>Is the speculation that people will just move to using China's hosted version of it rather than running it on their own infrastructure? β
That's a part of it. The bigger issue (probably) is buried in the last paragraph. Not only did they allegedly do it with lower grade equipment, but at a fraction of the cost:
>The startup allegedly spent just $5.6 million to train its latest models, a fraction of what American AI firms have been spending to build out their own models. DeepSeek also relied on chips with reduced capabilities from Nvidia due to stringent U.S. export controls.
