Despite recently reaching large-scale procurement agreements with NVIDIA and AMD, the social media giant Meta is still firmly advancing its "chip-making" initiative. According to a report by Bloomberg on March 5, Meta's Chief Financial Officer Susan Li publicly stated at the Morgan Stanley Technology Conference on Wednesday that the company is working on developing custom chips, with the ultimate goal of being able to independently support the training of future AI models.

Currently, although Meta is not a cloud service provider, it is one of the largest operators of data centers in the world. Susan Li revealed that Meta's self-developed chip strategy follows an "easy-to-hard" approach: the initial custom chips are mainly used for specific workloads such as sorting and recommendations, and have already been deployed on a large scale. However, she emphasized that Meta hopes to gradually expand the application scope over time, eventually developing processors capable of handling core AI model training tasks.

When discussing the chip procurement strategy, Susan Li showed a strong pragmatic attitude. She pointed out that Meta will seek the best balance between general-purpose chips from NVIDIA and AMD and its own custom chips based on the needs of different application scenarios. This "two legs walking" approach can ensure that the current explosive demand for AI business is met, while also reducing long-term reliance on a single supplier and optimizing computing power efficiency according to its own algorithms.

As Meta continues to invest heavily in the metaverse and generative AI, having an independent and controllable underlying hardware architecture has become a core strategy. This comprehensive evolution from software to hardware marks Meta's attempt to gain more bargaining power and technological leadership in the computing race of the AI era.

Key Points

  • 🚀 Custom Chip Development: Meta plans to expand the application of custom chips from simple recommendation algorithms to complex AI model training.

  • 🤝 Hybrid Procurement Strategy: While developing its own chips, Meta will continue to maintain deep cooperation with NVIDIA and AMD to ensure flexibility in the computing power supply chain.

  • 💡 Customized Computing Power: The CFO emphasized that custom chips have a significant advantage in processing specific workloads, which is key to Meta's cost optimization in data centers.