The Token Tsunami: Global AI Usage Hits 28.9 Trillion as Chinese Models Eclipse U.S. Giants

As of May 25, 2026, the digital world is witnessing a seismic shift that many analysts predicted, but few believed would materialize so swiftly. According to the latest data from Futu Niuniu, the total global weekly usage volume of Large Language Models (LLMs) has reached a staggering 28.9 trillion tokens. This marks the fifth consecutive week of rising global demand, but the real story lies in the geographic distribution of this growth: domestic Chinese AI models have outpaced their U.S. counterparts in usage volume for the fourth straight week, with DeepSeek-V4-Flash claiming the top spot.

The Rise of Chinese Efficiency

The dominance of DeepSeek-V4-Flash is no accident. In a market that is now maturing, users and enterprises are moving away from the novelty of "general capabilities" toward the raw necessity of speed, cost-effectiveness, and reliability. DeepSeek’s model, utilizing a next-generation Mixture-of-Experts (MoE) architecture, has managed to deliver a performance-to-price ratio that U.S. models, such as OpenAI’s GPT-5 or Anthropic’s Claude 4, struggle to match at scale. The Chinese strategy has pivoted toward the "democratization of tokens," slashing entry barriers for businesses looking to integrate AI into their daily operations.

This surge also reflects a deeper evolution within China’s tech ecosystem. While the U.S. remains the leader in cutting-edge research and "frontier" models, China has mastered the art of application and scaling. DeepSeek-V4-Flash has become the de facto tool for millions of developers across Asia and Europe, offering the ultra-low latency required for real-time applications—from customer service bots to automated code synthesis.

Economic Implications and the "Token Economy"

The 28.9 trillion tokens per week figure is not just a metric; it is the new currency of the global economy. This sustained growth over five weeks suggests that AI has ceased to be an experimental tool and has become foundational infrastructure, akin to electricity or cloud computing. The economic activity generated by these tokens translates into billions of dollars in productivity gains, but also places immense demands on energy resources and data centers.

• Infrastructure Costs: The insatiable need for compute is driving a new arms race in AI silicon, with NVIDIA and Huawei battling for dominance in Southeast Asian data centers.
• Subscription vs. Pay-as-you-go: The plummeting price of tokens, driven by Chinese supply, is forcing U.S. firms to re-evaluate their business models, shifting toward high-value specialized services.
• Data Geopolitics: The preference for Chinese models in many emerging markets creates a new landscape for digital sovereignty and the influence of global AI standards.

The U.S. Response and the Innovation Challenge

Washington and Silicon Valley are watching with growing concern. Despite U.S. models still being regarded as "smarter" in complex reasoning benchmarks, the market appears to be voting based on volume and accessibility. OpenAI and Google are attempting to counter with their own "Flash" models, but China’s supply chain integration and proximity to hardware manufacturing give it a distinct advantage in speed-to-market for new iterations.

According to analysts, if this trend continues, 2026 will be remembered as the year China became the world’s "intelligence factory." The challenge for the West is now to prove that superior quality can overcome raw quantity and lower costs. However, in the economy of scale, quantity often begets its own quality, as data from trillions of interactions fuels the next generation of model refinements.

"We are no longer in the era of 'who builds the best AI,' but in the era of 'who powers the world with the most AI,'" says a leading tech industry executive.

In conclusion, the rise to 28.9 trillion tokens is a reminder that the AI revolution is relentless. The shift of power toward the East, led by DeepSeek, is reshaping global balances and raising new questions about the future of technological hegemony. The coming months will be critical in determining whether Silicon Valley can reclaim its lost ground or if the "Token Economy" will henceforth carry a Chinese watermark.