Snapdragon Summit 2025: Ambitious Agentic Vision, but Hurdles Remain

This Insight is written by ABI Research Industry Analyst Paul Shell, who was invited as a guest to the Snapdragon Summit. All experiences were hosted, but no additional compensation was received.

The primary product announcements at Snapdragon Summit 2025 revolved around the smartphone and Personal Computer (PC) platforms, with Snapdragon X2 following on from the first-generation X PC systems, and the 8 Elite Gen 5 becoming the new flagship smartphone System-on-Chip (SoC), with support from Original Equipment Manufacturers (OEMs) such as Samsung, HONOR, and Xiaomi. This was augmented with a close partnership with Saudi Arabia’s HUMAIN, an Artificial Intelligence (AI) initiative launched and funded by the crown prince, on an Original Design Manufacturer (ODM)-made Windows laptop with a multi-function AI chatbot style interface and application integrations. Aside from the usual performance upgrades, standout announcements include:

• Sensing Hub: Additional micro Neural Processing Units (NPUs) and Image Signal Processors (ISPs) for always-on AI processing of audio, visual, and sensor data. This enables constructing personal knowledge graphs to be leveraged by Agentic AI applications for a more proactive experience of agents taking action in anticipation of issues such as a calendar clash.
• Snapdragon Guardian: A connectivity sub-system built into the Snapdragon X2 SoC with always-on connectivity for enterprise security, running on carrier networks to locate PCs and wipe data remotely. This manageability offering targets large enterprises.
• Android PC: Google and Qualcomm revealed plans to bring Android and Gemini to PCs with an evolution of ChromeOS and Chromebooks. This involves a common technical foundation for PCs, using the full AI stack and developer communities of both companies.
• Distributed Inference: The distribution of edge-based AI inferencing was touted after a demo of AnythingLLM, with plans to take this to wearable devices.

Absent were any major announcements related to Qualcomm’s cloud inferencing strategy, although an upcoming announcement was hinted at related to a new system for hybrid AI spanning edge and cloud environments. Whether (or to what extent) this involves Qualcomm’s existing data center IP in the form of the Cloud AI 100 PCIe card was also unclear. We also have yet to hear which Agentic AI protocols will be leveraged to enable the agentic use cases Qualcomm presented.

Cristiano Amon, Qualcomm’s Chief Executive Officer (CEO), set out his vision for the technological platform that can enable a hybrid Agentic AI primarily leveraging edge compute. This approach is built on always-on sensing and personalized, user data-based knowledge graphs that will inform proactive decision-making for productivity-enhancing applications such as personal assistants.

Although Qualcomm recognized that on-device AI has yet to gain recognition to date, particularly for consumer markets, it has set out to achieve a multi-device AI platform with features unique to its silicon. This should culminate in a smartphone, earbuds, a smartwatch, a PC, and smart glasses running on Qualcomm silicon, building and leveraging personalized knowledge graphs and distributing inference across devices.

But achieving this distributed inference framework within a single, closed Snapdragon ecosystem will be challenging. It will require coordination with OEMs, Independent Software Vendors (ISVs), and the maturation of (and industry coalescence around) future-proof data-sharing protocols to effectively manage these complex tasks. This is a complex undertaking and has similarities with Apple in terms of feature exclusivity to one brand, without the inherent advantage of the ownership of the technology stack under one OEM. Although Qualcomm’s commercial reach is larger than Apple’s in terms of a larger Total Addressable Market (TAM), it also necessitates a much more complex coordination between the sprawling Android and Windows ecosystems, and the need for consumer and enterprise buyers to connect agentic value (which remains largely unproven) exclusively with Snapdragon-powered devices.

Competition for the attention of AI developers developing cutting-edge applications is stiff, as evidenced by the billions of dollars that have been invested in the performance improvements of silicon offered by Qualcomm, MediaTek, Apple, Intel, AMD, NVIDIA, and Google to enable more on-device AI. Apple may have dropped the ball on its internal AI development—despite the inherent advantages of almost complete ownership of its hardware and software stacks—but this is a much simpler foundation upon which to build (Agentic) AI features and experiences. Google is able to launch its AI features with day-zero support on its in-house Pixel platform, but obviously on a much smaller scale than Apple. This leaves Qualcomm, MediaTek, Intel, AMD, and NVIDIA vying for a larger share of the AI demand in the mass-market Windows and Android ecosystems.

To drive demand for its silicon, Qualcomm might consider the following strategic moves:

• Agentic AI (or Adjacent) Application Acquisitions: Qualcomm should consider acquiring individual best-of-breed Agentic AI startup solutions that are leading the way in addressing specific areas applicable to Agentic AI, such as calendar and scheduling planning, and building a best-in-class Agentic AI solution exclusive to OEM solutions featuring Snapdragon’s silicon. This will not only speed up the development of genuinely productivity-enhancing applications running on-device, but also drive demand for Snapdragon platforms. The resulting applications could be distributed via a freemium model, licensed to OEMs, or given away for the first generation to stake a claim in on-device Agentic AI.
• Agentic AI Protocol Leadership: The software ecosystem supporting Agentic AI applications able to take action and share data between applications with varying degrees of “mission criticality” is still embryonic. In order to improve the success of its internal and external developer communities, Qualcomm should invest resources and take a leadership role in the committees steering the direction of protocols for both on-device and cloud-based Agentic AI to align with its cross-device and hybrid AI strategy.

• Ecosystem Enablement: Given the significant value residing in the newly announced Sensing Hub and Personal Scribe in both the smartphone and PC platforms, Qualcomm should create blueprints for the use of these hardware features in software for developing applications ultimately optimized for Snapdragon. These must go further and demonstrate the commercial value of their implementation so developers can understand the value-add, rather than viewing them as a “nice to have.” Naturally, the hardware must also be deeply integrated in Qualcomm’s existing AI Stack, including Edge Impulse.

Significant hurdles remain, especially for Qualcomm to coordinate across OEMs, ISVs, and protocol implementation, as well as to create thought leadership for agentic use cases to provide clarity for the ecosystem to develop and invest. If Qualcomm can lock experiences into its exclusive ecosystem and gain a head start, there is significant commercial potential for genuinely transformative productivity enhancing Agentic AI use cases supported by distributed inference.