The Perfect Solution to the Generative AI Environmental Crisis? Found It!

Mother Nature is being destroyed by Generative AI. What can we do about it?

Generative AI and the Environment

Generative AI is an environmental disaster unfolding as we speak.

When entire countries are struggling for water, and the world is suffering from the catastrophic effects of runaway global warming:

The installation and operation of LLMs is guzzling fresh-water and emitting CO2 like crazy, almost without limits.

The greed of Microsoft, OpenAI, xAI, Meta Platforms, and even Google is killing the earth itself.

This is corporate greed at its worst.

Prioritizing profits in developed countries at the expense of hundreds of millions in the poorest.

Here are some statistics about Large Language Models and their destructive threat to the environment.

OpenAI o1 (and soon, o3) is perhaps the worst offender.

But it’s not all gloom and doom—there is a solution to the problem, outlined in this very article!

A viable, economical (as estimated so far), and perfect solution!

But - the statistics first:

The italized section below alone was generated by Perplexity.ai (yes, I am aware of the irony). Sources are listed at the end of this article.

The Critical Environmental Costs of LLMs and Estimates for 2030

Energy Consumption

• Annual Energy Demand: By 2030, AI systems are projected to consume around 1,400 terawatt-hours (TWh) of electricity. This is equivalent to the total annual energy consumption of approximately 125 million average U.S. households, which represents about 100% of all U.S. households based on an average household consumption of 11,000 kWh per year.

  
  

• Daily AI Query Energy Use: If ChatGPT were to handle 1 billion queries daily, the energy required could reach around 10 TWh annually, comparable to the annual electricity consumption of about 1.5 million EU citizens.

Carbon Emissions

• Projected Emissions*: By 2030, data centers supporting AI operations are expected to emit approximately 2.5 billion metric tons of CO2 equivalent annually. This figure is roughly equivalent to the total annual emissions from countries like Brazil or Canada, which reported emissions around 2.4 billion and 2.5 billion metric tons respectively in recent years.*

  
  

• Single Model Training Emissions*: Training a single advanced AI model could produce between 20,000 and 30,000 metric tons of CO2 equivalent, comparable to the lifetime emissions of about 4,500 cars, assuming an average car emits around 4.6 metric tons annually.*

Water Usage

• Daily Water Consumption: Data centers are projected to consume around 10 million gallons of water daily by 2030, which could meet the daily water needs of approximately 90,000 people based on an average daily consumption of 110 gallons per person.

  
  

• Water Use for Model Training: Training a single generative AI model may require up to 1 million liters (approximately 264,000 gallons) of water, enough to manufacture approximately 400 electric vehicles, given that producing one electric vehicle requires about 2,500 liters (660 gallons) of water.

E-Waste Generation

• Projected E-Waste by 2030: Cumulative e-waste from generative AI is expected to reach around 16 million tons by 2030, which is comparable to the total e-waste generated by countries like Germany, which produced about 1.5 million tons in 2019 alone.

  
  

• Global E-Waste Context: The global e-waste generation is projected at around 74 million metric tons annually, making AI's contribution significant as it could represent about 22% of total global e-waste generation by 2030.

This is Horribly Alarming! Is There A Solution?

I don’t have to educate you on the effects of global warming, climate change, and CO2 levels.

That has been rehashed in detail, elsewhere.

Practically all of us know about the very real risks of:

• Temperature increases
• Ocean-level increases
• Erratic rainfall
• Increased flooding
• Increased droughts
• More powerful cyclones
• Shortages of drinking water
• Environment effects of e-waste

Many of us have even experienced some of the above first-hand!

Note: Google that last point about e-waste if you don’t know what e-waste is.

I want to give you information that is not found elsewhere on the Internet easily and is just starting to pick up traction.

But I have good news for you, the average Generative AI user.

There is a solution.

A perfect solution.

And it’s even viable for big tech.

Space-Based Data Centers!

There are new startups that are focusing on installing data centers in outer space, orbiting the earth!

This is a viable solution, especially with startups like SpaceX, which use reusable rockets, drastically reducing the costs required to install and maintain data centers in space.

This is the best viable solution, and big tech CTOs and CEOs must prioritize the environment.

Advantages of Outer Space for Data Centers

1. Abundant Source of Free Energy

   Without the atmosphere of the earth, the sun is a fantastic source of infinite energy available for free. A few solar panel, and boom: all the energy you will ever need, for life.

2. There’s Plenty of Space in Space

   Instead of buying real estate, tech billlionaires could focus on key points in space without having to worry about—well, space. Despite the huge amount of space debris orbiting the earth, space is so vast that there is plenty of room for data centers.

   
   

3. Free International Connectivity

   The rotation of the earth will ensure that the data center is accessible to almost every continent on earth. No more need for additional expensive undersa optical fibre cables!

   
   

4. Natural Cooling

   Space is cold—very cold. The challenge of cooling the chips at work will not be a problem in space. No more heavy consumption of fresh water! And no worries about CO2 emissions.

5. High Safety and Reliability

   It is not easy to reach a particular point in space. Space data centers do not need to worry about natural disasters, political upheavals, societal problems, or environmental problems. This is a huge advantage of over data centers on earth and systemic reliability!

6. Savings Over Time

   One authoritative article published says that operating a data center in space could save a $140M+ USD, a savings of 95%, over a period of ten years (source listed at the very end of this article). This is a reason why a data center in space is so attractive for more than environmental reasons.

However, there are a few disadvantages.

This article would be incomplete without listing them.

Disadvantages of Space Data Centers

1. High Costs

   The initial costs of installing a single data center in space is huge. That is also a problem of logistics and synchronization with existing space launches to decrease costs.

   
   

2. Maintenance

   If something goes wrong with the hardware systems, the costs of fixing the error may be so high that it may be less costly to simply launch another data center in the place of the malfunctioning components.

3. Technical Challenges

   Operating in the vacuum of space is not easy. Radiation, space debirs, the extreme cold, cosmic rays, and many other issues have to be overcome and planned for before a working system can operate successfully.

   
   

4. Legal Complexities and Regulation

   There is still a lot of complexity over who owns what part of space and where satellites can be placed. Many countries are still working out this thorny problem. Unless regulation is addressed, space data centers will still remain a theory without any applicability.

5. Scalability

   Scalability is a huge issue for data centers. Because of the huge upfront costs, space is a difficult option to scale easily.

Startups Already Operating In This Sector

There are startups that are looking to take advantage of this environmental disaster.

Some of them are:

Lumen Orbit

Lumen Orbit, founded in 2024 in the USA, aims to build modular data centers in space.

They have raised $11 million in funding and plan to launch a prototype satellite in May 2025.

Most importantly, their operational cost estimates suggest significant savings compared to traditional data centers.

The article below estimates $167M USD to operate a data center on earth for ten years versus $8M USD for ten years for space-based data centers.

There is genuine reason to be enthused here.

For more, see:

Axiom Space

Axiom Space is partnering with Kepler Space and Skyloom to establish an orbital data center on its first space module.

The launch date is expected to launch between 2026 and 2027.

They plan to leverage the unique conditions of space for AI applications.

For more, see:

I strongly expect more startups to emerge in this sector very soon.

The savings are too high to expect otherwise.

Saving an estimated 140M USD per data center over 10 years is not to be sneezed at!

Especially when we are looking at building over a hundred data centers!

Per tech giant!

And this is not accounting for human, ecological, politcal, and social disasters!

To which the space-based data centers are immune!

Conclusion

This is the perfect solution to the environmental problem of Generative AI.

Saving an estimated 140 million USD over ten years offsets all upfront costs of launching data centers into space.

Space-Based Data Centers that are solar powered are the best way to turn the Generative AI revolution green.

It is a viable and effective solution.

And cost-effective to boot!

Google, OpenAI, Microsoft, Meta Platforms, xAI, Alibaba are you reading this?

China should be very strongly interested in this as well!

And shout-out to @Aishwarya Srinivasan whose LinkedIn post (at this link: Aishwarya Srinivasan on linkedin.com)

was the original inspiration for this article!

This is exactly what the environment needs.

A shift from Earth-based data centers to space-based data centers could save the earth for our children (and their children as well).

Is not that single fact worth making this effort?

Not to mention a 90% cost reduction in operating costs over ten years (see this link for more!)

I hope this is the start of the Green Revolution for Generative AI.

Worldwide.

Now wouldn’t that be something!

References

1. The Sustainable Agency - Environmental Impact of Generative AI | Stats & Facts for 2025: https://thesustainableagency.com/blog/environmental-impact-of-generative-ai/

   
   

2. Statista - Environmental impact of AI - statistics & facts: https://www.statista.com/topics/12959/environmental-impact-of-ai/

   
   

3. Harvard Business Review - The Uneven Distribution of AI's Environmental Impacts: https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts

   
   

4. United Nations Environment Programme - AI has an environmental problem. Here's what the world can do about it: https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about

   
   

5. MIT - The Climate and Sustainability Implications of Generative AI: https://mit-genai.pubpub.org/pub/8ulgrckc/release/2

   
   

6. World Economic Forum - Generative AI energy emissions: https://www.weforum.org/stories/2024/07/generative-ai-energy-emissions/

   
   

7. Scientific American - Generative AI Is Poised to Worsen the E-Waste Crisis: https://www.scientificamerican.com/article/generative-ai-could-generate-millions-more-tons-of-e-waste-by-2030/

   
   

8. Teck Nexus - Generative AI Could Produce Massive E-Waste Equivalent by 2030: https://tecknexus.com/generative-ai-could-produce-massive-e-waste-equivalent-by-2030/27/

   
   

9. The Potential and Challenges of Space-Based Data Centers https://newspaceeconomy.ca/2024/06/24/the-potential-and-challenges-of-space-based-data-centers/

   
   

10. Are Space-Based Data Centers the Future or Science Fiction? https://www.datacenters.com/news/are-space-based-data-centers-the-future-or-science-fiction

    
    

11. Could we put data centers in space? https://phys.org/news/2024-06-centers-space.html

    
    

12. Data centers in Space https://www.linkedin.com/pulse/data-centers-space-kaneshwaran-govindasamy-sx6nc/

    
    

13. This Startup Wants to Tackle AI Energy Demands With Data Centers in Space https://www.pcmag.com/news/this-startup-wants-to-tackle-ai-energy-demands-with-data-centers-in-space

14. Axiom Station: A Breakthrough Orbital Platform https://www.axiomspace.com/axiom-station

15. Lumen Orbit White Paper: https://lumenorbit.github.io/wp.pdf