Supported by multiple positive signals, Meta's stock price surged by over 15% after trading on Thursday (February 1st) Eastern Time, and is expected to reach a historic high upon opening.
According to a Meta report, Q4 revenue was $40.1 billion, with analysts expecting $39 billion and core "family app" profits nearly doubling to $21 billion.
Meta also announced its first-ever dividend and massive repurchase, offering investors a dividend of 50 cents per share and an additional $50 billion in repurchase authorization. Prior to this, the company repurchased a total of $20 billion in stocks in the 2023 fiscal year.
According to the latest Reuters report, Meta plans to deploy self-developed custom chips this year to support the development of its AI business. This chip will be produced on Meta's second-generation internal chip production line and will be put into production in 2024, working in conjunction with hundreds of thousands of purchased GPUs.
Flipping through Meta's AI roadmap, the company is not only integrating AI functionality into family applications such as Facebook, Instagram, and WhatsApp, but also developing ChatGPT like chatbots, large-scale language models stronger than Llama2, and Ray Ban smart glasses to accelerate the deployment of data centers.
The chips, infrastructure, and energy required to catch up with the AI trend have become a huge burden for technology companies, including Meta, to some extent offsetting the benefits brought by AI.
According to the latest report from market research firm Omdia Research, in 2023, Meta and Microsoft tied for first place with a purchase volume of 150000 H100 GPUs, both of which stated they would also purchase AMD's latest AI chips. This is still not enough. Zuckerberg previously revealed that by the end of 2024, Meta's infrastructure will include 350000 Nvidia H100 graphics cards.
Analysts at investment bank Raymond James estimate that Nvidia's H100 is priced at $25000 to $30000, and if Meta pays for the lower end of the price range, this expenditure would be close to $9 billion.
No wonder Tesla CEO Elon Musk compares the current AI competition to a high-risk poker game, warning that companies need to spend billions of dollars annually on AI hardware to remain competitive.
Undoubtedly, self-developed chips are expected to save Meta a significant amount of expenses. SemiAnalysis founder Dylan Patel stated that with Meta's operational scale, successfully deploying their own chips could potentially save billions of dollars in energy costs and billions of dollars in chip procurement costs annually.