Apple eyes PrismML tech to run bigger AI on iPhones

Apple is reportedly in evaluation of AI model compression technology developed by Silicon Valley startup PrismML. The technology is intended to address a key challenge in mobile AI: fitting larger, more capable AI models onto iPhones so they can run directly on the device rather than requiring constant communication with cloud servers. The appeal of this approach is twofold. First, on-device inference reduces latency—users get AI-powered answers and features instantly without waiting for a round trip to a remote data center. Second, it reduces the need to send user data to the cloud, which aligns with Apple's longstanding commitment to privacy and on-device processing. PrismML's compression technology appears to be central to making this feasible, as it would allow larger models to be reduced to a size that fits within the storage and memory constraints of an iPhone while maintaining the quality and capability users would expect from such features. Whether this evaluation leads to actual deployment in future iPhones will likely depend on PrismML's ability to prove that its compression does not compromise the performance of the AI models themselves.

Apple's reported interest in PrismML's compression technology reflects a broader strategic shift in how the company wants to deliver artificial intelligence features to users. Currently, running large AI models typically requires offloading computation to cloud servers, which introduces latency and raises data privacy concerns. By evaluating technology that compresses these models to fit on iPhone hardware, Apple is attempting to solve both problems at once: delivering faster AI responses while keeping user data on the device itself. This aligns with Apple's historical emphasis on on-device processing and privacy. The success of such an approach would hinge on PrismML's ability to compress models without significant loss of capability—a technical challenge the startup's technology is apparently designed to address.