The global artificial intelligence (AI) landscape is becoming increasingly competitive, with China making significant strides to close the gap with the United States.
According to Stanford University’s latest report on AI trends, Chinese AI capabilities are rapidly advancing, with models from Chinese companies performing comparably to their US counterparts on the LMSYS benchmark. While the US still leads in producing notable AI models, China’s dominance in publishing AI research papers and filing patents underscores its growing influence in the field.
Meanwhile, emerging players from regions such as the Middle East, Latin America, and Southeast Asia are also contributing to the global AI ecosystem.
China’sChina’s Growing AI Footprint
China’s progress in AI is evident in its quantitative output. The country publishes more research papers and files more patents related to artificial intelligence than the US. However, in terms of cutting-edge AI models, the US remains ahead, producing 40 frontier models compared to China’s 15 and Europe’s three. Despite this disparity, China’s advancements signal its commitment to becoming a global leader in AI innovation.
The Rise of Open-Weight Models
One of the most notable trends highlighted in Stanford’s report is the rise of “open-weight” AI models that can be freely downloaded and customised. Meta has been a frontrunner in this movement with its Llama model, first introduced in February 2023. Over time, Meta has continued to refine its offerings, recently releasing Llama 4. Other companies, such as France’s Mistral and DeepSeek, have also developed advanced open-weight models.
OpenAI, a major player in the industry, announced plans to release an open-source model, its first since GPT-2, in mid-2023. This shift towards openness has significantly narrowed the performance gap between open and closed models. In 2024, this gap decreased from 8% to just 1.7%. However, closed models still dominate the market, accounting for 60.7% of advanced systems.
Efficiency Gains and Cost Reductions
The AI industry has seen remarkable improvements in efficiency over the past year. Hardware used for training and running AI models has become 40% more efficient, leading to reduced costs for querying these systems. This enhanced efficiency has also enabled relatively powerful AI models to operate on personal devices.
Despite these efficiency gains, most developers continue to demand more computing power rather than less. Training state-of-the-art AI models now involves processing tens of trillions of tokens (data components like words) and utilising tens of billions of petaflops of computational power. However, a looming challenge is the potential exhaustion of internet-based training data by 2026–2032. This limitation is expected to accelerate the adoption of synthetic or AI-generated data for training purposes.
The Economic Impact of AI
Stanford’s report offers a comprehensive view of how AI is reshaping economies worldwide. Demand for professionals with machine learning expertise has surged as organisations increasingly integrate AI into their operations. Surveys indicate that a growing number of workers anticipate significant changes to their roles due to advancements in this technology.
Private investment in AI reached an all-time high of $150.8 billion in 2024, reflecting heightened interest from businesses and venture capitalists alike. Governments worldwide have also committed substantial resources to support AI development during this period. In parallel, legislation related to artificial intelligence has doubled in the United States since 2022, a testament to how rapidly policymakers are responding to this transformative technology.
Challenges and Risks
While AI offers immense innovation and economic growth potential, it also presents significant challenges. Incidents involving misuse or malfunctioning of AI systems have increased over the past year. This trend has spurred greater efforts toward improving safety and reliability through rigorous research and testing.
Moreover, as companies become more secretive about their methods for developing frontier models, academic research continues to thrive and improve in quality. This duality highlights both the opportunities and risks associated with widespread AI adoption.
The Pursuit of AGI: A Distant Horizon?
The concept of artificial general intelligence (AGI) – machines capable of performing any intellectual task that humans can do – remains a highly debated topic within the industry. Stanford’s report notes that some current AI systems already outperform humans on benchmarks testing specific skills like image classification, language comprehension, and mathematical reasoning. However, these achievements are primarily attributed to models being optimised for such tasks rather than representing true general intelligence.
Despite these advancements, AGI remains an aspirational goal rather than an imminent reality. Nonetheless, progress in specialised areas underscores how rapidly AI technology is evolving.
Conclusion: A Global Shift in AI Leadership
The race for dominance in artificial intelligence is no longer confined to a few key players; it has become a global endeavour. While the United States maintains its leadership position through groundbreaking innovations and frontier models, China’s rapid advancements signal a shift in global dynamics. Emerging contributions from other regions further highlight how accessible and widespread this technology has become.
As efficiency improves and open-weight models gain traction, the barriers to entry for developing sophisticated AI systems are lowering. However, challenges such as data limitations, ethical concerns, and safety risks must be addressed proactively to ensure this transformative technology benefits society.