Qualcomm CEO Cristiano Amon explains why AI agents change chips, software, and budgets

If you are trying to map the AI hype cycle to real-world spending, the phrase “AI agents” is the fork in the road. In a CNBC interview on The Tech Download, Arjun Kharpal sat down with Qualcomm CEO Cristiano Amon to talk through what this new world actually means, not just for models, but for the devices and chips that have to keep up when AI stops being a chatbot and starts taking actions.

Amon, speaking as the CEO of a company built around mobile and edge compute, frames AI agents as a shift in the underlying requirements. The big change is that agents are not merely answering questions. They are coordinating steps. They need enough processing headroom, low enough latency, and the right systems integration to support workflows that feel continuous rather than request-and-response. In plain terms: if AI becomes something that does work, the compute question becomes sharper, and the “where does it run” question turns from background detail into a budget line.

To understand why Qualcomm’s CEO would care, you have to zoom out to how chip and platform strategies usually evolve. Mobile chips have long been optimized around a tight set of constraints, power consumption being the headline. When AI workloads were smaller or more centralized, companies could often treat on-device AI as a nice-to-have, while the heavy lifting stayed in data centers. AI agents complicate that. Agents imply more frequent interactions, more state, more coordination, and more time spent actually executing tasks. That can raise pressure on edge compute, because waiting for round trips to the cloud can break the “agent” experience. The relevance for executives is simple: performance, responsiveness, and efficiency become more intertwined, and that tends to push procurement and roadmaps toward silicon and platforms designed for sustained workloads, not just occasional inference.

There is also the software reality that executives cannot ignore. AI agents are mostly an orchestration layer over models and tools, which means the “agent ecosystem” becomes a strategic battleground. Chipmakers like Qualcomm live at the intersection of hardware capability and platform compatibility. If agents proliferate, developers will want predictable performance across devices, plus a straightforward path from models to deployment. That raises the stakes for standards, developer tooling, and partnerships. Even if you are not a chip company, your organization still gets pulled into that ecosystem: your app roadmap, your deployment strategy, and your customer experience all depend on whether the underlying compute stack can support agents reliably.

Regulation is another reason this conversation matters. While the podcast segment is focused on the new world of AI agents, the broader policy environment around AI is increasingly about control, accountability, and risk. Agents bring a particular twist because they can take actions. That changes the compliance posture for enterprises and industries, since monitoring and governance become harder when systems do more than generate text. For board members and executives, the second-order implication is that “AI readiness” now includes not only accuracy, but also observability and guardrails. That, in turn, affects engineering priorities and can alter platform choices, because the best-performing approach is not always the one that is easiest to govern.

Now connect that to incentives. Tech leaders are trying to time bets so they do not miss the next platform shift, but they also do not want to fund hype. The smart money typically goes toward enabling layers that make multiple downstream products easier to build. In the agent era, that can mean focusing on compute efficiency, deployment flexibility, and integration with tooling developers already use. If agents start becoming a default UX pattern, organizations will want hardware and platforms that reduce friction when scaling from prototypes to production.

This is where the interview becomes relevant beyond one company. Qualcomm CEO Cristiano Amon’s perspective gives a clue about how chip CEOs think about the future workload. The question for other executives is whether your strategy is built for the old model of AI and the new model of agents. If your current plan assumes AI is something users “call,” but the market moves toward AI that “acts,” then your tech stack, your governance approach, and even your vendor partnerships need an update. The winners in this phase will be the organizations that align compute capability with agent workflows, so you get responsiveness and execution without turning operations into chaos.

For decision-makers, the takeaway is not that agents are coming. The point is that agents stress the entire chain: the user experience, the orchestration layer, and the compute underneath it. When Amon discusses the new world of AI agents on The Tech Download, he is effectively highlighting a shift in the cost and capability equation for AI systems. That shift will shape which vendors get budget, which architectures get adopted, and which teams get blamed when “AI that does things” does them poorly.