Meta Platforms’ (NASDAQ:META) Llama family of AI models could be worth between $10 and $20 billion, which would represent more than a 1% upside to its current market capitalization.
Such valuation, according to Jefferies, reflects its leading share in open-source deployments, access to data, compute resources, and Meta's extensive consumer user base.
Analysts see this valuation as comparable to companies like xAI and Anthropic but lower than OpenAI's $86 billion due to Llama's lack of monetization.
The analysts also draw parallels with Red Hat, which was acquired by IBM (NYSE:IBM) for $34 billion in 2019 at 10x its FY19 revenue. Red Hat’s model involved selling a low-cost software package for the Linux OS while generating revenue through technical support and training for the complex system.
A more fitting comparison, analysts note, might be to the creators of Spark, which built a highly successful business around their open-source project.
Meta has reported that Llama’s token volume across major cloud service providers (CSPs) more than doubled between May and July 2024, following the release of Llama 3.1. Moreover, token volume usage grew tenfold from January to July 2024 for some of Meta’s largest CSP partners.
“While Meta has struck revenue share deals with the major CSPs who sell Llama 2 and has indicated expectations to have similar deals for Llama 3.1, we expect revenue from said deals to be immaterial to the overall business,” Jefferies analysts wrote.
Given its vast market share and the absence of its own industry-leading foundational model, Jefferies believes Amazon (NASDAQ:AMZN) Web Services (AWS) stands to benefit most from Llama’s increasing usage.
Analysts also highlight that executives from Amazon (AMZN) and Meta have been in discussions, as reported in August, about a potential deal that could lead to Meta spending over $1 billion annually on AWS.