AI is shifting from cloud-based models to embedded systems, and the next phase of AI could be defined by on-device intelligence, agent-driven systems and synthetic data generation. 

The next phase of AI is agentic: AI systems that continuously reason, act and adapt in real time, driving sustained demand for token generation. 

AI must be distributed across device, edge, cloud and data centre to optimise cost, performance, power and latency as workloads become continuous 

As AI moves beyond traditional cloud-based models, companies are rethinking how intelligence is embedded onto devices, applications and user interfaces.

Generative AI and agentic AI enable more personalised and context-aware digital experiences, while maintaining strict standards around privacy and security.

Vinesh Sukumar, Vice President of Product Management, AI/Gen AI at Qualcomm Technologies, says: “There’s a fundamental shift underway from using big models to cloud-only AI inference to smaller models, driving towards distributed AI that runs directly on devices.

“This is being driven by the need for real-time responsiveness, increasing privacy expectations and the cost of scaling cloud infrastructure.

“Running AI locally allows these devices to process data instantly, near the user, while controlling the cloud inference cost.”

Innovation in the era of AI

Qualcomm was founded in 1985, taking on early research into code-division multiple access wireless technology, as well as semiconductor products. Since then, it has evolved into a next-generation computing technology company, branching out into embedded AI, IoT and augmented reality.

Its operations focus on developing and designing technologies for global communications and computing, including mobile networks, smart devices and semiconductors.

The company’s core mission is to enable intelligent computing everywhere, bringing together connectivity, compute and AI to transform how people, businesses and industries operate.

It focuses on engineering human progress by making advanced technologies like AI more accessible and scalable, as well as impactful across the society.

Vinesh joined the company in 2020 after spending six years as Head of AI Product and Strategic AI Architecture at Intel.

“I have the absolute privilege of leading the AI product team at Qualcomm,” Vinesh says. “As part of this role, I drive AI systems, hardware and software strategy and the vision for all the business units.

“I started my journey on computer vision, and transitioned towards AI. With time, my passion for perception-related use cases increased, from pattern recognition to object recognition, to segmentation use cases, then slowly moved into the consumer industry.

“When I joined Qualcomm, it was very synonymous with handset or mobile.

“There's been a transition now to look at efficient computing and domain-specific AI engines like Hexagon NPU Designs. We have prioritised power-efficient AI on-device. The philosophy has now scaled well beyond smartphones into Snapdragon PCs, IoT devices, automotive ADAS and idea environments, and now data centres as well.

“What changed is definitely the scope. We’re now delivering AI across the distributed systems –  devices, edge infrastructure and cloud environments – while staying true to its efficiency being the first foundation.”