Google’s head of data center energy on the urgency of expanding the grid

The world has changed radically since Google first launched its 24/7 carbon-free energy program in the fall of 2020.

The COVID-19 pandemic had already pushed internet usage way up, prompting a rapid expansion of data center capacity. But energy demand was increasing only modestly — until the AI arms race kicked off in 2022. 

This summer, Google and Microsoft both reported significant increases in both their energy consumption and carbon emissions — driven primarily by increasing data center workloads and capacity expansions. But the build-out has just begun. Hyperscalers like Google, AWS, and Microsoft are set to dominate new capacity expansions, and will account for most of future AI workloads.

Meanwhile, grid capacity constraints and an interconnection backlog for clean generation have only gotten worse. As Amanda Peterson Corio, Google’s global head of data center energy, put it, “structural issues” are presenting the biggest barriers to scaling up clean power faster.

“As we look at the load growth, we're seeing a number of challenges — and it really comes down to an underinvestment,” said Peterson Corio in an interview with Latitude Media.

Google has not backed away from its original goal of matching clean energy to its operations on an hourly basis. But we’re nearly halfway to the company’s 2030 target, and the job has certainly gotten much harder — pushing the company to develop a new tariff for emerging technologies like enhanced geothermal, as well as an orderbook for small modular nuclear plants. The company has also identified point-source CCS combined with existing fossil fuel plants as a possible solution.

I caught up with Peterson Corio to discuss some of the ways Google is tackling constraints and exploring new solutions, four years after the launch of its 24/7 carbon-free energy program. The interview has been edited for brevity and clarity. 

Amanda Peterson Corio will be speaking at our upcoming Transition-AI event in Washington, DC on December 3. She’ll be joining Chris Shelton of AES, Jeremy Renshaw of EPRI, and Page Crahan of Tapestry to talk about creative ways to meet the load growth challenge.

Stephen Lacey: What are some of the big constraints that you're thinking about right now?

Amanda Peterson Corio: We're seeing a number of challenges, and it really comes down to underinvestment in our grids. I would put that in three categories. In my mind it's an underinvestment in transmission infrastructure first and foremost. I think in 2021, we built under 400 miles of new lines in the U.S. — and we need 1,500 to 10,000 miles of new transmission lines each year. And those really need to be inter-regional lines to bring stranded generation to where load growth is happening on our grids.

I think the other big challenge that we're seeing is investing. Not enough investment has been made to date in either flexible or firm carbon-free generation resources to provide that needed capacity, which is really what we're seeing as the challenge to growth on our grids. As you know, we are making investments in some of those new firming assets, but we need a lot more if we're going to do this at scale and in a way that's clean.

I think the last thing that we're seeing an underinvestment in is the evolution of our market structures to really bring and incentivize more flexible designs for that two-way grid. There are a lot of structural issues that are really presenting challenges. 

Stephen Lacey: Do you see more creativity on the grid capacity expansion side? There's obviously a lot of renewed focus on grid-enhancing technologies, maybe reconductoring. What are some of the ways that you see possibly helping alleviate these constraints?

Amanda Peterson Corio: DOE put out a report that said grid-enhancing technologies can open up to 100 gigawatts of capacity in the U.S., and we are strongly supportive of better utilizing the grid infrastructure that is already there today through those enhancements. It's just going to be easier to do than building new transmission lines. 

We have been looking at new structures to support this. One such structure is the clean transition tariff that we worked on with NV Energy in Nevada. There, we work to design a new rate class that could put new geothermal onto the grid in Nevada — so we take the risk on the technology and the capacity showing up, but we also get the benefit of that capacity as a rate class customer. 

The same can be done to get new grid-enhancing technologies, or energy efficiency, or grid storage built into the system. How can we as a customer pay to help the utilities build this, and then pass the capacity onto customers that are willing to pay? We're working structures to see this adopted more broadly.

Stephen Lacey: When you think about the use of AI to benefit the broader grid or to modulate data center demand, how are you thinking about the opportunities for AI to manage energy itself?

Amanda Peterson Corio: I think the next frontier for AI is looking at how we can work with transmission operators and ISOs to do better grid planning and actually advance grid tools. And this is something we're looking forward to talking about at the Transition-AI conference — how can we use AI to actually help solve the grid's problems and bring a little more sophistication to accelerate trapped generation or load that can't grid interconnect online in a more sophisticated way.

Also on the load side, we continue to look at demand response applications. Our carbon-aware computing platform is already moving certain compute tasks to different times of day where the grid is cleaner. We do that without reducing any reliability to our customers. We've used this in Europe in times of intense grid stress. And we're continuing to evolve and think through how we can do that more efficiently for ourselves. I also think AI has applications to help other customers do that and figure out how this whole ecosystem can incentivize load shifting or load dropping as well.

Stephen Lacey: In the U.S., there are not just capacity issues, but also long interconnection queues that slow the ability to build generation fast enough. How significant of a challenge is that for you in trying to build this clean generation?

Amanda Peterson Corio: It's a big challenge for the U.S. If generation cannot connect to the grid fast enough because of the long interconnection study process, that just delays when we can get this new firm capacity onto the grid. So it is a big challenge.

We work with Tapestry [the Google X project focused on the electric grid] to figure out how we can work with our utility partners and ISO partners to bring simplification to their modeling — and then build out on top of that better scenario planning to actually reduce the timelines for new generation interconnection.