The shower knob problem of a green cloud

It’s easy to blame technology for guzzling power and adding up to the carbon burden of the planet. But we also expect technology to be real-time, to be scalable, to be cost-savvy, and to be available every minute (and everywhere) like an invisible friend. For technology providers, these expectations can be like trying to balance hot water and cold water in the shower. Finding just the right sweet spot is not that simple. And every shower tap is a whole new beast, in itself. Bumping into Tracy Baldwin, Global Lead, AWS Sustainability at the AWS Re: Invent this year, we stumbled upon a lot of front-end and back-end action that is happening to thread this needle. Let’s get a close look at how AWS is moving these tricky levers.

Green Cloud. Can we say these two words in the same breath? Is it still a paradox of sorts?

We talk about the cloud being the most efficient most cleanly, and with our ability to scale smartly, I think we can handle this challenge. I’ve been with AWS for 11 years, focusing on sustainability for the majority of that time, so I’ve worked both on our sustainability work and then also on how we work with customers, to help them achieve their sustainability goals.

Is there a trade-off between GreenOps and FinOps—especially when FinOps has become more pronounced today due to cloud wastage issues and the problem of hidden expensiveness?

These two ops actively complement each other. We started with pillars like performance optimization and two years back we created a sustainability pillar specifically for the well-architected program. We have created initiatives to help customers turn off resources they do not need, and how to re-architect their systems if needed. Have they architected an application most efficiently, if they migrate something over without re-architecting, can they, then, take the opportunity now to optimize that for the cloud? Sometimes, customers will optimize every architecture refactor when they move over, but sometimes they don’t, sometimes they just want to get their workloads shifted into AWS. And even for a new workload, we work with them on how they build that.

But one of the interesting things I found last year was, that we did our first hybrid wind and solar projects in India, which you don’t often find with the conditions that are right for that.

Interesting. So it’s not an either-or situation?

A lot of times they go together, it’s one of those nice, non-compromise areas, because generally, if you’re turning off resources you don’t need, you are more sustainable and you’re saving cost. And the other thing is you don’t have to compare to on-premises, you don’t have to own that equipment, you don’t have to have equipment that you only use once a month or an hour a day, etc. That’s part of what we work on with them as well. You just spin up and down as needed, especially when you get into serverless.

Your thrust on renewables appears to be quite strong. Is it done in a greenfield way or a brownfield way?

We look at local climate paths and explore the most cooling-efficient alternatives. We also look at designing and customizing as per region. We make it possible through both ‘what’s available’ and ‘what to create’ routes. Come to think of it—the best and most sustainable energy is energy that we don’t use, right? We have been the largest corporate procurer of renewable energy for the last three years.

Can you elaborate? Specially for India?

Recently, we announced a new set of projects. That includes those utility-scale wind and solar farms, and also solar rooftops on fulfillment centers. In India, specifically, we’ve gone very big because we have operations in two regions in India. So we’re now at over 1.1 gigawatts of capacity in India. And we announced one large project in October, with 198 megawatts. But one of the interesting things I found last year was, that we did our first hybrid wind and solar projects in India, which you don’t often find with the conditions that are right for that. Since their load generation profiles are complementary, since solar will peak in the middle of the day, and then wind will pick up generally in the late afternoon and evening, we’re able to deliver more clean energy onto the grid, by doing those projects in combination.

Are RECs (Renewable Energy Credits) part of your strategy?

We have published our renewable energy methodology. Amazon has focused on enabling new projects, a lot of the ones that we’ve talked about. How we track the accomplishment of courses by renewable energy credits or guarantees, differs by region, of course. In some cases, we would procure renewable attributes and advance as this project is going online, especially with some projects that have been delayed over the past few years due to supply chain issues.

How crucial are metrics to fixing the carbon problem—especially when other players like IBM and Google are exploring PUE numbers and carbon calculators?

We enable our customers with carbon footprint tools and reports that help them track how they are doing and how to help them reduce their footprint. That kind of tracks how they are doing and then we focus on those optimization architecture patterns that we were talking about.

When I’m talking about efficiency, like for cooling, we would look at the most energy- and water-efficient methods. From the angle of renewable energy, we do focus on the availability of options and also try to create options.

Do you also consider embodied carbon emissions—because they form a big part of any computing system?

Yes. The core of what we focused on from the beginning is efficiency across every aspect of our infrastructure. And that’s everything from designing our data centers and our hardware to modeling and tracking performance to make sure that we continue to identify ways to increase efficiency. So that’s one of the things that you’ve probably seen at Re: Invent this time, and what is most visible is the investment in AWS chips. We announced Graviton4, which is the highest-performance, most energy-efficient chip we’ve ever built for a broad range of workloads. Then we also announced Trainium2, getting into the ML AI world, which is designed to be the highest performance, most energy-efficient way to train AI workloads in the cloud. And we also think of areas like backup generators or building data centers and hardware with a low-carbon approach. In the majority of data centers constructed this year, we have reduced emissions with this mindset.

When you start an availability zone, you go to a new country. This is a Scope 3 problem, like, you may not have control over the Greenfield setup that you’re starting. So how much can you control the limitations that you confront there in terms of the availability of alternative energy, renewable energy, and all the other associated Scope 3 issues?

We think about several things when we go into a region. For example, when I’m talking about efficiency, like for cooling, we would look at the most energy- and water-efficient methods. From the angle of renewable energy, we do focus on the availability of options and also try to create options when they don’t exist. So when I joined, there were a lot of countries where it was very difficult, either from a regulatory perspective or cost or just available lands to do renewable energy projects. So we’ve made a ton of progress in doing projects where it wasn’t easy to do before, like either you couldn’t do one directly as a corporate or some other regulatory constraint. We announced our first projects over the past three or four years in Singapore, Japan, India, and now South Korea two weeks ago, as well. One thing we focused on is not only how we can do that for ourselves, but how can we work with our peer companies, and our customers across the industry. Because that’s an industry-wide issue, right? So our policy teams collaborate with others, to figure out ways to make it more possible to do renewable energy.

Anything on recycling or upcycling?

Circular economies are important for both getting to our net zero carbon goal and for reducing waste. We think about that in three areas: design, operation, and recovery. Design is, we try to design reusable, lower carbon, server rack systems to begin with more recycled materials. And then we operate our equipment efficiently. As long as it is operationally efficient. And then the third is recovery. We have reverse logistics hubs. We take securely decommissioned rack systems, and to these hubs, and we have a process where we can test and repair components and then we can take them to the data centers where we need them. And then if they cannot be repaired, we have ways to either have companies reuse, complete components, or recycle them.

Circular economies are important for both getting to our net zero carbon goal and for reducing waste. We think about that in three areas: design, operation, and recovery.

From that lens, how do you view the whole debate around the right to repair and, a locked-in system where customers do not have the right to maintenance, maybe, shift if they want to?

We are very focused on repairing and reusing hardware, as long as it is efficient, and as long as it is delivering what it needs to deliver for customers. That also helps that we design a lot of AWS hardware as custom hardware beyond the chips that I talked about. So we are able and actually, we feed the data from our nitro cards, for example. When we go and we identify areas we need to repair, we feed all that back into our design team, so that they can continue to improve the design of those cards going forward. So there is also that feedback loop.

Can quantum computers solve the heavy-compute needs that we face today?

I cannot comment on that, but Gen AI helps a lot in the use of services to accelerate sustainability. For instance, AI is being used for work in controlling deforestation.

What is the biggest advantage and challenge that Big Tech has in addressing the carbon footprint problem?

It is going to be a collaborative effort across the industry. For the sustainability approach of AWS, I think about that in three buckets. One is that AWS is building the most sustainable infrastructure we can, to deliver services to customers. Second, how do we help customers architect sustainably, which is also a theme you’ll hear in Werner’s keynote, a lot each year. Third, how can customers use AWS services to either build solutions that help them achieve their sustainability goals, or build solutions that help solve bigger sustainability problems or help other companies do that? But overall, really, our scale helps us, and with a focus on innovation, we can be a lot more efficient than on-premises data centers. 

Written by Pratima H