Qualcomm's Nitin Kumar outlines the company's vision for the agentic AI era, where smartphones, PCs, wearables, XR devices, and cloud infrastructure work together.
The year 2026 can be largely related to the widespread adoption of Artificial Intelligence (AI) and the year of agentic AI. We have seen all the major brands building new AI agents, whether it be Meta, OpenAI, Google, Anthropic, and more. However, only a handful of players are focusing on the hardware that can power the new AI features. A few years down the line, the conversation around chips used to be fairly simple: faster cores, better cameras, longer battery life, and more.
However, today the same chip that powers your smartphone is also expected to run a plethora of AI features. That said, though many big players are adapting to this fast-paced AI environment, one brand has invested in infrastructure even before the era of AI began. Yes, we are talking about Qualcomm.
The company has spent the last several years pushing its Snapdragon architecture well beyond smartphones, into PCs, wearables, XR glasses, cars, robotics, and now data centres. The company is now betting on a future where intelligence will become distributed across an ecosystem of products, moving seamlessly between different gadgets and cloud infrastructure, depending on the task at hand. We had the chance to speak with Nitin Kumar, Vice President of Product Management for Snapdragon Chipsets at Qualcomm., to understand how silicon can help drive the new agentic AI era.
For years, Qualcomm has been associated with the smartphone business. Snapdragon processors became synonymous with premium Android experiences, with popular brands like Samsung, Xiaomi, OnePlus, OPPO, Vivo, and countless others using their chipsets on their mid-range to flagship models. However, this identity has now changed entirely in 2026.
Qualcomm now has silicon products spanning an extraordinary range of computing categories. At one end of the spectrum are earbuds and wearables operating within milliwatt power budgets. At the other end are AI infrastructure products designed for data centre deployments consuming thousands of times more power. “We have a large spectrum of Qualcomm device portfolio that we're offering that we showed yesterday, right from earbuds to wearable devices to XR glasses... to smartphones, to PCs, to tablets, to automobiles, to robotics, all the way to data centre,” Kumar explained.
He revealed that the credit for this continuum goes to the years of upfront investment rather than a sudden pivot. "We have invested over many years in our technology stack across all different IPs, whether it's CPU, GPU, AI capability, audio capability, from devices that will be sub one watt and actually milliwatt kind of a use case scenario, whether it's your earbud devices, all the way into like kilowatt kind of a scenario, which would be on the data center," he explained.
The breadth of that portfolio, ranging from a few milliwatts to the AI200 data centre rack at roughly around 160 kW, matters because the next generation of AI experiences will almost certainly not be confined to a single form factor. The challenge for the industry is figuring out how these experiences connect. However, for Qualcomm, this gives it an edge over rivals that specialise in just one end of this range. "We're the only ones who can run an AI in a sub one-watt device and all the way into a kilowatt kind of a device," Kumar said.
Qualcomm's AI vision is not focused on a specific product category, but a conglomerate of all. Kumar revealed that owning silicon across categories matters the most for the next phase of AI. "If you look at all your personal device categories for a second... I have my Windows PC here, I have my smartphone here, I'm wearing a watch powered by Snapdragon, and I'm wearing my Ray-Ban Meta glasses that are also powered by Snapdragon. Because it is powered by Snapdragon, and for the next generation of AI experiences, having that context will be super important to drive contextual AI," he said.
Nitin Kumar, Vice President of Product Management for Snapdragon Chipsets at Qualcomm.
He further added that this provides a structural advantage that no other silicon vendor currently offers. "Because all these are powered by Snapdragon, it gives us a unique advantage in terms of like, we're the only silicon player that can tie a common thread between an XR glasses that are available today to a wearable to my earbuds that I have in my pocket to a PC to a tablet to a smartphone. It gives us that unique ability to stitch this contextual thread together and drive the best experience across all these devices powered by Snapdragon, so that there is an enhanced capability that you're able to get out of your device," Kumar said.
When asked who decides when a workload should stay local or when it should escalate to larger cloud models, Kumar revealed that there is no fixed rulebook for it. "The world is actually changing very fast... as models evolve, the capability of the models evolves, the accuracy of the model gets fine-tuned, the space is changing in terms of what might exist or what did exist in terms of capability last year versus what exists now will be very different from what might even exist in one year," he noted.
He believes that the future belongs to neither camp completely, but to what he called “distributed compute”. "The general theme from us is very clear, that we believe in more of a distributed compute, if you will, that the best application from a user perspective is when the technology just blends in the back, and the user just gets the output that they are looking for in terms of the capability," he said. He added that this depends heavily on how capable the underlying hardware is at any given time, since models themselves keep shrinking. "A lot of that depends on the capability of the platform, the system capability of the platform, which by itself is evolving as models get quantised to a smaller footprint by maintaining the accuracy to an acceptable degree. That intelligent orchestration is where we'll have an advantage in how we can distribute the workload," Kumar explained.
The AI space is changing rapidly. There was a time when AI was used only for trivial matters; now it is a full-blown experience for consumers. Talking about the future of AI and whether the next breakout consumer AI experience will be fully local, cloud-assisted, or a hybrid mix, Kumar reveals that the space is evolving at a rapid pace. “So it's what was there a year ago is very different today, and might be very different. It really changes very fast, every few months, I think," he said.
Qualcomm showcased a bunch of AI usecases along with its Dragonwing at Computex 2026.
However, he believes that as the world is moving towards agentic AI, there will be a bigger shift in how people will interact with their devices altogether. “The landscape changes very fast — every few months, I think. But one thing is clear if you look at the trend: the world is moving towards more AI. The world is moving towards agentic AI. In fact, our belief is that AI is the new UI as well," he reiterated. As of today, he explained that users are the ones giving commands to AI. However, this will change in the near future as agentic AI comes into play. “Right now, we are the ones giving AI commands as users. In tomorrow's world, agents themselves will be giving more AI commands, in terms of how they orchestrate between local, cloud, or some form of hybrid or distributed AI workloads," he further explained.
That said, he still believes that, even with an agentic AI future in mind, on-device AI will still depend on the cloud to some extent. “I believe that as the world continues to evolve and the demand for space continues to evolve, it will always remain a distributed AI approach. As much of the AI as possible will run locally, and then there will be certain applications that will require cloud usage,” he added further.
Nitin says that the capability of local devices will continue to improve exponentially, enabling them to run a lot more. At the same time, demand for more AI is increasing at a very aggressive pace. “I believe distributed AI will be the right architecture, one that fits across a variety of needs for users. A user can buy a smaller capable machine, a large capable machine, or a really powerful machine locally. But some use cases might still exceed local capability, and for those, you may want to go to the cloud," he explained.
As agentic AI becomes more popular each day, one question that remains uncertain is how to gauge whether a device is agent-ready. When asked what the minimum technical requirement should be for a device to qualify as agent-ready in this new AI era, Kumar revealed that it is a difficult question to answer “because the AI agent space is also very, very wide."
Qualcomm has been quietly building AI into its chipsets even before AI took centre stage. "Let me give you a couple of wide-spectrum use cases on both, on both sides. Okay, take a smartphone as an example. Believe it or not, AI is, on a relative basis, a new term. An NPU integrated into an SoC is a new term. Qualcomm has been integrating an NPU on a mobile chip for maybe about 12, 13 years," he said.
The Dragonwing IQ10 Robotics Reference Design.
" If you go back several years, you would take a picture and be able to blur the portrait out of it, which was called the bokeh effect. That was using an NPU, an AI engine integrated on the chip. When you can use a feature like Hey Snapdragon on a smartphone, and the phone wakes up, that is also running on a very tiny NPU that is part of the audio engine we have. It is a very small engine, but it is still an AI engine. It is an NPU actually running a neural network whose one job is to detect context from the world around it. The reason it requires AI processing is that it has to do that all the time, listening continuously at much lower power," he explained.
“On the other side of the spectrum, like, on a PC, we have an 80 TOPS of NPU capability, and as you go into the server side, we have far more AI capability that exists in several of the offerings that have been announced, and more to come on that," Kumar said. Given how wide that range is, he revealed that it is really hard to set fixed parameters for a device to be labelled as agentic AI-ready. "It is very hard to say what the minimum size is and what the maximum size is, as the use cases are very wide in spectrum. One thing is for sure, that we'll have the best solution that will provide capability for that solution within the constrained form factor of that device," he said.
Qualcomm is ready for the agentic AI revolution. Kumar says that the company is offering an end-to-end solution for the developers. "Leveraging Qualcomm Stack, Qualcomm toolchain, Qualcomm processes, Qualcomm support, they can optimise an AI algorithm on Snapdragon architecture, and then run that on a smartphone, run that on a PC, run that on a tablet. That gives them a unique advantage," he said.
"We are working with several of our ISV partners across different segments, whether it's from content creation, document summarisation, music creation, or different vertical industries, where we are enabling the partners to leverage the on-device AI capability and offer a different use case than it was non-existent," he explained.
That being said, Qualcomm is building an ecosystem of silicon while using its expertise in the performance-per-watt compute. "Our DNA has been in terms of providing the best performance while preserving battery life, that is a performance per watt advantage that we have had for decades now, and the reason we are like that is that our back of the house wealth has been from a mobile industry," he said.
He explained that engineers at Qualcomm always design a new architecture with power optimisation and power saving in mind. Whether it be an architecture for earbuds to XR glasses or a Windows PC, the core philosophy remains the same: deliver the maximum performance while preserving the battery life.
“We will continue to innovate and lead the market in that direction. That is in our DNA, that is our core strength, and our technology stack is the proof point of that. We continue to invest very heavily in our technology stack to make sure that we maintain a leading position, because that is the fundamental promise that Snapdragon provides to the end customer, and we must deliver on that promise. When you are buying a Snapdragon device, there are some fundamental things that you are going to get from that device, no matter what device category it is. And one of them is best-in-class performance per watt," he explained.
Get your daily dose of tech news, reviews, and insights, in under 80 characters on Gadgets 360 Turbo. Connect with fellow tech lovers on our Forum. Follow us on X, Facebook, WhatsApp, Threads and Google News for instant updates. Catch all the action on our YouTube channel.