Kimi K3 is exploding in popularity and is approaching its compute power limit, as Menlo Partners’ partner says GPUs are still the “AI hard currency”

robot
Abstract generation in progress
ME News message. On July 20 (UTC+8), according to Beating monitoring, Menlo Ventures partner Deedy Das, riding the breakout of Kimi K3, proposed a “computing power long” logic. K3 entered the top ten in OpenRouter usage within two days of launch, processing about 1400 billion Tokens per day. Demand then moved close to the capacity limit of the “dark side of the moon,” and new subscriptions were subsequently paused. Das believes that over the next few years, the most valuable may be those who lock in GPUs, cloud resources, and electricity in advance. As long as the strongest models continue to get smarter, users will still pursue top-tier capabilities, and as long as inference prices do not crash, computing power will remain in long-term shortage. This logic has clear premises. Once model capability reaches a ceiling, routine tasks shift to cheaper models, or inference costs suddenly drop, and scarcity of computing power will ease. Model vendors might also generate very high revenue, yet still be dragged into losses by training, talent, and customer-acquisition costs. (Source: BlockBeats)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned