According to the statement made by Mark Zuckerberg Meta, It will have 1.3 million AI-focused GPUs by the end of the year.
Meta, which already has very large server systems, will continue its investments in this field without slowing down in 2025. CEO stated that Meta will spend between 60 billion and 80 billion dollars to invest in different areas in 2025. Mark Zuckerberg1.3 million GPUs, which is targeted to be reached by the end of the year (Most of it will come from Nvidia) will greatly accelerate its productive artificial systems and work. Last week, it was brought to the agenda that Microsoft plans to spend $80 billion for new artificial intelligence-focused data centers in 2025.
YOU MAY BE INTERESTED IN
Meta made a splash in recent months with its artificial intelligence model named Llama 3.3 70B. In the statement made about the new model, the following was stated: “The Llama 3.3 70B offers the performance of our 405B model, but is our newest model that is easier to operate and more affordable. “By leveraging the latest advances in post-training techniques, including online optimizations, 70B significantly improves performance at a low cost, making it a more accessible option for the entire open source community.” Before this, Llama models were made available to government institutions and companies working in the field of national security, which made a splash.
More than a dozen private sector companies and government agencies partner with the U.S. government on national security, including Amazon Web Services, Oracle and Microsoft, as well as defense companies such as Palantir and Lockheed Martin. You will now be able to easily access Llama artificial intelligence models signed by Meta.. It is stated that this step will accelerate the work carried out and provide great benefits in many areas. For example, Oracle uses Llama to synthesize aircraft maintenance documents. This allows technicians to diagnose problems more quickly and accurately, speed up repair time, and return critical aircraft to service.
It is reported that Meta also provides similar access to Llama to governments and private companies working with the state in the UK, Canada, Australia and New Zealand. Meta CEO Mark Zuckerbergwithin the scope of the last quarter negotiations, the company said, “llama 4A new open source LLM, or large language model family, called ” On a cluster with over 100k Nvidia H100 AI GPUs or that it trains in a larger cluster than any system that has come up about what others have done. The company, which has invested billions of dollars in servers to provide more power to artificial intelligence systems, aims to make Llama 4 available in the first quarter of 2025.