HONG KONG —
The newly unveiled Chinese AI system Kimi K3 has sparked considerable interest in the U.S. technology community and, shortly after its introduction, has temporarily halted new sign‑ups because demand has quickly outpaced available capacity.
This situation underscores the hurdles Chinese AI systems encounter while catering to an expanding domestic and international user base amid an intensifying U.S.-China technology rivalry.
Chinese open‑source, cost‑effective advanced AI models — such as DeepSeek’s recently released V4 — are gaining rapid global traction. In early 2025, DeepSeek’s breakthrough helped position China as a formidable counterpart to the United States, where many cutting‑edge AI models remain openly accessible.
“Kimi K3 has garnered far more attention than anticipated,” Moonshot AI, a Beijing‑based startup, noted in a post on X late Sunday. “In the past 48 hours, demand has approached the limits of our existing infrastructure.”
Moonshot announced that it will prioritize current subscribers and temporarily suspend new enrollments. “We are expanding capacity as quickly as possible and will reinstate new subscription slots in phases,” the company added, also sharing the same statement on Chinese social platforms.
K3’s public launch late last week, intended to compete with more sophisticated models from Anthropic, OpenAI and others, has unsettled U.S. rivals. With 2.8 trillion parameters, it is currently recognized as the largest open‑source AI model in the world.
“New model releases typically generate overwhelming interest, which can stress existing compute resources,” said Lian Jye Su, chief analyst at research firm Omdia. “This highlights that Moonshot AI lacks enough compute chips to meet the current surge in demand.”
Su added that the primary cause was likely Moonshot’s failure to fully anticipate the surge in K3’s popularity. He noted that K3 requires substantial compute resources, making allocation both complex and costly.
In terms of front‑end coding performance — a benchmark used to evaluate AI capabilities — Kimi K3 ranked at the top among major advanced models on the Arena evaluation platform following its public release.
K3’s introduction has also placed pressure on the stocks of major U.S. technology firms, which worry that more economical Chinese AI solutions could erode pricing power and demand for domestic services, even as U.S. restrictions limit China’s access to the most advanced chips and technologies.
K3 ranks among the newest generation of Chinese AI models that have elicited strong worldwide reactions.
Over the weekend, Alibaba previewed its latest Qwen3.8 Max model, which the company claims contains 2.4 trillion parameters and ranks as one of the world’s most powerful AI systems, trailing only Anthropic’s Fable 5. Earlier last month, Chinese startup Zhipu (also known as Z.ai) launched its GLM‑5.2 model, which quickly gained global adoption.