In his keynote address at COMPUTEX 2026, Qualcomm CEO Cristiano Amon set a clear tone for future technological developments, emphasizing that 2026 is the "Year One of AI Agents." He further pointed out that AI development has reached an astonishing scale and has evolved from simply responding to prompts to becoming an agent capable of taking autonomous actions, thereby changing traditional computing architectures and the way terminal devices are used.
From "Passive Tool" to "Active Agent"
For the past few years, smartphones have been the absolute center of users' digital lives. However, with technological evolution, the center of the future digital experience will shift to AI agents, with various physical devices such as mobile phones and PCs becoming end points for AI agents, allowing AI agents to seamlessly "roam" across various devices.
These agents operate very differently from traditional applications: they run continuously in the background and are able to perpetuate the user's context.
Cristiano Amon emphasizes that AI agents act based on the user's intent, proactively understand tasks, and safely and reliably coordinate the execution of multiple tasks.
To support the massive AI ecosystem, Cristiano Amon also reviewed the current growth scale of global terminal devices in his speech, including the fact that smartphones have accumulated to more than 60 billion units in the global market, personal AI devices evolved from wearable devices have accumulated to more than 20 billion units, PCs have exceeded 20 billion units globally, and connected vehicles have reached 5 million units. These devices will become the "stage" on which AI agents can play a role in the future.
A fundamental change in the hardware architecture of terminal devices
As "AI agents" become central to daily life, existing device architectures will face significant challenges. Cristiano Amon stated that current devices are not designed for these entirely new AI agent experiences, meaning that future device hardware architectures must undergo fundamental changes.
In response to this trend, future hardware designs must meet the following conditions:
• Strict power consumption and latency limits:The device faces strict power consumption and latency limitations, making battery life a critical test.
• Strong task coordination capabilities:The system requires a powerful and energy-efficient CPU to coordinate various AI tasks.
• Local model execution capability:It must be equipped with energy-efficient NPUs and GPUs with high computing density to facilitate the smooth execution of local AI models on the device.
• Deep integration:Deep type integration is necessary between different computing domains.
Distributed computing and the full launch of the ecosystem
This transformation is not limited to single hardware devices; future AI computing will take the form of "distributed computing," spanning across the cloud and terminal devices. In this architecture, the data collected by sensors and contextual information will become extremely important, and Cristiano Amon further points out that AI agents lacking contextual information will be unable to realize their practicality and initiative.
In terms of software and operating system ecosystem, the adoption and use of AI is accelerating. For example, orchestrators such as Open Cloud and Hermes can already run natively on Snapdragon PCs, while Google is integrating agent AI directly into the Android system through Gemini smart technology. Microsoft is also carrying out the same OS-level layout to make AI applications more widespread.
Expanding into physical AI and 6G communication deployment
Beyond personal computing devices, Qualcomm is also focusing on the physical world, explaining that fields including automobiles, robotics, and industrial systems will undergo a transformation driven by Physical AI. Taking automobiles as an example, Cristiano Amon predicts that future connected cars will possess a "two-layer intelligent architecture": one layer is personalized AI responsible for the cabin experience, and the other layer is the physical AI system actually responsible for driving the vehicle.
Furthermore, the distributed AI experience that underpins all of this relies on more advanced network infrastructure. Therefore, Cristiano Amon pointed out in his speech that the next-generation wireless communication technology "6G" must have three core pillars: connectivity, distributed computing, and sensing, to ensure that AI computing can be quickly, stably, and seamlessly transferred across different scenarios and devices.
Qualcomm's "all-round" computing ambitions
Qualcomm's market ambitions now extend beyond mobile phone chips. Cristiano Amon emphasizes that the systems Qualcomm is building span a wide range, from headphones equipped with low-power Wi-Fi (requiring less than 2 milliwatts of power consumption) to the now ubiquitous mobile phones and PCs, as well as various AI devices, robots, connected vehicles, and even data centers, all involving a massive computing power scale of kilowatts.
Meanwhile, in response to the current trend of AI agents driving new computing paradigms, Cristiano Amon believes that tokens will become the next "new currency," with demand increasing 40 times between 2026 and 2030, and the number of tokens after 2030 will be beyond imagination.
Regarding the announcement last year of its renewed foray into the data center market and the unveiling of its new data center product brand "Dragonfly" in this keynote speech, Cristiano Amon kept the details under wraps, stating that they would be announced at the investor day "at the end of this month".
