PulseAugur
EN
LIVE 09:18:31
commentary · [1 source] · · 中文(ZH) 未来推理将吃掉70%算力,30%留给训练丨硅谷投资人张璐@AIGC2026

VC: AI inference to dominate compute, communication energy use to soar

A venture capitalist predicts that AI inference will soon consume 70% of computing power, up from its current 50%, while training will decrease to 30%. This shift is driven by the increasing use of intelligent agents over simple chatbots. The VC also highlighted that communication within data centers consumes significantly more energy than computation itself, suggesting a critical need for advancements in optical communication technologies. Furthermore, the focus is moving towards acquiring high-quality, real-world data, particularly in sectors like healthcare, rather than solely pursuing larger models. AI

Summary written by gemini-2.5-flash-lite from 1 sources. How we write summaries →

IMPACT Predicts a significant shift in AI compute allocation towards inference and highlights the critical energy demands of data center communication.

RANK_REASON The cluster summarizes a VC's perspective and predictions about the future of AI infrastructure and applications, rather than reporting on a new release, product, or research finding.

Read on 量子位 (QbitAI) →

COVERAGE [1]

  1. 量子位 (QbitAI) TIER_1 中文(ZH) · 允中 ·

    Future reasoning will consume 70% of computing power, leaving 30% for training | Silicon Valley investor Zhang Lu @AIGC2026

    技术创新只是起点,产业整合速度才是AI落地的真正竞争力