Deutsche Bank strategists analyses China’s AI (Artificial intelligence) ecosystem, noting that Chinese models are rapidly catching up in capability while being priced near mid-tier US systems. They argue that open-weight Chinese models could accelerate global AI adoption but pressure United States (US) AI capex returns and margins. Recent Chinese releases have already weighed on AI and semiconductor stocks and may hasten commoditisation.
Open-weight Chinese models disrupt AI economics
“Geopolitical fears also interacted with fresh concerns around the AI trade, which meant that equities took a hit around the world. That was particularly clear for chip stocks, with the Philly semiconductor index down -9.97% last week (-1.63% Friday), marking its biggest weekly decline since the week of the Liberation Day tariff announcements last year.”
“The micro implications are as important as some Chinese models are being priced at levels broadly comparable to mid-tier US models (e.g. Anthropic Sonnet), despite performance that approaches higher-end systems, implying a materially lower cost-to-intelligence ratio.”
“A key differentiator is approach: Chinese models are increasingly released as open-weight systems, allowing developers and enterprises to download, modify and run them locally, whereas US leaders have largely pursued closed, proprietary models delivered via APIs.”
“The immediate market reaction—pressure on AI and semiconductor names—reflects a reassessment of whether the industry’s current capex trajectory is sustainable if similar performance can be delivered more cheaply.”
“At the macro level, this could encourage faster, wider and cheaper adoption of AI which will be more positive for productivity. However, it also raises the risk of a capex overcycle in the US if returns on AI infrastructure come under pressure, while also intensifying geopolitical fragmentation as competing technology stacks evolve.”
(This article was created with the help of an Artificial Intelligence tool and reviewed by an editor. Know more.)
Read the full article here
