US-China Competition Intensifies
The global AI race is increasingly shaped by two competing strengths: US companies lead many frontier systems, while Chinese developers are gaining ground with cheaper and more customizable models. Recent launches, funding and security disputes show that competition is widening beyond model performance. Access to chips, computing power, data and talent is becoming equally important.
DeepSeek Seeks USD 11.93 Billion Funding
China’s DeepSeek is attracting fresh capital as competition in artificial intelligence intensifies. The startup is close to raising at least 80 billion yuan, or about USD 11.93 billion, in a funding round backed by major investors including Tencent. Reuters reported that the claims came from Bloomberg sources, and Reuters could not independently verify the report.
US Startup Reflection Launches Beam
The US is responding with new open-weight models designed to challenge China’s growing presence. Nvidia-backed Reflection AI launched Beam on October 5. The company says the model has 501 billion total parameters, with 23 billion activated for each task. Beam is positioned as a lower-cost competitor for coding and agentic workloads against leading Chinese models.
Chinese Open Models Gain Ground
Chinese companies are gaining attention for open-weight AI, where developers can access model weights and customize systems for their own needs. Reuters reported that firms such as DeepSeek, Moonshot AI and Alibaba have produced models that compete with Western systems on several tasks. Their lower costs and flexibility have become important factors in the global competition.
US Accuses Chinese Firms of Model Distillation
US officials have accused Chinese AI companies of using distillation to copy capabilities from American models. The September allegations named six Chinese firms, including DeepSeek, Moonshot AI and Alibaba. Officials said the companies used outputs from US-developed systems to train their own models. Chinese firms were accused of seeking faster development while lowering training costs.
AI Model Weights Become a Security Concern
The competition is also moving into AI security. On October 1, US lawmaker Ro Khanna asked OpenAI, Anthropic, Google, Meta and SpaceXAI for information on attempts to obtain their AI model weights. These weights contain the parameters that determine how models operate. The request reflects concern that theft could weaken the US technological advantage over China.
AI Safety Joins the Global Race
Washington and Beijing are also facing pressure to manage risks from increasingly capable AI systems. The two countries prepared bilateral AI safety talks in September, while US officials continued to focus on competition and technology security. The race for more powerful models is therefore developing alongside a parallel debate over safeguards, national security and international cooperation.