U.K. companies are facing some of the highest industrial electricity prices in the developed world, and AI’s growing energy demands are sparking concerns over how companies can invest in the tech without jeopardizing sustainability. The problem, Greg Holmes, field CTO of IBM-owned Apptio, says in this Q&A, is compounded because organizations have little to no visibility into how much energy their AI workloads consume, or where that energy even comes from.
Apptio’s platform, acquired by IBM in 2023, looks to meet the challenge by helping organizations track the financial and resource costs of running data across hybrid environments.
Holmes argues that greater visibility doesn’t just help companies cut costs but can also make the case for further AI investment by showing where the technology is actually delivering value.
Why is energy consumption a key consideration for companies deploying AI?
Greg Holmes: From a cost perspective, AI is one of the factors pushing up energy demand because it’s a very high-demand item. The way most organizations consume AI is also a little bit at arm’s length. If you’re making a simple query in Claude or ChatGPT, for example, it’s not clear how much energy that consumes, or even where that energy comes from.
Most firms have a lot of choices when it comes to how they want to do AI. They can build platforms on premises, use infrastructure-as-a-service through the cloud, or access AI platforms directly or consume AI through SaaS products. There are a lot of different ways they can consume AI, and very little control for organizations in terms of knowing how that translates to energy consumption.
A lot of the emissions are also outside of direct control. But it’s still one of the key factors in determining how sustainable one use of AI is compared to others.
Are there any regulations governing renewable energy use in AI and cloud infrastructure?
Holmes: At the moment, the focus is really on disclosure.
Organizations are increasingly required to report on their ESG and sustainability performance, and those requirements are extending to larger and larger firms. By the end of this year, pretty much all organizations should be disclosing their sustainability position and consumption.
To be able to do that effectively with their AI and cloud workloads, they need to get better, more reliable information from their suppliers, and then they need to make a judgment call on how sustainable their energy consumption actually is.
How can companies manage AI’s environmental impact given this limited transparency?
Holmes: One approach is to estimate the energy use of workloads based on available information. That’s something we’re able to help customers with, whether they’re using cloud providers or AI services.
By combining workload data with information from sources such as energy maps, organizations can get a better understanding of the sustainability of their energy consumption. However, that picture can vary significantly depending on where a workload is running.
It differs across countries, across regions and even between different availability zones within the same cloud provider. All of those factors affect the sustainability profile of the energy being used.
The figures won’t be perfect, but they provide an order-of-magnitude understanding that helps organizations make better decisions. For example, some workloads could be moved to different locations or run at different times of day if that improves sustainability without affecting performance.
How does the U.K. compare with other countries in sustainable energy?
Holmes: I’d say the U.K. sits somewhere in the middle. Energy costs are relatively high, but the U.K.’s energy sustainability is better than in many countries. It’s not the best — Scandinavian nations and other countries like France tend to perform particularly well — but it’s also certainly not the worst.
There is a lot of work being done through government policy to improve the sustainability of energy consumption. At the same time, however, we’re seeing growing demand for new data centers, many of which still rely on more carbon-intensive backup power sources. That’s why it’s important to remember that not all energy is equal.
Organizations only have so much influence over how energy is produced, but larger firms can make a difference. They can invest in their own energy generation, make sustainability a factor in procurement decisions and include it in RFP requirements. If large organizations create stronger demand for renewable energy and sustainable sourcing, energy providers will respond by producing more of it.
What are the risks of AI adoption if factors such as energy go unmanaged?
Holmes: Governance is a major concern. Organizations need to think about whether the data they’re using is appropriate from a legal and compliance perspective. Are they processing sensitive customer information? Is data being stored in locations where it shouldn’t be? Is it critical to the operation of the business?
Those governance issues sit alongside concerns about uncontrolled energy consumption and sustainability.
Can greater visibility into AI usage change behavior?
Holmes: Absolutely. If people understand how much a workload costs to run, along with its sustainability and ESG impact, they’re more likely to make informed decisions. If people don’t see the impact of their actions, they can’t realize how it contributes to the organization’s energy consumption.
At the same time, that information needs to be connected to business value. If AI is saving significant amounts of work, reducing travel or replacing more energy-intensive processes, then it can deliver a positive overall outcome.
In some cases, greater visibility could actually support the case for greater AI adoption, as it helps organizations understand where the technology is delivering measurable value.
Looking ahead, what trends are you watching?
Holmes: We’re already seeing major organizations place greater emphasis on sustainability and transparency. Companies are investing in projects that provide greater visibility into resource consumption, and sustainability is increasingly becoming a factor in strategic planning.
We’re also seeing significant growth in data center construction. U.K. capacity is expected to increase substantially over the coming years as new facilities come online. Some of those facilities may be considerably more energy-efficient than the data centers they will eventually replace.
In general, though, this conversation should become more of a board-level issue. Historically, technology costs and energy consumption associated with technology have been viewed primarily as IT concerns. But for many organizations, technology is becoming one of the largest sources of Scope 2 emissions.
If CEOs, CFOs and boards care about reducing emissions, then technology-related energy use needs to matter to everyone, not just the technologists.
Editor’s note: This interview has been edited for clarity and conciseness.
