Public trust in AI is falling fast while Silicon Valley keeps spending hundreds of billions, setting up a possible bubble crash that would hurt investors and the broader economy.

• The US has poured $285 billion into AI, but 73% of industry insiders are excited while only 10% of the public is — the biggest expert-public gap since 2008.
• AI nails hard tests like the bar exam and math olympiads but still misreads analog clocks half the time and invents fake court cases for lawyers.
• A new Apple study found AI doesn't actually reason — it just pattern-matches, and falls apart on multi-step problems like the Tower of Hanoi.
• Data centers for AI are eating huge amounts of power and water, and demand from AI firms has pushed RAM chip prices up 2000%, hurting gamers and regular buyers.
• The US leads in AI investment but ranks around 24th in actual workplace adoption, far behind countries like Singapore, the UAE, and France.

Outlook: A bad AI model report or a public disaster could trigger a sell-off and a 2008-style crash, and the worse scenario is governments wiring AI into power grids and militaries before fixing its flaws.