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

$200B$800BAIMLLLM$1.8T20242030
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

GPT-3175BGPT-3.5175BGPT-41.76T10x

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

$$4.6MGPT-3$$100MGPT-4$$500M+Future20x5x

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

MMLU BenchmarkClaude 3.588.7%🏆GPT-4o88%Gemini83.7%🏁

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