US EditionPublished • loading... • Updated
AI's $400 bn problem: Are chips getting old too fast?
AI chips lose up to 90% of their value in 3-4 years due to high failure rates and rapid tech advances, challenging traditional six-year asset lifespans.
- This year, the tech industry plunked down about $400 billion on specialized chips and data centers, while Mihir Kshirsagar, Princeton University researcher, and Michael Burry, investor, say chip lifespans may be two or three years.
- Rapid generation turnover and intense usage combine to shorten AI processor life, as chip makers release faster, more powerful models and AI workloads cause chips to run hot and fail.
- Market data show steep depreciation as chips lose 85 to 90 percent of value within four years, while Nvidia's Blackwell and Rubin generations deliver leaps sidelining older models.
- Heavily indebted AI specialists face added danger because some loans use chips as collateral, risking financial strain if lifespans shrink to two or three years, analysts say.
- Analysts warn the fallout could ripple through an AI-dependent economy, while some companies plan to resell or repurpose older chips to mitigate losses.
Insights by Ground AI
47 Articles
47 Articles
AI's US$400bil problem: Are chips getting old too fast?
NEW YORK: In pursuit of the AI dream, the tech industry this year has plunked down about US$400 billion on specialised chips and data centres, but questions are mounting about the wisdom of such unprecedented levels of investment.
·Malaysia
Read Full Article
Daily Journal+43 Reposted by 43 other sources
AI's $400 bn problem: Are chips getting old too fast?
In pursuit of the AI dream, the tech industry this year has plunked down about $400 billion on specialized chips and data centers, but questions are mounting about the wisdom of such unprecedented levels of investment.
·Cherokee County, United States
Read Full ArticleCoverage Details
Total News Sources47
Leaning Left3Leaning Right7Center13Last UpdatedBias Distribution57% Center
Bias Distribution
- 57% of the sources are Center
57% Center
13%
C 57%
R 30%
Factuality
To view factuality data please Upgrade to Premium
