The release of Kimi K3, a massive open-source AI model from Chinese startup Moonshot AI, triggered selloffs in US semiconductor stocks this week as investors reassessed the competitive landscape in artificial intelligence development.
Kimi K3 is a 2.8-trillion-parameter large language model that the company says is now the largest open-weight AI model ever released. It outperforms several leading US-developed models on key benchmarks, particularly in coding tasks and efficient hardware utilization, according to Moonshots internal testing and independent evaluations.
US chip stocks, particularly those tied to AI hardware demand, saw declines following the announcement. The market reaction reflects growing concerns that Chinese AI companies may be closing the gap with their American counterparts despite US export controls designed to limit Chinas access to advanced semiconductors.
The model was built using Moonshots proprietary Kimi Delta Attention mechanism and Attention Residuals architecture, which the company says allows it to achieve frontier-level performance with lower computational requirements. Full model weights are scheduled for release on July 27, making the technology available for developers worldwide to download, run and customize.
Kimi K3 supports a 1-million-token context window and includes native visual understanding capabilities. It competes directly with Anthropics Claude Fable 5 and OpenAIs GPT-5.6 Sol on several industry benchmarks, and in some tests involving AI hardware efficiency, it reportedly surpasses both.
The release signals a significant escalation in the global AI arms race, particularly in the open-source segment. While US companies like OpenAI and Anthropic have moved toward proprietary, API-access-only models, Chinese firms including Moonshot and DeepSeek have embraced open-weight releases, allowing developers to run the models on their own hardware without API fees or usage restrictions.
However, analysts caution that the chip stock selloff may be overblown. Demand for AI computing hardware remains strong globally, and US companies like Nvidia continue to report record revenue. The emergence of more efficient models could actually expand the AI market rather than diminish it, as lower computational costs enable broader adoption.
This article was adapted from BeInCrypto. Read the original here.
