SCX.ai Brings Australian-Hosted AI to Apple Developers

SCX.ai has introduced a new tool that connects Apple’s native AI framework to models hosted in Australian data centers. This move makes the company the first in the country to offer this specific integration for developers working on iPhone, iPad, and Mac applications. The package, known as SCXSwiftKit, allows developers to choose between Apple’s local on-device AI and remote, cloud-based models without changing their existing code structure.

The launch addresses a growing complexity in software design where AI features are no longer standalone chatbots but are embedded into everyday apps. Developers must now decide whether a specific task should stay on the user’s device or be sent to a remote server. This choice has significant implications for data privacy, regulatory compliance, and computing costs, factors that are becoming central to business strategy.

Flexible Routing for Local and Cloud AI

SCXSwiftKit is distributed through the Swift Package Manager and conforms to Apple’s standard LanguageModel interface. This compatibility means developers can continue using familiar Apple tools for streaming, tool calling, and session management. A single application can use Apple’s local model for simple, low-latency tasks while routing more complex requests to SCX.ai’s hosted models, which offer larger context windows and greater computing power.

The system is designed for flexibility. Models are resolved at runtime, meaning new AI capabilities can be added to the service without requiring developers to update their software packages. While the primary focus is Australian hosting, SCX.ai also offers models hosted in the United States, Germany, and Singapore, allowing developers to select the optimal location based on their specific needs.

Data Residency and Sovereign Infrastructure

A central component of the offering is Project MAGPiE, described by SCX.ai as an Australian sovereign model. Hosted locally, it supports a context window of 131,000 tokens, which is particularly useful for processing longer documents. For businesses, the ability to verify where data is processed is critical. The platform exposes the hosting region as country codes, enabling applications to check the data destination before a prompt is submitted.

David Keane, founder of SCX.ai, notes that consumers are often unaware of where their AI requests are processed. He emphasizes that as AI becomes a standard part of software development, the decision of where inference occurs becomes increasingly important. He argues that Australian developers need a genuine local option to ensure data residency when it matters for compliance and privacy.

Economics of Everyday AI Inference

The company is positioning this infrastructure around the rising demand for inference, the computing required to run AI models, rather than the high-cost process of training them. SCX.ai operates SambaNova inference processors from its facility at the Equinix SY5 data center in Sydney. This setup is optimized to maximize the amount of inference work delivered from available power and physical space, rather than focusing on dedicated training sites.

As more software products embed language models, the economic focus is shifting toward the cost of daily usage. Keane highlights that while training models requires massive computing power, the next challenge is managing the continuous, high-volume requests from millions of users. This approach aims to provide a sustainable and efficient option for developers who need reliable AI capabilities without the overhead of building their own global infrastructure.