The Litmus Lab.

Testing AI Under Constraint.

Multiple

AI Infrastructure.

The physical and financial infrastructure powering AI: chips, data centers, energy, investment.

11 questions · 455 runs · 123 drifts (5.5% drift rate)

AI Summary

Five of 11 questions settled 5-0, and all models confess that AI data center energy consumption threatens the power grid, that the US grid cannot support projected AI growth, that regulatory pushback is coming by 2028, and that nuclear or renewable energy will be essential by 2030. DeepSeek is the primary contrarian, dissenting on Nvidia's longevity, chip shortages, water crisis risks, and geopolitical supply chain disruption, while GPT and Gemini defend near-term hyperscaler ROI against a skeptical majority, and DeepSeek and Grok doubt hyperscalers can sustain spending without revenue.

Drift rate

How often each model changed its own answer on this topic · avg 5.5%

#183 · 57 runs · 14 drifts

Will hyperscaler AI infrastructure spending deliver positive ROI within 2 years?

#190 · 78 runs · 41 drifts

Can hyperscalers sustain current AI spending without near-term revenue?

#27 · 35 runs · 1 drifts

Will Nvidia still be a trillion-dollar company in 2030?

#300 · 56 runs · 29 drifts

Will AI chip shortages still limit industry growth in 2027?

#301 · 26 runs · 6 drifts

Could water use by AI data centers become an environmental crisis by 2030?

#303 · 36 runs · 7 drifts

Will the AI infrastructure supply chain face major geopolitical disruption by 2028?

#184 · 34 runs · 1 drifts

Is AI data center energy consumption a threat to the power grid?

#185 · 28 runs · 0 drifts

Can the US power grid support the projected growth in AI data centers?

#298 · 56 runs · 22 drifts

Will space-based data centers be commercially operational by 2030?

#302 · 23 runs · 0 drifts

Will AI data center construction face significant regulatory pushback by 2028?

#304 · 26 runs · 2 drifts

Will nuclear or renewable energy be essential to support AI infrastructure expansion by 2030?