open_in_newOriginal article
Jack Ma-backed Ant touts AI breakthrough built on Chinese chips

Jack Ma-backed Ant touts AI breakthrough built on Chinese chips

  • 24.03.2025 03:55
  • businesstimes.com.sg
  • Keywords: Ant Group, Alibaba Group Holding

Ant Group, backed by Jack Ma, achieved an AI breakthrough using Chinese semiconductors to develop cost-effective models that rival Nvidia's performance, showcasing China's advancements in AI while reducing reliance on US technology.

Alphabet NewsAlphabet ProductsBABAsentiment_satisfied

Estimated market influence

Ant Group

Positivesentiment_satisfied
Analyst rating: N/A

Ant Group is making progress in AI and reducing costs using Chinese chips, which could challenge Nvidia's dominance.

Alibaba Group Holding

Alibaba Group Holding

Positivesentiment_satisfied
Analyst rating: Strong buy

Alibaba provides chips to Ant Group for AI model training, contributing to the advancement of local AI technology.

Context

Analysis of Ant Group's AI Breakthrough Using Chinese Chips

Key Facts and Data Points

  • Cost Reduction: Ant Group claims its AI model training costs are reduced by 20% using Chinese-made semiconductors.
  • Chips Used: Domestic chips from Alibaba Group Holding and Huawei were utilized, alongside alternatives like AMD.
  • Mixture of Experts (MoE) Technique: This approach divides tasks into smaller data sets for efficiency, similar to Google and DeepSeek.
  • Performance Comparison: Ant's models reportedly match or outperform Nvidia's H800 chips in certain benchmarks.
  • Training Costs:
    • High-performance hardware: ~6.35 million yuan (S$1.17 million) per trillion tokens.
    • Optimized approach with lower-spec hardware: Reduced to ~5.1 million yuan.
  • Model Parameters:
    • Ling-Lite: 16.8 billion parameters.
    • Ling-Plus: 290 billion parameters.
    • Comparison: ChatGPT's GPT-4.5 has 1.8 trillion parameters (MIT Technology Review estimate).
  • Applications: Ant plans to leverage its AI models for healthcare and finance, including through acquisitions like Haodf.com and apps like Zhixiaobao and Maxiaocai.

Market Implications

  • Shift in Semiconductor Dependency: Chinese companies are reducing reliance on US-based Nvidia chips, potentially reshaping the global semiconductor market.
  • Cost Efficiency: Lower training costs could democratize AI adoption, particularly benefiting smaller firms and fostering competition.
  • Global AI Landscape: The move highlights a growing competitive landscape between Chinese and US AI technologies.

Competitive Dynamics

  • Strategic Investments: Companies are investing in alternative chip technologies and optimizing AI training methods to reduce dependency on premium GPUs.
  • Open Source Models: Ant's decision to make Ling models open-source underscores the importance of collaboration and accessibility in AI development.

Long-Term Effects and Regulatory Considerations

  • Fragmentation Risks: Increased reliance on local solutions may lead to a fragmented global AI ecosystem, with regional tech hubs emerging.
  • Regulatory Adaptation: Regulators will need to adapt as local AI solutions become more prevalent, potentially leading to new policies or standards.

Strategic Considerations

  • Efficiency Focus: Companies must prioritize cost-effective AI training methods and hardware optimization to stay competitive.
  • Real-World Applications: As highlighted by Shengshang Tech's CTO, practical applications are key to proving the value of advanced AI models.