2020 has been an incredible year for Scaleway with the launch of more than 20 new services and features. We also worked hard to prepare for the upcoming turning point in our GPU/AI ecosystem which will take place in 2021 — stay tuned for that! And, as for each product and feature that we develop, we aim to build a Cloud that makes sense.
Building AI services that make sense
With the race to AI that we are currently witnessing, all companies want to build the most performant algorithms. To do so, they need to train larger and larger models, usually needing two types of fuel:
- Compute power
As the AI market is growing, the need for these two fuels will increase. Numerous studies and articles have been pointing out that while the AI new frontier has been opening new opportunities, it has also been opening a new type of competition based on pure power consumption. Creating a whole new thirst for computing power.
An OpenAI blogpost from 2018 already stated back then that the amount of compute power used in the largest AI training runs has been increasing exponentially with a 3.4 month doubling time (by comparison, Moore’s Law had a 18-month doubling period). An exciting growth potential. But this competition for compute power has an impact on our Carbon Footprint.
This competition for compute power has an impact on our Carbon Footprint.
A recent study from the Allen Institute for AI argued for the prioritization of “Green AI” efforts, focusing on the energy efficiency of AI systems. They caused a lot of buzz after publishing their Green AI manifesto, advocating for more transparency in reporting the environmental and energy costs, which are often hidden — or at least, not “marketed” — in AI products and scientific publications.
But that’s where we are working to help you.
Sustainable AI at Scaleway
At Scaleway, we are working to provide you with the best fuel for your training model.
Upcoming and exciting products are coming for data valorization. Talking about computing power, world-class, affordable and brand new GPUs are also in our roadmap. All the while keeping in mind the environmental, energy and water costs — water being used for cooling purposes.
This is why we are delighted to bring a major contribution: our highly acclaimed P100 NVIDIA GPUs are now available in the eco-AZ fr-par-2, our carbon neutral neutral & environmentally efficient region.
With this new AZ, we want to help you make what’s good for your business, all the while minimizing the environmental impact of GPUs that are known for being extremely energy intensive. By doing so, we are hoping to help you gain market shares, based on innovation and values we care about.
This lower environmental impact is the result of different and drastic measures combined:
- Controlling energy and water consumption with a weighted index that goes well beyond the simple PUE indicator
- Defending the circular economy and banning toxic products in our data centers
- Defending renewable energy standards by powering our data centers with 100% renewable energy since 2017, mainly using hydropower
Last but not least, with an innovative and trademarked adiabatic process that allows us to operate without air conditioning, we are drastically reducing power consumption. While controlling our water consumption, often forgotten by major providers.
we are really excited to have you join our effort for a better, more efficient way to launch AI projets. You can learn more about our policies here: “the most environmentally efficient and transparent cloud worldwide”. We will also be delighted to help you build the future of AI through our startup program.