Regulation watch: The next year of AI data center transparency

The race to build AI infrastructure is ever-growing, but with that comes backlash, critique and protest.

As public unease and anxiety around data centers continue to grow, those in charge are being held accountable in ways they haven't been in the past. As compliance expectations shift, CIOs must understand what is changing and what is coming their way to prepare their infrastructure accordingly.

In this Q&A, Subodha Kumar, founding director of the Center for Business Analytics and Disruptive Technologies and a faculty member at Temple University, spoke with TechTarget to share his insights on how CIOs should tackle the evolving regulatory landscape. Kumar researches AI, AI policy, sustainability in IT and other topics.

Editor's note: This Q&A has been edited for clarity and conciseness.

What are the most significant AI data center transparency regulations taking effect in the next 12-18 months, and how do the EU and U.S. approaches differ?

Subodha Kumar: There is a different approach from the EU vs. the U.S., and what we are already seeing is that the EU is ahead in the game with some of the reporting guidelines. They also have the Energy Efficiency Directive, where operators above a certain demand must report their energy performance and their sustainability indicators.

The European Commission is also preparing a broader data center energy efficiency package, where they have a rating scheme and a minimum standard.

In the U.S., what we are going to see in the next 12 to 18 months will be more state-wide. This is what the U.S. has always done, and it means different states will be following different guidelines. The U.S. approach is more fragmented, while in the EU, things are more centralized and very metrics-based.

The focus here in the U.S. is more around who is bearing the cost. The biggest protest right now is that the gain is going to the organizations. They are making money, but the cost is incurred on society, and that's where the regulations are coming.

If you are setting up a data center, most of the regulations recently -- for example, in New Jersey -- will be on more than just reporting. It is about how you ensure the cost to society does not go up.

What do you mean when you say 'cost to society'?

Kumar: Many states are seeing an increase in energy and water prices lately, and that has been a big concern. Now, how much of that increase relates to data centers is not that obvious, but communities feel these data centers are either responsible or will become responsible for it in the next 12 to 18 months.

Water use is humongous, and it has to come from somewhere. Energy use is big, and it has to come from somewhere. There are also intangible costs, like increasing noise and traffic in the area. Who's going to build the highways to cover for that?

What specific transparency disclosures will AI data center operators need to provide?

Kumar: Operators will have to report on things like power usage effectiveness (PUE), the ratio of total facility power divided by IT equipment power. The lower the number, the better. If the PUE is between 1.1 and 1.2, this is considered excellent, and anything above 1.7 is considered poor.

The second biggest thing is water use. If you think of large data centers, which companies like OpenAI are running, they can go through more than a million gallons of water per day. So, water usage effectiveness (WUE), which is also defined as annual water use divided by the IT equipment energy, is another key disclosure.

And the third big concern is space usage effectiveness.

There are some other metrics that CIOs and other executives should look at. For example, carbon usage effectiveness, renewable energy factors, energy reuse factors and carbon intensity.

What are the biggest challenges organizations face when collecting and validating AI data center transparency data?

Kumar: There has been a mad rush when it comes to creating these data centers and being ahead of the game. There's tough competition between OpenAI, Anthropic, Microsoft and Google. Whenever this IT growth happens -- and we have seen this before in the late 2000s with the dot-com boom -- we start focusing on efficiency and forget about everything else.

The system is not built for that, and it will require resources to create this data and report it accurately in a timely manner, so how to allocate your resources will be a challenge.

Another big challenge is balancing intellectual property and transparency. Companies don't want to disclose how they are operating. Executives have to be very careful when balancing transparency with security and separating operational transparency from sensitive architectural information.

The challenge is getting what is needed for compliance, how they can use it to improve their efficiency in the long run, and protecting their intellectual property.

What sustainability transparency requirements are emerging, and how do they differ from traditional data center regulations?

Kumar: Some regulations, like carbon efficiency, are not all that different. What is different is that we never bothered much about electricity and water. You have to understand why we did not have regulations around that. If you look at the older enterprise data centers, they were using much less electricity and much less water.

Previously, data centers were actually not very efficient when it came to some of the measures which have now been implemented, but we did not have policies and regulations then because the impact was much smaller. Now, we are talking about data centers as large as Philadelphia International Airport, so the scale has increased and the impact becomes much larger.

What sustainability metrics and KPIs should organizations track to understand and communicate the environmental impact of their AI infrastructure?

Kumar: Whatever is coming from the EU is the best place to start. Carbon intensity is very important, as well as energy reuse factors and renewable energy factors. Of course, PUE and WUE are very important, and are where CIOs should start.

This is not just the task of the CIO, because this should be a cross-functional approach within the organization. Organizations should bring in their chief compliance officer and their chief sustainability officer.

What are the potential business, financial and reputational consequences of failing to meet emerging AI data center transparency requirements?

Kumar: The risk is not just the fines. The biggest problem will be non-compliance, which can lead to project delays, some loss of tax incentives and permit denials. They may have higher utility cost obligations in the long run. Community opposition, which we are already seeing, is another challenge.

You can't forget about the reputational part, as a poor reputation can create damage. All of these data centers [face pressure from] investors right now. Any negative publicity can create broad concerns, so they have to look at the bigger picture and not just at the numbers.

Looking ahead, what advice would you give CIOs for 2027 and beyond?

Kumar: For 2027, the immediate concern is to bring these teams together as soon as possible. Do not delay on that. Any delay would cause a big concern.

In the longer term, CIOs have to look for more innovative ways to deal with this problem. For example, look at how China is putting a data center in the ocean.

Transparency regulations should be treated as an opportunity rather than a burden. If these regulations have been properly considered and included in culture and strategy, there will be long-term impacts for several decades.

The immediate thing should be to get your act together and put the teams together. Reduce impact where possible and improve these measures wherever needed. In the long run, find innovative ways to improve your efficiency while remaining compliant, talk to people and listen to the community. Don't just sit in your room making decisions.

Rosa Heaton is a content manager and writer for the IT Strategy team at TechTarget.