AI Stats
The numbers behind the AI revolution. Real data, beautifully visualized. Click any card to share.
⚡
Energy & Environment
GPT-4 Training Energy
4,600homes/year
Training GPT-4 used enough energy to power 4,600 homes for a year
4,600 US households for 1 year
Source: Epoch AI (2024)racetoagi.org
AI Water Consumption
700Kliters
per 100 conversations
Every 100 ChatGPT conversations use 700,000 liters of water
Per 100 conversations (data center cooling)
Source: UC Riverside / Nature (2023)racetoagi.org
Training CO₂ Footprint
500flights
NYC ↔ SF round-trips
Training one AI model emits as much CO₂ as 500 cross-country flights
500 round-trip flights NYC ↔ San Francisco
Source: Strubell et al. (ACL) (2019)racetoagi.org
💰
Economic Impact
AI Investment Surge
$40B
💵💵
2022
$65B
💵💵💵💵
2023
$100B
💵💵💵💵💵💵
2024
↑ 150%
AI startup investment hit $100B+ in 2024, up from $40B in 2022
150% growth in 2 years
Source: PitchBook / CB Insights (2024)racetoagi.org
Jobs Transformed by AI
Current
Transformed
300Mjobs
300 million jobs worldwide will be transformed by AI automation
Exposed to AI automation globally
Source: Goldman Sachs Research (2023)racetoagi.org
AI Market Explosion
9x growth
AI market will grow 9x from $200B to $1.8T by 2030
$200B (2024) → $1.8T (2030)
Source: Grand View Research (2024)racetoagi.org
🤖
Model Comparisons
The Parameter Race
AI models grew 10x in size: 175B → 1.76T parameters in 3 years
GPT-3 (175B) → GPT-4 (1.76T estimated)
Source: SemiAnalysis (2023)racetoagi.org
Training Cost Explosion
AI training costs jumped from $4.6M to $100M+ per model
$4.6M (GPT-3) → $100M+ (GPT-4) → $500M+ (future)
Source: Lambda Labs (2023)racetoagi.org
Benchmark Showdown
Claude 3.5 edges out GPT-4o on MMLU: 88.7% vs 88.0%
Claude 3.5: 88.7% | GPT-4o: 88.0% | Gemini Ultra: 83.7%
Source: Papers with Code (2024)racetoagi.org