible. “The failure to eliminate China doesn’t mean the contest is over. On the contrary, as knocking China out becomes less feasible, the AI race between the US and China simply shifts to more specific and hard-fought fronts,” the commentary said. “As long as China stays true to a development path suited to its own realities, it will earn the standing it deserves in shaping AI’s norms, rules and standards of access.”

That reference to “China’s realities” speaks to the long-running impact of U.S. export controls that have constrained China’s access to advanced chip manufacturing technology and high-end AI hardware, a topic rarely addressed directly by Chinese state media until recently. The U.S. first blocked Dutch chip equipment giant ASML from selling extreme ultraviolet (EUV) lithography machines to China in 2019, extended restrictions to cover cutting-edge deep ultraviolet (DUV) lithography systems in October 2023, and banned exports of high-end Nvidia AI graphics processing units (GPUs) to China starting in October 2022.

In response to these restrictions, Chinese firms initially adapted by sourcing second-hand mid-tier DUV machines for domestic AI chip production, routing imports of Nvidia chips through smuggling networks and shell companies, and training large models at overseas data centers in Southeast Asia. As controls tightened further, however, firms pivoted to the cost-effective distillation strategy that has now enabled their competitive breakthroughs.

For its part, the Trump administration has raised formal concerns over intellectual property practices in China’s AI sector. Speaking on July 21, Treasury Secretary Bessent said the administration has gathered evidence indicating that leading Chinese AI models draw heavily on foundational development work from U.S. systems, emphasizing that Washington does not tolerate what it frames as intellectual property theft.

Chinese analysts and policymakers have framed their country’s AI strategy as fundamentally different from the U.S. approach. Zhu Min, former deputy governor of the People’s Bank of China, outlined China’s priorities during a June World Economic Forum panel, noting that Beijing’s top goal is to integrate AI across China’s massive industrial economy, with a particular focus on manufacturing applications.

“China’s greatest advantage was never about building the largest model. It lies in having more use cases, more factories, more diverse industries and cost-conscious business owners,” Zhu explained. “The real winners will not be the AI model makers but those doing deployment, integration and process restructuring. The vendors installing factory systems, connecting data pipelines, retraining workers and collecting annual fees will pocket more than anyone.”

Zhu emphasized that manufacturing offers the clearest return on AI investment in China: every small improvement in production yield directly boosts profits, every avoided unplanned outage cuts avoidable losses, and every shortened delivery cycle increases business turnover. By contrast, he noted, U.S. investment capital overwhelmingly focuses on backing a small number of top-tier foundational model developers, betting that a handful of firms will capture the bulk of the global AI market. Chinese investment, meanwhile, flows disproportionately to application layers and industrial integration, targeting returns from the vast network of domestic factories, niche use cases and commercial orders.

Not all observers expect major breakthroughs from the September talks. Shandong-based political commentator Chen Xia argues that the U.S. is unlikely to make meaningful concessions on core issues China cares about, including easing export controls and expanding access to advanced technology. “The most likely outcome is a handful of toothless risk management clauses. On the issues China truly cares about, including access to technology and the easing of export controls, the US will not compromise on any of them,” Chen said. “The US only wants to walk away with maximum political gain at minimum cost.”

Chen also frames Trump’s outreach for AI talks as a tactical political move ahead of U.S. midterm elections in early November, where Trump’s Republican Party is fighting to defend its narrow majorities in both the U.S. Senate and House of Representatives. After the election, Chen argues, Washington’s underlying strategy of containing China’s AI development will become more explicit. He added that the U.S. is using the dialogue process to push for global AI governance rules that advance its own strategic interests and lock other nations into a disadvantaged position, while China seeks to secure a more equitable voice in shaping global AI rules.

As China mobilizes response and recovery efforts following the weekend landfall of Typhoon Noul, authorities are confronting a growing, man-made secondary hazard that is complicating emergency work: AI-fueled disinformation flooding social media platforms. In recent months, widespread storm and flood events across the country have been accompanied by a tidal wave of fabricated AI content, ranging from false depictions of mass casualties and severe flooding in unaffected regions to altered clips that distort the reality of official emergency response operations.

This tidal wave of false content has already triggered tangible harm. In multiple affected regions, AI-generated videos spreading baseless claims about widespread power outages have sparked panic buying of emergency essentials, straining local supply chains and adding unnecessary pressure to rescue operations. Bad actors have also weaponized real disaster events to amplify misleading content, capitalizing on public anxiety to drive engagement. When early July flooding in Guangxi’s Hengzhou City forced hundreds of captive snakes to escape a local breeding farm, bad actors quickly circulated fake images claiming crocodiles had been released into a public river, stoking widespread fear across local communities.

For Chinese authorities, who are working to manage mass evacuations and complex, life-saving rescue operations during active disasters, the top priority is preventing unnecessary public panic. What makes this challenge especially acute is the growing accessibility of AI technology – a sector China has invested billions of dollars into to advance its global tech leadership goals – which now gives bad actors low-cost, simple tools to deceive mass audiences.

“Many of these content creators exploit disaster news to grow their follower bases, knowing sensational, exaggerated content earns more likes and views. They deliberately fabricate information to grab public attention,” Huang Zhihua, Deputy Chief of online public security for the Zhejiang Provincial Police Department, told Chinese state media. In recent weeks, law enforcement has already carried out dozens of arrests and issued penalties ranging from administrative fines to pre-trial detention for creators of bogus disaster-related content, but experts warn an unprecedented national misinformation crisis is on the horizon.

In a recent article published by the Communist Party Central Party School’s newspaper, Professor Chen Bin noted that AI has dramatically lowered the technical barriers to creating convincing false content. “Individuals no longer require advanced technical skills. By entering a simple text prompt into an AI tool, they can rapidly generate highly realistic but entirely fabricated text, images, audio and video at scale,” Chen explained.

Speaking to the BBC, Professor Gao Fuping of the East China University of Political Science and Law outlined the wide range of motivations driving the spread of this content, from casual creators looking to make a joke without considering downstream harm to bad actors seeking to drive web traffic for profit, and even those intent on deliberately inciting public panic and social unrest. “Many users incorrectly assume the internet is an anonymous, unregulated space where they cannot be traced,” Gao added.

Policy experts have floated a range of potential solutions, from mandatory watermarking for all AI-generated content to more rigorous geographic and timestamp verification for disaster-related media, but the core challenge remains: detection technology still lags far behind AI content generation tools. “We are stuck in a constant cat-and-mouse game. We cannot assume technology alone will be able to identify all AI-generated content with a single click,” warned Professor Xu Xiaoke of Beijing Normal University.

China has positioned itself as a global leader in AI development, first naming the sector a “core driving force” for national economic and technological progress in 2017. Billions in public and private investment has fueled rapid growth, with state news agency Xinhua reporting the country is now home to more than 6,000 active AI companies. Professor Xu argues that in the face of growing misinformation risks, China must build a national “trusted information infrastructure” that includes centralized, official disaster early warning systems and verified data interfaces that news organizations and the public can access during crises.

However, a legacy of official underreporting and downplaying of disaster severity has created a public trust vacuum that bad actors have been quick to exploit. Earlier in July, local officials in the Nanning area instructed journalists to soften language describing a major breach of the Liulan Dam, referring to it instead as an “opening” or “gap.” The intentional downplaying of the risk had deadly consequences: water surging through the breached dam killed at least 26 people in downstream communities. This incident is not an outlier: in 2021, local officials in Henan Province were arrested on charges of concealing or underreporting 139 deaths from catastrophic flooding.

Despite these trust issues, law enforcement is urging the public to rely only on accredited official sources during emergencies. “When you encounter disaster-related information from an unknown source, approach it with skepticism and prioritize official announcements,” said Gao Kai of the Ningbo Public Security Bureau. He stressed that the internet is not a lawless space, noting “Public security authorities will continue to intensify efforts to crack down on illegal online activity.” Authorities have established a specialized task force to monitor social media, particularly short video platforms, to investigate and penalize fake content creators rapidly. Penalties can range from small fines to up to seven years of prison for cases where misinformation causes severe harm.

On July 23, China’s national Cyberspace Administration launched a nationwide crackdown on disaster-related fake internet content, ordering local regulators to prioritize cracking down on fabricated data, maliciously edited content, staged disaster scenes, misrepresented historical events, and impersonation of government officials. The regulator also ordered social media platforms to strengthen pre-publication content screening for disaster-related material.

Professor Gao Fuping argues that the most effective solution combines targeted public education with strict law enforcement, noting “We need to widely publicize real cases of people who have been penalized for creating fake content.” He added that “The harm caused by AI-generated fake content is ultimately the result of how people choose to use the technology – not the technology itself,” and argued existing Chinese laws are sufficient to address the problem when enforced consistently.

Other experts are less optimistic. Professor Chen pointed out that AI technology is evolving far faster than national legal frameworks, meaning new regulatory challenges are constantly emerging, leaving AI misinformation a persistent and growing challenge. Peter Pang, a Shanghai and Washington D.C.-based international lawyer, notes legal systems around the world are already seeing an explosion in cases related to AI-generated misinformation. He argues that AI app developers and social media platforms that host fake content should share legal liability alongside the individual creators of false content.

“Platforms are the first line of responsibility, because they are the enablers that allow this harmful content to reach mass audiences,” Pang told the BBC. He used a hypothetical example of a car crash that kills a pedestrian after a driver panicked while fleeing a danger falsely reported in a viral AI post: “The content creator might claim the crash is the driver’s problem, not theirs. But they created the false narrative that triggered the panic, and they should be held responsible for the resulting harm.”

Looking ahead, experts warn AI-generated misinformation will only grow in both volume and speed of spread. “Natural disasters, political elections, public health emergencies, armed conflicts, and key livelihood issues will all become high-risk targets for large-scale AI disinformation,” Professor Xu said. “Instead of waiting for false content to go viral before trying to contain it, governments and platforms need to put proactive prevention systems in place now.”