Apple's M7 chip and German AI research drive local generative AI

Apple announced that its next‑generation M7 processor, slated for release in the first half of 2027, will skip the Pro, Max and Ultra variants and instead focus on a specialized Neural Engine. The M7 Ultra version is projected to support up to 1.5 TB of RAM – a capacity currently found only in server‑class hardware – and aims to enable true on‑device generative‑AI workloads such as 2K image creation, reducing reliance on cloud services.

In Germany, the Fraunhofer Institute for Integrated Circuits IIS is developing the HEGEMON OTTO project, a comprehensive evaluation framework for generative AI models used in security‑critical contexts. The framework tests accuracy, transparency, robustness against manipulation and resource usage, employing multimodal models that process text, images and geospatial data. The goal is to determine which AI systems are trustworthy for agencies that need reliable, reproducible analysis.

Separately, research on future AI‑PCs highlights the emergence of neural‑processing units (NPUs) in notebooks since 2024/2025 and explores neuromorphic chips such as the Genesis prototype. These chips could enable continuous‑learning local AI assistants that adapt to tasks in creative software, development environments, office productivity and smart‑home robotics, offering a vision of personal AI that operates entirely on the device.