Rickards Warns $5 Trillion AI Buildout Risks Dot-Com-Style Bubble

1 articles · Updated · The Manila Times · Jun 28

$5 trillion could be spent on U.S. AI data centers alone, Jim Rickards said, arguing the sector is already consuming more capital and energy than the dot-com boom and may be forming a bubble.
80 million miles of fiber laid in the 1990s left about 85% unused; Rickards says AI differs because data centers, chips, power and cooling face hard physical limits that software booms did not.
George Noble, Tim Dettmers and Marc Andreessen are cited as seeing diminishing returns, with each incremental AI gain requiring far more compute, money and electricity and hardware optimization potentially hitting a wall by 2026 or 2027.
Lucent fell from $75 to $0.76 and Cisco from $50 to $8 after dot-com financing loops unraveled; Rickards says similar circular revenue may be emerging as AI investors fund startups that then buy their chips and cloud capacity.
July 29 could be a key test, with Meta expected to update guidance around then; Rickards says disappointing AI earnings this summer could challenge valuations much as March 2000 warnings preceded the Nasdaq's nearly 80% collapse.