AI winter
- Definition
- A period of reduced funding, hype, and progress in AI research. The field experienced major winters in the 1970s and late 1980s. Investors watch for signs of a new winter whenever growth narratives stall.
- Why it matters
- Understanding AI winters is essential for calibrating investment timelines and strategic bets. Previous winters were triggered by overpromising and underdelivering: expert systems failed to generalize in the 1980s, and neural nets hit compute limits in the 1990s. The current wave has more commercial traction than any prior cycle, but the warning signs to watch are the same. If enterprise ROI from AI fails to materialize at scale, if model capability gains plateau, or if regulatory headwinds choke deployment, capital could flee. Smart companies build AI strategies that survive a winter: focus on measurable ROI, avoid dependency on cutting-edge capabilities, and keep infrastructure flexible.
- In practice
- Sequoia Capital published a widely-discussed analysis in 2024 asking 'Where's the revenue?' noting that AI companies had raised $50B+ but generated only $3B in revenue. The framing echoed pre-winter dynamics. However, by 2025, AI revenue grew significantly with companies like OpenAI exceeding $5B ARR and Anthropic crossing $1B ARR, reducing winter fears. The market bifurcated: AI infrastructure companies thrived while many application-layer startups struggled with retention and unit economics. History suggests AI winters are industry-wide events, but this cycle's commercial adoption depth may make a full winter less likely.
We cover business & strategy every week.
Get the 5 AI stories that matter — free, every Friday.
Related terms
Scaling laws
Empirical relationships showing that model performance improves predictably as you increase data, compute, and parameters. Scaling laws are why labs are pouring billions into ever-larger training runs.
Moat
A sustainable competitive advantage that prevents rivals from replicating your position. In AI, moats can come from proprietary data, distribution, fine-tuned models, vertical expertise, or switching costs, but raw model capability is rarely a moat.
Inference economics
The study of costs, pricing models, and margin structures around running AI models in production, encompassing hardware costs, model efficiency, pricing strategies, and the competitive dynamics of the inference market.
Know the terms. Know the moves.
Get the 5 AI stories that matter every Friday — free.
Free forever. No spam.