Meta has already invested millions of dollars into the development of open-source AI models in the Llama family. However, training large language models (LLMs) requires growing financial resources. As The Information reports, Meta has turned to Microsoft and Amazon for assistance—an unexpected move given Meta’s own significant computing capacity.
Partnerships and Strategic Incentives
According to four sources, Meta offered “incentives” to these companies in exchange for support. These included the potential to influence the direction of future Llama AI models. What Microsoft and Amazon stand to gain is unclear, especially considering both are developing their own AI projects, primarily commercial in nature, while Meta’s core revenue comes from advertising. Since Llama is open-source, sponsoring a potential competitor might not seem profitable at first glance.
In addition to serious talks with Microsoft and Amazon, negotiations also took place with Databricks, IBM, Oracle, and at least one Middle Eastern investor. Meta reportedly discussed forming a Llama Consortium in early 2025. While still in the planning stages, the conversations largely centered around acquiring funding, server access, and other resources to reduce the cost of model training.
Exploring Additional Forms of Support
Meta also considered other types of collaboration. These included having senior company representatives attend partner events or offering more transparency about the training process of Llama models. The outcomes of these discussions have not yet been made public, but we’ll keep you updated as more information becomes available. It may also become clear which companies Meta ends up partnering with based on its future presence at various industry events.
Notably, in November 2024, it was revealed that training the then-new Llama-4 model required more than 100,000 NVIDIA H100 accelerators, reminds NIX Solutions. More recently, in March, NVIDIA introduced Llama Nemotron AI models, which offer adjustable “intelligence” and are based on Meta’s existing technologies.
The development of Llama models remains an ongoing effort, and the question of long-term partnerships and support remains open. Whether Meta’s strategy will strengthen its position in the open-source AI space is yet to be seen.