Apple Reportedly in Talks With AI Startup PrismML for On-Device Processing

Apple Inc. is reportedly in talks with PrismML, a stealth startup from the California Institute of Technology, to acquire technology that enables large-scale artificial intelligence models to run directly on devices like the iPhone. The move signals Apple's strategic push to enhance on-device AI capabilities and lessen its dependence on cloud-based processing.

A Breakthrough in Model Compression

The discussions, first reported by *The Information*, center on a significant technical achievement by PrismML. The startup successfully compressed Alibaba’s 27-billion-parameter open-source Qwen 3.6 model, allowing it to run locally on an iPhone 17 Pro.

Key details of the technology include:

• Size Reduction: The model was compressed from a massive 54 GB down to less than 4 GB.
• Full Power: Unlike typical "sparse architecture" methods that only activate parts of a model, PrismML’s technology reportedly keeps all 27 billion parameters active simultaneously without degrading performance.
• Core Innovation: The breakthrough is built on ultra-dense 1-bit and ternary weight architectures, which dramatically reduce the model's memory footprint while increasing processing speed.

Strategic Implications for Apple

Ad

Acquiring this technology would align with Apple's core strategies of user privacy, performance, and cost efficiency. By processing AI tasks on the device, Apple can offer faster response times and offline functionality while ensuring user data does not need to leave the device.

This approach also presents a path to reducing the substantial costs associated with running AI workloads in large data centers. While Apple's recent WWDC announcements included advanced Siri features, many still rely on sending data to cloud models like Google's Gemini. An acquisition of PrismML could accelerate Apple's goal of shifting these heavy AI workloads back to the user's device.

Diverging from Industry Trends

While competitors such as Microsoft, Amazon, and Meta are investing hundreds of billions of dollars into building vast AI data center infrastructure, Apple appears to be pursuing a different path. By focusing on powerful and efficient on-device AI, the company could differentiate itself in the market by making the centralized data center less critical for everyday consumer AI interactions.