The Next Revolution in AI Must Be an Industry-Wide Effort to Reduce Carbon Use

Artificial intelligence models are now advancing at a pace that can be measured in days rather than years.

The US has just announced its US$500 billion Stargate program to build AI infrastructure. The data centres at the heart of the project are designed to consume massive amounts of electricity to power AI models. Each centre would use the same amount of energy to run up to 400,000 cars every year.

In the same month, the DeepSeek AI search model was introduced. The model built by a Chinese startup was able to compete with standard large-language-model AI like ChatGPT using much less computing power. This efficiency is attributed to innovative training methods and optimization techniques, which optimize memory usage and computational resources.

DeepSeek’s performance unlocked possible solutions to one of the nagging problems of AI language models: balancing performance with cost and energy efficiency.

Curbing hardware power consumption for even basic AI is one small step for the planet. Without cloud infrastructure, though, there is no AI. That’s why cloud platforms are crucial to the next round of advances that will rein in AI’s many massive energy-sucking needs. 

AI’s Growing Carbon Footprint

AI’s total carbon footprint is more like a colossal crater. While the energy demand of AI at its current scale is significant, with the right tools and responsible practices, this challenge can be turned into an opportunity to transition toward more sustainable technologies.

The timing of an AI revolution is very inconvenient, as its power consumption clashes directly with the urgent need for carbon-neutral solutions to curb climate change. The hardware and chips used to power AI will, thankfully, evolve quickly, so they use less energy than current versions. To balance innovation with climate sustainability, we must educate users about the carbon impact of this equipment, ensuring AI is used judiciously.

The most important lesson is the value of sharing data and best practices. Transparency is essential because it provides the whole industry with the tools to get the job done right the first time. Once organizations have the information on how to reduce power usage or use fewer components without sacrificing performance, the shared results of that knowledge will ripple through the industry.

Right now, however, most users and companies utilize services blindly without access to information about which ones are more or less carbon-intensive. That’s not their fault. AI providers need to come clean with their customers. 

How Cloud Providers Can Lead the Way

Cloud providers are uniquely positioned to bridge the gap between AI efficiency and sustainability. By optimizing data center operations, integrating renewable energy, and developing transparent tools to track carbon impact, they can drive systemic change in AI’s environmental footprint.

If we are to get serious about reducing AI’s environmental impact, we must collect accurate data and share it for the greater good. That means rewarding companies for participating in good faith, not critiquing them for being carbon-intensive. 

AI chips depend on critical minerals requiring energy-intensive extraction. More collaboration across the supply chain can lead to cleaner sourcing to curb environmental impact.

Cloud providers can take a leadership role in steering AI sectors toward a more sustainable carbon profile. They have the data and tools to help their customers make a serious dent in their carbon emissions and share their achievements with others in the industry without conceding competitive advantages.

Our firm, OVHcloud, has developed a carbon calculator that allows clients to understand the entire carbon footprint of their product, including other inputs, such as freight, manufacturing, cooling and waste management. Indirect emissions like these can make up a significant portion of the total carbon output. Users can also adjust their activities to lower their emission profile, such as moving heavy workloads to data centres powered by renewable energy.

Landmarks on the Road to Sustainable AI

Cloud computing tools can provide the roadmap to innovation sectors, leading them to improve their sustainability performance while maintaining growth. Here are some of the important landmarks on that roadmap:

• Educate users about the carbon impact of AI and the energy consumption of hardware. This will help them make informed choices about their AI usage and its environmental consequences.
• Promote transparency within the AI industry, encouraging cloud providers to share data on energy usage, carbon footprints, and environmental impact. This allows users to pursue more sustainable options.
• Encourage collaboration across the supply chain to reduce the environmental impact of AI, including cleaner sourcing of critical minerals used in AI chips. Extraction is a huge environmental cost in AI infrastructure, but there are rapid advancements being made to reduce it.
• Deploy transparent carbon tracking tools, like the carbon calculator mentioned above, to help companies understand, measure, and reduce their emissions. These tools allow users to track and reduce their carbon footprint in real time.
• Use renewable energy in data centre operations and enable companies to transfer their activities, such as moving workloads, to renewable-powered AI infrastructure. This can lower their emissions profile and radically curb fossil-fuel usage.
• Share best practices and data across the industry, ensuring collective progress toward more sustainable AI solutions. An open forum for collaboration and improvement helps everybody advance faster.
• Utilize cloud providers’ unique position to optimize AI functions in a way that targets energy use. This will provide leverage for the industry to pursue more sustainable practices while driving innovation in energy-efficient AI.

To achieve these landmarks, the first step is gathering and publishing data and using it to achieve lower emissions and spark a cycle of advancement based on informed decision-making.