China's low-cost AI models are changing the global AI race. Here's why Silicon Valley is worried

The global artificial intelligence (AI) race is entering a new phase as Chinese companies rapidly gain ground with cheaper and customizable AI models, posing a fresh challenge to US technology giants such as OpenAI and Anthropic.

For years, the competition focused on building the most powerful AI models. But the latest developments suggest that affordability, flexibility and ease of deployment may now matter just as much as raw performance.

The shift has gained momentum after Chinese AI startup Moonshot AI unveiled its latest model, Kimi K3. The launch has intensified concerns in Silicon Valley that businesses may increasingly prefer low-cost "open-weight" models over premium AI systems that are significantly more expensive to use.

Chinese models dominate developer platforms

The trend was already becoming visible even before Moonshot's latest release. On OpenRouter, a marketplace where developers can access hundreds of AI models, Chinese systems now occupy the top five positions by weekly token usage.

More from Tech

These models are developed by companies including Tencent, Xiaomi, DeepSeek, MiniMax and Z.ai. They are all "open-weight" models, meaning users can download, customise and run them on their own infrastructure instead of relying entirely on cloud-based AI services.

This flexibility has made them attractive to businesses looking to reduce costs while retaining greater control over their AI systems.

Why businesses are switching

Industry experts say most companies do not require the most advanced AI model for everyday work.

Routine tasks such as coding assistance, document summarisation, customer support and data extraction can often be handled by cheaper models with little difference in performance.

According to some estimates, open-source or open-weight models could eventually handle around 95 per cent of enterprise AI queries, leaving only the most complex tasks to premium systems from companies such as OpenAI or Anthropic.

In an interview with Axios, Kong CEO Augusto Marietti said demand for open-weight models has surged in recent months because flagship AI models have become too expensive for many users.

Mozilla Chief Technology Officer Raffi Krikorian compared using frontier AI models for routine office work to "driving a Ferrari to Whole Foods," arguing that less expensive models are often more than capable while costing up to 50 times less.

US firms race to respond

The rapid progress of Chinese AI companies has also surprised industry observers. In May, Anthropic CEO Dario Amodei said China was still six to 12 months behind the US in advanced AI capabilities. However, within weeks, Moonshot released a model that rivals leading American systems on several key benchmarks, highlighting how quickly the gap is narrowing.

American companies are now responding by expanding their own open AI offerings. Thinking Machines, the startup founded by former OpenAI CTO Mira Murati, has launched an open-weight model focused on deep customisation.

Chipmaker Nvidia is also expanding its Nemotron family of open AI models, while SpaceXAI has open-sourced Grok Build, the software powering its AI coding assistant.

A changing business model for AI

The rise of open-weight AI could reshape the economics of the industry. If businesses can meet most of their needs using affordable models they control themselves, premium AI services may find it harder to justify their higher prices.

That could have wider consequences beyond Silicon Valley. Companies like OpenAI and Anthropic are preparing for major public listings, with valuations built on the assumption that frontier AI models will remain scarce and highly valuable.

As Chinese firms continue to improve performance while keeping prices low, the AI race is no longer just about building the smartest model—it is increasingly becoming a battle over who can deliver the best value.