AI outbids everyone
- Data-center spending > $700B this year
- AI memory (HBM) uses ~3× the wafer per bit
- DRAM + NAND up 300%+ vs a year ago
- Relief only when new fabs ramp: late 2027–2028
Chips alone won't cause a crisis. 2026 is risky because three shocks are stacked — AI outbidding everyone for memory, the Iran-war energy shock, and tariffs — while central banks raise rates.
The money isn't disappearing — it's moving from people who buy devices to the companies that make chips and the shareholders of AI firms.
The catch: what would make gadgets cheap again (an AI slowdown and a memory glut) is exactly what would crash AI stocks. Relief for one side means pain for the other.
Any one of these alone would be manageable. Together, while interest rates rise, they create real stagflation risk.
Price changes announced or forecast in 2026. Bars scale to the largest item in the current filter.
Industry revenue can hold up even as unit sales collapse — budget buyers and budget brands get pushed out.
Click each step to see the term economists use and the 2026 evidence.
Established ideas from economics, each applied to what's happening now.
Scroll sideways → · solid = happened · dashed = expected
My judgment, not a forecast. Pick a scenario to see what happens to each part of the economy.
Indicators that show which scenario is unfolding.
| Signal | Now | Relief if… | Danger if… |
|---|
The price system is routing scarce chips to the highest bidder: AI. Costs land on consumers, especially the lower half; gains land on chipmakers and shareholders. Chips alone mean pricier gadgets and slower progress, not a crisis. Real crisis risk comes from energy shock + high rates + AI-concentrated markets. Don't expect 2025 prices back soon — waiting for a crash is a bet on scenario 2.