[Apple in Talks with a Startup That Compresses AI Models for Local iPhone Operation] According to Sina Finance, Apple is evaluating the technology of PrismML. This startup claims that its technology can deeply compress high-performance AI models, enabling them to run directly on iPhones while reducing memory usage by up to 15 times. This breakthrough could allow more AI computations to be performed locally on phones, eliminating the need to send requests to the cloud, thereby improving Siri's speed and enhancing user privacy protection. If PrismML's claimed performance is fully realized in practical tests, this technology could reshape market demand for memory and data center computing power. However, analysts note that the AI industry as a whole will continue to consume significant chip resources.