A survey of 100 manufacturing leaders in the United States found that none reported a significant revenue increase or cost savings from artificial intelligence, even though 64% claimed efficiency gains. The Grant Thornton data, published this month, also showed that 48% of manufacturers remain stuck in pilot projects, against 34% across other industries. The zero result on financial returns exposes a widening gap between operational AI activity and measurable impact on the profit-and-loss statement.

In retail, AI’s influence is highly concentrated. Euromonitor International data indicates that AI referral traffic grew more than 300% last year, yet beauty and personal care captured roughly 45% of generative-AI referrals and consumer health another 28%, while categories such as snacks and pet care remained in the low single digits. The same research found that 78% of consumers have discovered a new brand through generative AI, but only 10 to 15% of US skin-care brands appear in AI-driven recommendations at all. Visibility is binary and earned through structured product data, not guaranteed by shelf position or media spend.

Consumer trust in AI remains fragile on both sides of the Atlantic. A survey of 1,250 American adults by Insuranceopedia found that 25% have turned to AI chatbots for medical advice because they cannot afford a doctor, with the figure rising to 53% among those aged 18 to 24. A study in the British Medical Journal last October concluded that AI chatbots provide problematic medical advice about half the time. In Italy, a GoStudent survey of 500 parents and children aged 11 to 17 revealed that 49% of teenagers who use AI for personal matters discuss schoolwork—the highest share in Europe—but only 8% believe a chatbot gives more useful answers than a real person, and 52% feel less capable of completing the same task alone.

Beijing has shifted its stance in the US-China chip stand-off, informing Alibaba, ByteDance and DeepSeek that they will be allowed to purchase limited numbers of Nvidia H200 processors, according to a source cited by the South China Morning Post. The move aims to temporarily ease a training bottleneck while China maintains its long-term goal of tech self-sufficiency. Analysts in Shanghai note that as AI models become a commodity, the contest is increasingly a race over electricity costs, pitting China’s renewable-energy build-out against America’s hydrocarbon-based grid. In Singapore, investors at the Reuters NEXT Asia conference said they are directing capital toward AI infrastructure such as data centres and liquid cooling, rather than front-end models, amid concerns about over-exuberance.

The first H200 shipments to the approved Chinese firms are expected in the coming months, a regulatory step that will test whether access to cutting-edge silicon accelerates China’s AI capabilities or merely narrows a temporary computing-power gap.