Ahead of Apple’s Worldwide Developers Conference (WWDC), running from June 8 to 12 local time, interest appears to be rising again in Apple’s AI strategy.

How disruptive the AI features Apple is putting into the iPhone, including the next version of Siri, is the biggest point of interest. Observers also say Apple is likely to stress its on-device AI capabilities, running AI on devices rather than in the cloud.

According to The Information, Apple plans to give significant weight at this WWDC to messaging around its competitiveness in on-device AI, citing its experience developing its own chips for the iPhone, Apple Watch and Mac computers over the past 15 years.

The Information reported this, citing sources familiar with WWDC plans.

Most current AI models are delivered via the cloud through data centres packed with advanced AI chips. Apple devices are no exception. Some user queries for the next version of Siri, which Apple is expected to unveil at WWDC, are handled on Google Cloud under an agreement between Apple and Google.

Cloud-based AI delivers strong performance, but has limitations in cost and privacy. If AI runs directly on a device, users can cut the risk of data leaks while also reducing costs.

With AI token costs recently emerging as an issue at many companies, on-device AI is drawing more attention as an alternative to cloud AI.

From Apple’s perspective, on-device AI could also be a business with much to gain because it does not need to pour astronomical costs into building the data centres required for AI, unlike other big tech companies.

ㆍ[Tech Insight] Apple, which spends the least money on AI, will win the consumer AI war?

Technology analyst Richard Kramer (리차드 크라머) estimated in a recent report to investors that the value of Apple’s on-device chip compute capacity is about $50 billion. He said such compute capacity is effectively “funded by users”.

Apple unveiled the M5 chip last October with a neural engine accelerator embedded in all GPU cores. By its own performance measurements, the M5 can run a 30 billion-parameter model within 3 seconds. Apple’s push to expand on-device AI appears to be broadening beyond efficient chip development to include AI models running on devices and the software capabilities that support them.

According to The Information, as part of an agreement with Google, Apple has been able to use Google’s Gemini models to train small models that can run on its devices. Apple is also reviewing acquisitions of smaller companies that could help shrink AI models so they can run on devices.

Among the companies Apple is considering acquiring is LiquidAI, a startup specialised in running AI on devices, The Information reported, citing sources.

Startups are also emerging that develop on-device AI-based applications rather than cloud-based ones. WebAI is one of them. WebAI, which specialises in enterprise AI applications that run on Apple chips, provides AI tools for customers in the aviation industry.

The company’s AI applications run on an iPad or a Mac rather than in the cloud, and were developed by learning vast manuals that explain complex Boeing engine structures and operating principles. They are used to support engine maintenance.