Qualcomm CEO Cristiano Amon has forecasted a significant increase in AI token generation, predicting 1.27 trillion AI inference tokens will be produced every 10 seconds by 2030. This projection, shared during his COMPUTEX 2026 keynote in Taipei, marks a 40-fold increase from the current rate of 31.7 billion tokens every 10 seconds. These tokens are essential units of text, code, and reasoning generated by large language models. Amon emphasized the need for the semiconductor industry to address the infrastructure challenges posed by this growth. He advocated for moving AI processing to the edge, utilizing Qualcomm's Snapdragon platforms to handle workloads locally on devices like phones and PCs. This approach could reduce token usage by up to 60% and cut costs by four times, offering a strategic advantage for Qualcomm in the evolving AI landscape.