Qualcomm CEO Cristiano Amon took the stage at COMPUTEX 2026 in Taipei on June 1 and dropped a number designed to make every chip investor lean forward. AI systems currently generate roughly 31.7 billion tokens every 10 seconds. By 2030, Amon projects that figure will balloon to 1.27 trillion tokens in the same window, a roughly 40x increase that would translate to an annual total exceeding 4 quadrillion tokens.

Before anyone starts checking their crypto wallets: these are not blockchain tokens. Amon is talking about AI inference tokens, the atomic units of computation that large language models produce when they think, write, or act.

From answers to agents

Today’s AI mostly generates answers. You ask a chatbot a question, it spits back a paragraph. That interaction might consume around 10,000 tokens. The next phase, what the industry calls “agentic AI,” is a different beast entirely. Instead of answering a single question, an AI agent executes multi-step tasks autonomously: booking travel, managing supply chains, negotiating contracts, coordinating with other agents. A single agentic task can scale to over 1 million tokens, according to Amon’s presentation.

Amon called tokens “the new currency of AI,” a phrase that neatly captures how inference workloads are becoming the dominant measure of compute demand in the industry.

Qualcomm’s edge play

Most AI inference today runs in massive centralized data centers operated by the likes of Microsoft, Google, and Amazon. Qualcomm is betting that as token demand explodes, a significant share of inference will migrate to the edge, meaning the devices in your pocket, on your desk, and in your car.

Qualcomm, which already dominates the mobile processor market with its Snapdragon lineup, is positioning its edge-AI chip portfolio to capture this wave. Amon emphasized the company’s commitment to advancing on-device inference capabilities across consumer electronics and automotive applications.

What this means for investors

For crypto-adjacent investors, the terminology overlap is worth addressing directly. AI tokens and blockchain tokens share a name and nothing else. Amon made no reference to any blockchain or cryptocurrency protocols during his keynote. The entire presentation focused on semiconductor capabilities and inference workloads.

The competitive landscape is the key variable to watch. Qualcomm is not the only company chasing edge AI silicon. Apple continues to invest heavily in its Neural Engine, MediaTek is pushing its own on-device AI capabilities, and Nvidia’s reach extends well beyond data centers.