DeepSeek open source success a positive development for Meta, says Stifel

Stifel notes reports that suggest China’s DeepSeek was, in part, refined through distillation of Meta‘s (META) Llama models, which the firm believes to be “a violation of Meta’s policies.” Meta allows Llama models to be used for free until those apps or projects achieve 700M MAUs, which “is a big number, but perhaps this might entice Meta to lower the MAU threshold to begin explicit monetization,” the analyst tells investors. At “a minimum,” the rapid timeline at DeepSeek is “a testament to open source models,” which the firm views as “a positive development for Meta.” Cheaper and more efficient models, if DeepSeek is, in fact, “cheaper,” should equate to a pickup in demand for AI workloads, which should translate into greater consumption across Amazon’s (AMZN) AWS and likely Google’s (GOOGL) GCP, adds the analyst, who views these developments as “neutral to positive” for Amazon and Google and suggestive that “current capex levels are likely justified in the near term.”