The smartphone industry is entering a phase where artificial intelligence is no longer confined to isolated features but is becoming central to how mobile phones are designed and used. Conversations with chipmakers suggest that the device in a user's pocket could increasingly act as their primary AI computing platform, driven by advances in on-device processing.

"MediaTek firmly believes that smartphones will play a central role in driving breakthroughs and advancements in AI technology," said JC Hsu, Corporate Senior Vice President at MediaTek and General Manager of its Wireless Communications Business Unit.

This shift is being underpinned by a surge in AI usage globally. "AI inference is exploding, with global monthly token usage now exceeding 6,000 trillion tokens," Hsu said, pointing to rising demand for computing that can operate beyond cloud dependency.

For years, AI on smartphones powered functions such as face unlock, camera enhancements and predictive typing. The current transition is broader. Chipmakers are reworking the architecture of mobile processors to place AI at the centre rather than treating it as an add-on.

"Generative AI is pushing the smartphone industry toward AI-first device design, where the AI engine (NPU) becomes central to the Chipset architecture rather than an auxiliary component," Hsu said.

He added, "Neural Processing Units (NPUs) are moving from secondary components to the centerpiece of the System-on-Chip (SoC). Chipmakers are now competing on 'TOPS per watt' (efficiency) rather than just raw speed."

In simple terms, this means the part of the chip that handles AI is no longer a supporting element - it is becoming the main engine. Instead of only making processors faster, chip makers are now trying to make them smarter and more efficient so that AI tasks can be performed without draining the battery. This shift could change how consumers judge phones, moving attention from specs like clock speed to how well AI works in daily use.

At the centre of this transition is "edge AI," a term used to describe AI models running directly on devices rather than relying on remote servers.

"Edge AI refers to running advanced AI models directly on EDGE devices like smartphones, using dedicated chipset hardware for real-time, local processing without cloud dependency," Hsu explained.

"On-device GenAI deliver smooth, real-time AI experiences anytime, anywhere without relying on the network, keeping data secure and making intelligent services instantly accessible," he said.

This method reduces lag and reduces the need to send personal data to external servers. It also keeps features working, even with unreliable internet connections, which remains a real concern in some areas across the globe, including India.

The shift toward on-device AI is also expanding what smartphones can do. "MediaTek recently introduced a technology preview of Omni, based on the Qwen 3 Omni 4B model, marking the first 4B fullmodality AI model to run directly on a mobile device. As a true fullmodality model, Omni is designed to understand voice, video, and text together, enabling more holistic and contextual AI interactions entirely on-device," Hsu said.

Put simply, this means the phone can process different types of information at the same time. By combining voice, visuals, and text inputs, the AI gets a richer grasp of context, enabling it to generate more well-rounded, comprehensive responses. It can, for example, process spoken language while also interpreting what's happening through the camera. This could make interactions feel more natural, effectively changing the smartphone from a basic device into an assistant that understands its surroundings.

Chip manufacturers are reacting by pouring resources into specialised components, like NPUs, designed to manage AI tasks with greater efficiency. MediaTek's Dimensity 9500 chipset reflects this direction, boasting enhancements that seek to optimise both performance and power consumption. The wider industry is following suit, as partnerships between chipmakers and smartphone manufacturers speed up the introduction of AI capabilities. These features encompass real-time translation, image refinement, and personalised settings tailored to individual user data. Analysts believe this is reshaping the competitive environment.

Instead of focusing only on hardware differentiation, companies are building ecosystems where AI capabilities are embedded across functions and improve over time. Expansion beyond flagship devices

While generative AI features were initially limited to premium smartphones due to hardware requirements, this is beginning to change."Generative AI features were initially limited to flagship/premium smartphones because they require advanced processors, large memory capacity, and optimised AI hardware," Hsu noted, adding that "these capabilities are also expanding across various price tiers."

Premium devices are likely to continue offering more advanced capabilities, but mid-range smartphones are expected to adopt selected AI features at a faster pace than previous technology cycles.

The shift toward AI-centric smartphones is underway, though still in flux. The trajectory, however, is becoming evident. Phones are evolving from mere conduits to cloud services into self-sufficient computing platforms. This evolution has the potential to alter how we interact with our devices. Rather than constantly toggling between applications to get things done, AI systems might take over workflows, operating behind the scenes and reacting to our needs, not just our explicit instructions. If this trend persists, the smartphone might not only retain its status as the most ubiquitous device but also emerge as the central interface for daily AI interactions.