AI 2025: China’s Quantum Leap and the Global Race for Supremacy

By Technology Correspondent
April 10, 2025

China’s Origin Wukong quantum computer, which recently completed the world’s first fine-tuning of a billion-parameter AI model. (Photo: Anhui Quantum Computing Engineering Research Center)

The Stanford AI Index 2025 report, released this April, paints a picture of artificial intelligence as a field no longer in its infancy but sprinting toward adolescence. With breakthroughs in efficiency, geopolitical rivalries, and quantum computing, the report underscores a pivotal moment: AI is no longer a tool of the future—it is reshaping economies, industries, and global power dynamics today. While the U.S. retains its lead in innovation and investment, China’s rapid advances—from hyper-efficient models to quantum computing integration—signal a tectonic shift. As Andrew Swan of Man Group warns, “DeepSeek’s rise isn’t an endpoint—it’s the starting pistol for a new era of AI nationalism.”

The Efficiency Revolution: Smaller, Faster, Cheaper

The AI Index reveals a paradigm shift: bigger is no longer better. In 2022, achieving a 60% accuracy score on the MMLU benchmark required models like Google’s PaLM, a behemoth with 540 billion parameters. By 2024, Microsoft’s Phi-3-mini—a mere 3.8 billion parameters—matched this performance, slashing size by 142-fold. This trend toward compact, efficient models democratizes AI, enabling small businesses and startups to harness capabilities once reserved for tech giants.

Smaller Models

Equally striking is the collapse in inference costs. Querying a model with GPT-3.5-level performance cost $20 per million tokens in 2022; by late 2024, Google’s Gemini-1.5-Flash-8B reduced this to $0.07—a 280-fold drop. “Cost barriers are evaporating,” notes the report, enabling sectors from healthcare to logistics to adopt AI at scale.

Costs are Cheaper and Cheaper

Yet China’s most audacious play lies in quantum computing. Origin Quantum’s Wukong, a 72-qubit superconducting quantum computer, recently fine-tuned a billion-parameter AI model with 76% fewer parameters while boosting accuracy by 8.4%. As Dou Menghan, the project’s lead, explains, “It’s like equipping classical models with a quantum engine.” This hybrid approach—leveraging quantum parallelism to explore vast parameter combinations—could redefine AI’s computational limits.

Geopolitics: The U.S.-China Duel Intensifies

The U.S. produced 40 notable AI models in 2024, dwarfing China’s 15 and Europe’s three. Yet China’s models are closing the quality gap: performance differences on benchmarks like MMLU and HumanEval have shrunk from double-digit margins in 2023 to near parity. Beijing’s strategy—flooding the field with patents (China leads globally) and state-backed R&D—is paying dividends.

China is Catching Up

Investment tells a similar story. U.S. private AI funding hit $109 billion in 2024, dwarfing China’s $9.3 billion. But Beijing’s focus on quantum and semiconductor self-reliance, accelerated by Trump-era tariffs and export controls, reveals a long game. As the Global Times notes, Origin Wukong has already processed 350,000 tasks for users in 139 countries, spanning finance, biomedicine, and fluid dynamics.

Meanwhile, Europe risks irrelevance. Fragmented regulation and lukewarm investment ($4.5 billion in 2024) leave it trailing in the generative AI race. “The EU is debating ethics while China builds engines,” quips a Silicon Valley VC.

Corporate Adoption: From Experimentation to Integration

Businesses are all-in. In 2024, 78% of firms reported using AI—up from 55% in 2023—with generative tools driving the surge. DeepSeek’s R1 chatbot, launched in January, epitomizes this shift: its self-principled critique tuning (SPCT) and generative reward modeling (GRM) enable real-time, human-aligned responses at lower costs. By open-sourcing these frameworks, DeepSeek aims to undercut Western rivals like OpenAI and Meta.

Healthcare offers a glimpse of AI’s life-saving potential. The FDA approved 223 AI-enabled medical devices by 2023, from diagnostic algorithms to surgical robots. Yet risks loom: the AI Incident Database recorded 233 harmful events in 2024 (+56% YoY), including deepfake scams and chatbots linked to teen suicides.

Regulation: Patchwork Policies and Ethical Minefields

In the absence of U.S. federal legislation, states passed 131 AI laws in 2024—more than doubling 2023’s total. Topics range from deepfake bans to algorithmic bias audits. Conversely, the EU’s AI Act, mired in debate over “high-risk” classifications, struggles to keep pace.

China’s approach is characteristically top-down: strict censorship of AI outputs (DeepSeek’s models famously evade questions on Tiananmen) paired with aggressive R&D funding. Critics argue this stifles innovation; proponents counter that it prevents “toxic” outputs. Either way, the ethical calculus remains fraught: how to balance safety with the breakneck speed of progress?

Economic Realities: Optimism East, Skepticism West

Public sentiment diverges sharply. In China, 83% view AI as a net positive, driven by state narratives of technological ascendancy. The Asia-Pacific’s optimism extends to India’s AI education push and Singapore’s smart-city initiatives. In contrast, only 39% of Americans share this enthusiasm, with fears of job displacement and misinformation dominating discourse.

Trade wars further complicate the picture. Trump’s tariffs—now targeting Chinese EVs and semiconductors—have forced firms like Apple and Nvidia to rethink supply chains. Yet as Man Group’s Swan notes, “Asia’s tech hardware sector is poised to profit,” with AI-integrated devices spurring a $1 trillion upgrade cycle.

The Quantum Age Demands Collaboration

The Stanford report’s message is clear: AI’s next chapter will be written not by algorithms alone, but by geopolitics, ethics, and quantum leaps. For the West, the challenge is twofold: ramp up R&D funding to counter China’s state-backed juggernaut, and forge global standards to prevent a fragmented, dangerous AI landscape.

As Dou Menghan’s quantum-engineered models hint, the future belongs to those who blend classical and quantum, public and private, ambition and caution. The race is on—and complacency is not an option.

References

Stanford AI Index 2025 Report
China Daily: China’s Origin Wukong quantum computer breakthrough
Global Times: China’s quantum computing achievement
Euronews: DeepSeek’s AI advancements
Portfolio Adviser: Man Group’s analysis of China’s AI potential