The cost of AI (pt1)

Prelude

Chatting with AI about AI and its cost. This time the energy cost of using some features of AI. This is a quick and dirty experiment and should be entertained as such, I did not validate any of the AI’s citations nor did I studied this myself. Please don’t use this as a source…

As a trigger I asked and AI (Copilot) to change the style of an image. Then asked it to to help me figure out how much energy we would be spending (wasting) on such endeavor. The following is the relevant excerpt of that conversation.

Conversation

Me:

can you calculate how much energy you’ve spent to create this image?

Copilot:

Generating a single AI image—like the cyberpunk watercolor pirate you requested—typically consumes between 0.01 kWh and 0.29 kWh of energy, depending on several factors such as:

Model complexity (e.g., Stable Diffusion XL vs. smaller models)

Image resolution

Infrastructure efficiency (cloud vs. local GPU)

Prompt length and content (though this has minimal impact) ¹.

To put that into perspective:

0.01 kWh is roughly the energy used to charge a smartphone to about 10–15%.

0.29 kWh is closer to a full smartphone charge or driving about 0.5 km in a gasoline-powered car ³.

So, your image likely used about as much energy as charging a phone once, or less if a more efficient model was used.

Would you like to estimate the carbon footprint too, or explore ways to reduce the environmental impact of AI usage?

References

1: [2506.17016] The Hidden Cost of an Image: Quantifying the Energy …

2: AI Image Generation: Cost and Speed Estimates for 2025-2026

3: Making an image with generative AI uses as much energy as charging your …

---

Me:

can you calculate how many of theses requests AI models perform every hour?

Copilot:

Globally, AI models generate an astonishing number of images every hour. Here’s a breakdown based on the most recent data:

🌍 Global AI Image Generation Volume

34 million AI-generated images are created every day⁴.

That translates to approximately 1.42 million images per hour, or about 395 images per second.

🧠 Platform-Specific Insights

DALL·E 2: Generates around 1,389 images per minute ⁵.

Other text-to-image platforms (like Midjourney, Stable Diffusion, Adobe Firefly): Generate 19,028 images per minute combined ⁵.

That’s a total of ~20,417 images per minute, or ~1.22 million per hour from just these platforms. 🖼️ Cumulative Impact

Since 2022, over 15 billion AI-generated images have been created using platforms like Stable Diffusion, Midjourney, DALL·E 2, and Adobe Firefly ⁶.

So, if you’re generating a single image, you’re part of a global wave of over a million creations every hour—a scale that rivals the entire history of photography in just a few years. Would you like to estimate the total energy or carbon footprint of this global activity next?

References

4: The Generative AI Creative Economy: Stats and Trends in 2025

5: How much AI data is generated every 60 seconds? New report reveals …

6: AI Image Statistics: How Much Content Was Created by AI

---

Me:

can we you provide some comparisons to what that amount of energy is equivalent to?

Copilot:

Absolutely! Let’s compare 12,200 kWh—the estimated hourly energy usage of global AI image generation—to everyday energy uses:

⚡ What Can 12,200 kWh Power?

Activity	Energy Use	Equivalent to 12,200 kWh
Home electricity	~30 kWh/day per household (EU avg)	Powers ~406 homes for a day
Electric car (EV)	~0.15 kWh/km	Drives an EV ~81,300 km
Smartphone charging	~0.01 kWh/charge	Charges ~1.22 million phones
Laptop usage	~0.05 kWh/hour	Powers ~244,000 hours of laptop use
Refrigerator	~1.2 kWh/day	Runs ~10,166 fridges for a day
Air conditioning	~3 kWh/hour	Runs ~4,066 hours of AC

🌍 Environmental Comparison

Carbon footprint: If powered by fossil fuels, 12,200 kWh could emit ~6 metric tons of CO₂ (based on 0.5 kg CO₂/kWh). That’s roughly the same as: Flying round-trip from London to New York 6 times Driving a petrol car ~24,000 km

Images Summary