As Meta vigorously develops artificial intelligence (AI) technology, the technology giant will spend billions of dollars to purchase Nvidia AI chips, which are the core of AI research and projects.
On Thursday local time, Meta CEO Mark Zuckerberg stated on social media that the company's AI future roadmap requires it to establish a "large-scale computing infrastructure.". Zuckerberg stated that by the end of 2024, infrastructure will include 350000 Nvidia H100 graphics cards.
Zuckerberg did not disclose how many H100 graphics cards Meta had purchased, but H100 was not released until the end of 2022 and supply was limited.
According to analysts at investment bank Raymond James, Nvidia's H100 is priced at $25000 to $30000, and if Meta pays at the lower end of the price range, this expenditure would be close to $9 billion.
In addition, Zuckerberg stated that if other GPUs are included, Meta's computing infrastructure will include "computing power equivalent to nearly 600000 H100 sheets.".
Last December, technology companies such as Meta, OpenAI, and Microsoft announced that they would use AMD's latest artificial intelligence chip, the Instinct MI300X, which directly benchmarks against the Nvidia H100.
Stock up weapons for AI battles
Meta needs these computing chips, and the company is committed to research on General Artificial Intelligence (AGI), which Zuckerberg said is the company's "long-term vision.". OpenAI and Google's DeepMind department are also studying AGI.
AGI is a future form of artificial intelligence that is comparable to human intelligence.
Meta Chief Scientist Yann LeCun also emphasized the importance of GPUs during a media event in San Francisco last month.
"If you think AGI will be very popular, you have to buy more GPUs," LeCun said at the time. For Nvidia CEO Huang Renxun, LeCun said, "A battle of artificial intelligence is underway, and he is providing weapons."
Meta stated in its Q3 2023 financial report that its total expenses for 2024 will be between $94 billion and $99 billion, partly due to computing power expansion. Zuckerberg stated in a conference call with analysts that in terms of investment focus, artificial intelligence will become Meta's largest investment area in 2024, whether in engineering or computing resources.
Zuckerberg stated on Thursday that Meta plans to "responsibly open source" its undeveloped "universal artificial intelligence," and the company has also adopted this approach in the Llama series of large language models. He also stated that Meta is currently training Llama 3 and bringing its Basic Artificial Intelligence Research Team (FAIR) and GenAI research team closer together.