Qualcomm CEO Says AI Requires Gazillions of Tokens
Qualcomm CEO Cristiano Amon stated that future AI systems, especially Agentic AI, will require "gazillions" of tokens.
This statement emphasizes the enormous demand for computing power and token processing as AI evolves towards multi-agent and intelligent systems, while also mentioning opportunities in robotics, automotive AI, and AI-native 6G.
Market Mechanism: Semiconductor investors as buyers focus on long-term AI demand, with capital continuously flowing towards high-performance chip and edge computing companies. Qualcomm benefits from a strengthened narrative of diversified AI applications, while pure training chip suppliers face pressure under token scale expectations.
Supplementary Data: Qualcomm reiterated "Resistance is futile" during the keynote, and QCOM stock price briefly fell in pre-market trading.
Source: Public Information
ABAB AI Insight
Cristiano Amon has repeatedly pushed Qualcomm to transition from mobile chips to AI edge computing during his leadership. The phrase "gazillions of tokens" continues his long-term optimistic forecast for Agentic AI, having previously laid out a multimodal and edge-side AI strategy through the Snapdragon platform.
In terms of capital pathways, Qualcomm is shifting R&D resources and capital towards supporting chip architectures capable of handling massive token processing, motivated by capturing incremental markets as AI migrates from the cloud to edge in scenarios like robotics, automotive, and 6G, while complementing rather than directly competing with cloud giants like NVIDIA.
Similar to NVIDIA CEO Jensen Huang's repeated emphasis on the scale of AI infrastructure, Qualcomm is currently in the mid-transition phase from a mobile SoC leader to a multi-scenario AI chip supplier, focusing on capturing edge-side token processing share.
Structural Judgment: This essentially represents a technological substitution. Qualcomm replaces part of the centralized cloud training model with edge computing chips, reducing overall latency and energy consumption by supporting distributed processing of "gazillions of tokens," facilitating a structural leap of AI from centralized large models to multi-agent intelligent systems, driven by the explosive growth of AI application scenarios matching the surge in edge-side computing power.
ABAB News · Cognitive Law
AI does not save tokens; it requires massive tokens to be truly intelligent.
Edge-side processing capability determines the actual deployment speed of the next generation of AI.
The larger the scale, the more decentralized rather than centralized it needs to be.