Zuckerberg acknowledged DeepSeek as a new competitor but stated it is not an immediate threat, emphasizing the ongoing importance of GPUs for AI tasks.

The development of Llama 4 is progressing well, with its mini version already completing pre-training, and it is anticipated to set new standards in open-source AI.

Despite the ambitious plans, analysts express concern that Meta's spending could lead to resource wastage, especially as competitors like DeepSeek are developing AI models more cost-effectively.

DeepSeek, a Chinese startup, has emerged as a competitive threat with its R1 large language model, which reportedly outperforms Meta's models while requiring significantly less funding and computing power.

Meta Platforms CEO Mark Zuckerberg has announced a substantial investment plan of $60 billion to $65 billion in artificial intelligence (AI) for 2025, emphasizing the need for a robust AI infrastructure to leverage the company's large user base.

In contrast to DeepSeek's efficient training on 2,048 Nvidia H800 GPUs, Meta's previous Llama 3 model required 16,384 H100 GPUs, showcasing a stark difference in resource utilization.

Following Zuckerberg's announcement, Meta's stock initially fell in premarket trading but later rose by 1.7% when the market opened, reflecting positive analyst sentiment.

Meta's financial performance has shown strength, with a reported revenue of $48.4 billion in the fourth quarter of 2024, surpassing estimates and indicating a 21% year-over-year increase.

However, the Reality Labs division, which focuses on the metaverse, reported a significant loss of $4.9 billion, highlighting financial challenges in that sector.

Investor scrutiny will be heightened regarding Meta's spending and cash flow in 2025, especially in light of previous costly investments in the metaverse.

Overall, while Meta aims for long-term leadership in AI, the immediate financial pressures from infrastructure investments may challenge its stock performance.