National security concerns are driving the White House to keep a closer eye on artificial-intelligence companies' most advanced models. But the companies that supply the massive computing power needed for AI have little to fear.
The White House earlier this month hosted executives from companies including OpenAI and Anthropic to outline how it plans to monitor their models for potential security risks. The administration says it doesn't plan to release the process publicly, and it described the reviews as "voluntary." They're focused on the "closed" models controlled by private companies, rather than on the cheaper open-source models that are often created by Chinese companies.
Whatever direction the White House goes with the reviews, AI-infrastructure companies like CoreWeave are the best positioned for any policy outcome, analysts for Truist Securities said in a research note on Sunday.
"We view AI Infrastructure (CRWV, clouds, hardware) as best positioned across the widest range of outcomes, as growing inference demand may increasingly offset any moderation in training demand," the analysts wrote.
The thesis centers on the idea that cheaper, open-source models will close the gap with expensive frontier models if companies like OpenAI and Anthropic have to slow development to comply with the review process. That's negative for the model companies, but potentially a positive for companies, like Palantir and Duolingo, that integrate AI into their products, since cheaper open-source alternatives will become more attractive, the analysts wrote.
On the other hand, if the White House were to crack down on open-source models, a scenario that the Truist analysts find unlikely, that could make AI use more expensive.
Companies like CoreWeave can be mostly agnostic on how the regulation shakes out, the Truist analysts said. The company sells the compute needed to power AI inference and training. Even if demand for training new AI models drops somewhat as a result of regulation, demand for inference to perform AI-powered tasks should remain strong.
Of course, the federal government isn't the only regulatory threat to AI companies. In recent months, state and local governments have imposed or considered moratoriums on the data centers that companies including CoreWeave need to provide compute. New York last month imposed the first state-wide moratorium on data centers. Texas Governor Greg Abbott, whose state had been considered one of the most data-center friendly locales, earlier in August ordered a pause on new projects.
On CoreWeave's earnings call last week, CEO Michael Intrator said the moratoriums are "not going to impact the demand for this infrastructure. They are going to impact where this infrastructure gets built."
Comments