Project Hub

POWER

Discovered

Not listed in Korea

Mindshare

7d
0.1%+0.1pp#187New
Mentions
1
New
Views
86
First seen
Sep 8

Sentiment

7d · 1 message

Fewer than 10 messages analysed.

Mentions

1 mentions

First to mention

Since Oct 18, 01:02 · 1 channel

Related narratives

30d

Nothing to show yet.

Related projects

30d

Nothing to show yet.

Channel contribution

30d · 1 channel

Messages

30d
  • 열충9h ago

    Why AI Bubble Burst is Just Volatility, Not Collapse Ben Horowitz & David Sacks & Chamath - All-In Podcast 1. Decisive Difference from Dot-com Bubble: "Demand Shortage vs Extreme Supply Shortage" (Ben Horowitz) 1999 LoudCloud Trauma: Horowitz and Marc Andreessen essentially founded LoudCloud in 1999 as proto-cloud. Back then, as internet infrastructure, broadband, and browsers were just rolling out, there was critically insufficient end-consumer demand to consume the massive debt-funded optical fiber bandwidth they had laid down. Exact opposite AI market dynamics: But today's AI market is not facing demand shortage but rather 'absolute supply shortage.' Andy Jassy (Amazon CEO) testimony: The night before the recording Horowitz asked Jassy "is it true capacity runs out by 2028", and Jassy answered "there's no volume even after 2028. Earliest we'll get breathing room is end of 2028." Rising supply chain costs: While per-token prices are declining, GPU prices with sufficient power and cooling infrastructure and data center space rental costs are actually rising. Data center construction faces extreme difficulty due to labor shortage and local political and regulatory pushback, which acts as huge tailwind for CoreWeave which already secured supply chains. 2. GPU Lifespan Extension and 'Forced Scarcity' (David Sacks) Why it's just volatility: When valuing trillions in future value on exponential growth rates, markets react hypersensitively to small data points and show large volatility. But fundamentals haven't shaken at all. Surprising GPU economic lifespan extension (3~5 years ➔ 6~8 years): Previously GPU useful life was estimated at 3~5 years for depreciation, but exploding inference demand stretched lifespan to 6~8 years. Even old chips (A100, H100) that should have finished depreciation are currently renting at premium prices, giving equipment owners unexpected upside surprise. VCs and accelerators are even buying leftover computing power directly to lure startups with it as bait, showing how scarce it is. Supply glut (Glut) impossible because: In physical world (Atoms) gigawatt (GW)-level power and

    86