An investor argues that Macs could handle coding assistance, document analysis and local automation while larger or more complex tasks go to the cloud—not replace the infrastructure needed to train frontier models. The post identifies keeping some sensitive data on company-controlled hardware as a potential benefit, but warns that local execution does not automatically solve security problems. It also cautions that local AI is not always cheaper: hardware costs, utilization, electricity and maintenance matter. For Apple, stronger hardware demand remains a thesis. The author says there is not yet evidence that local AI is materially increasing Mac revenue or driving meaningful enterprise market-share gains.