Apple Eyes PrismML Acquisition to Boost On-Device AI Performance on iPhone

Apple is exploring the acquisition of chip technology and the startup PrismML to enable large-scale artificial intelligence models to run directly on iPhone hardware and enhance its AI server capabilities, according to reports from Investing.com and TradingKey. This strategy aims to shift heavy AI processing from the cloud to on-device execution for specific iPhone models.

Apple’s Potential Acquisition of PrismML for On-Device AI

Apple is targeting the acquisition of PrismML to integrate advanced AI capabilities directly into the iPhone ecosystem. According to Investing.com and TradingKey, the move is designed to support the processing of large AI models on-device, reducing reliance on external servers for complex tasks.

PrismML has developed a model known as Bonsai 27B, which Thisisgame Thailand reports is capable of running smoothly on iPhone hardware. The “27B” designation typically refers to the number of parameters in the model, a metric that usually requires significant memory and processing power, making its fluid operation on a mobile device a technical milestone.

By integrating PrismML’s technology, Apple intends to allow certain iPhone models to execute these large-scale models locally.

Chip Acquisitions and AI Server Infrastructure

Beyond mobile hardware, Apple is exploring the purchase of chip-related assets to bolster its AI server infrastructure. Investing.com reports that these explorations are focused on strengthening the backend capabilities that support Apple’s broader AI ecosystem.

This dual-track strategy addresses two different bottlenecks in AI deployment: the hardware limitations of mobile devices and the massive compute requirements of server-side AI.

Shift Toward Local AI Model Execution

The effort to run large models locally marks a shift in how Apple manages AI workloads. iPhone-Droid reports that Apple is actively seeking methods to call large AI models directly on specific iPhone versions, rather than relying on a cloud-only architecture.

Local execution requires highly optimized software and hardware integration.