Why Apple’s New CEO Must Prioritize On-Device AI for Privacy and Performance

“Apple has a new CEO; he’s a hardware guy.” This concise summary of Apple’s leadership transition quickly spread across Silicon Valley and the broader tech industry. The company’s choice, John Ternus, ascended through the ranks in hardware, taking over iPhone engineering in 2020 and all hardware engineering a year later.

Analysts suggest Ternus’s appointment to succeed Tim Cook indicates Apple will approach AI with a measured strategy: leveraging AI to enhance device functionality without overhauling its services or core business. While competitors like Microsoft, Google, and Meta invest tens of billions annually in AI research and data centers, Apple’s spending in these areas has remained comparatively stable. Its AI research group has not become the company’s primary focus.

However, appointing a “hardware guy” as CEO does not imply Apple’s AI initiatives will be marginalized or limited to minor features, such as removing unwanted objects from photos. Instead, Apple’s strategic advantage lies in running advanced AI models directly on its own hardware—not in the data centers of third-party corporations.

Apple’s personal AI models would operate within a secure enclave on its chips, similar to how Apple Pay protects financial data. By processing on-device, these models could handle personal and sensitive information with speed and efficiency while avoiding cloud transmission—ensuring privacy remains intact.

Do these considerations truly matter? Absolutely. As public distrust of major AI labs grows and regulatory oversight lags, Apple’s long-standing reputation for data privacy could become a key competitive advantage. The company has spent years building credibility in this area, and AI presents an opportunity to capitalize on that trust.

Currently, running large-scale AI models on laptops and smartphones remains a work in progress. Under Ternus’s leadership, Apple may have the ideal mix to achieve this goal. Ternus played a pivotal role in the transition to Apple Silicon, which serves as the foundation for the company’s AI strategy. Johny Srouji, who led Apple’s silicon engineering, will now oversee hardware in Ternus’s former role.

Ternus also shares a long and productive working relationship with Craig Federighi, Apple’s software chief. Federighi is taking charge of most of Apple’s AI research group and will be instrumental in integrating AI models into the company’s operating systems and applications.

To successfully run large AI models on compact chips, Apple’s hardware, silicon, and software teams must collaborate seamlessly. The company’s AI track record includes notable setbacks. Siri, for instance, remains a broken promise. In 2024, Apple announced plans to transform the command-based assistant into a systemwide AI agent powered by large language models, promising highly personalized features for iPhones. To date, these improvements have not materialized.

Yet, as previously noted, Apple still has an opportunity to lead from behind. While the company may not match OpenAI, Anthropic, and others in developing massive, general-purpose AI models, it can carve out a unique position by focusing on