In This Story
Despite growing concerns over future demand for artificial intelligence chips, both Meta (META+1.71%) and OpenAI can’t seem to get enough right now.
Both companies are seeing “significant capacity constraints” for graphics processing units, or GPUs, which are essential for training and inferencing generative AI models, analysts at Morgan Stanley (MS+0.81%) said in a note highlighting takeaways from the firm’s Tech, Media & Telecom 2025 conference.
“[A]t a high level these comments are consistent with our thematic framework where a small number of leading companies (or individuals) are focused on training/building the leading LLM ... which will require significant compute,” the analysts said.
Meta uses GPUs for more than training its AI models, the analysts noted, including for ranking and recommending content and ads across its social media platforms. The tech giant “indicated that multiple teams are still waiting for GPUs,” the analysts said.
Morgan Stanley noted that Meta “continues to be the company delivering the most incremental ROIC [return on invested capital]” from its investments in generative AI. Earlier this year, Meta chief executive Mark Zuckerberg said the tech giant is planning to invest between $60 billion and $65 billion in capital expenditures on AI in 2025.
Meanwhile, OpenAI said that its GPU supply “is completely saturated,” and that it has never been in the position “where it can’t sell out access to its GPUs at reasonable margins,” according to the analysts.
Access to data, however, is not a constraint for OpenAI’s model development, Morgan Stanley said. The AI startup said that it can create synthetic data — or non-human, computer generated data — with its existing compute to meet its needs.
Neither OpenAI nor Meta immediately responded to a request for comment from Quartz.
OpenAI chief executive Sam Altman also said at the conference that he sees AI’s potential to cause deflation “as most underappreciated/misunderstood by investors,” according to the analysts. Altman’s views are aligned with the analysts’ previous thematic work showing that increasing efficiency in accessing AI models could “help offset inflation,” Morgan Stanley said.
After Chinese AI startup DeepSeek demonstrated models on par with those from OpenAI and Meta for seemingly far less money and with weaker GPUs, investors were spooked, sparking a global sell-off of tech stocks. Despite challenging Big Tech’s billions in spending on AI infrastructure, analysts have said that more efficient AI models could make the technology cheaper and more accessible.