Data Center Blowback

How are data centers changing power industry business models? Growing local opposition is prompting states as varied as South Dakota, Oklahoma, Georgia, New York and Vermont to consider halting new data center projects for one to three years. Utility regulators are under pressure to "do something" about rising electricity prices. What are the potential effects on independent generators?

A panel talked about these questions at our 35th energy finance conference outside San Francisco in mid-June. The panelists are Vikas Anand, CEO of Origis Energy, Blake Nixon, CEO of Geronimo Power, The Honorable John Reynolds, President of the California Public Utilities Commission, and Jan Smutny-Jones, CEO of the California Independent Energy Producers Association. The moderator is Sameer Ghaznavi with Norton Rose Fulbright in Chicago and host of the data center podcast Uptime Now.

Generator Effects

MR. GHAZNAVI: Vikas Anand, how are data centers changing power industry business models?

MR. ANAND: Data centers have been a big procurement engine for power for some time, but the AI boom has evolved in the last 12 to 18 months to require two things.

One is a need for data centers that support training and inferencing, which by definition means larger data centers and stricter requirements around latency and reliability. There are nuances, but that is the big change.

The second thing is the no-regrets approach that most of the hyperscalers are taking to these investments. It is their strategic imperative. They spare no effort. They do not want to look back five years from now and feel they underinvested. Add to that the nationalistic view on AI adopted by the current administration that has made for a really different dynamic.

Therefore, what we are seeing is a demand for speed to power, and then lots of power. The demand signals are encouraging a lot of new innovation. We are all being more ambitious today than we were probably two years ago.

The power industry is changing: more innovation, more incubation, more bets. The NextEra-Dominion merger is an example. At Origins Energy, we are being more ambitious about our goals. I won't say we are making bigger bets on a per-project basis, but we definitely are open to taking more bets than two years ago. This is despite the uncertainty on tax credits, tariffs, Iran and everything else.

MR. NIXON: The demand has moved from virtual PPAs to a more physical world where you actually need the power versus the green marketing benefits that come with it. There has been a big shift back toward old school procurement through utilities to be on the grid to the extent you can and the physical buildout that comes with it. Demand for power has increased, but it is mostly a shift toward capacity. Capacity seems to be the thing that is in shortest supply right now.

MR. GHAZNAVI: Retail electricity prices are increasing in 39 US states. Have you seen a similar change in the wholesale prices at which you sell electricity?

MR. ANAND: Absolutely. Power contracts are pricing very differently than four or five years ago. Some of it is falling to the margin line, but not all of it. Costs have gone up substantially.

A lot happens between the prices to which we agree in PPAs and what retail customers pay. This is a topic on which we as an industry need to engage quickly or we risk losing the narrative to the anti-renewables groups that blame renewables for higher prices.

MR. NIXON: Renewable energy prices have doubled since 2020 across all markets. They continue to go up. If you can do something to enable a data center, the increase is much greater than that. Capacity has moved from being basically worthless to being worth a heck of a lot. Capacity that we used to contract at $5 to $7 a KWh month has tripled or quadrupled in most markets. That is where you see the greatest effect on residential power prices.

Regulatory Concerns

MR. GHAZNAVI: The increase in residential power prices is sparking opposition to data centers.

John Reynolds, when regulators say data centers should pay their own way, what does that mean in practice? And how is California dealing with the constraints that data centers are putting on power, water and other resources?

MR. REYNOLDS: Fortunately, we are only responsible for the power side, but I will speak broadly to both parts of the question.

When regulators say they want data centers to pay their own way, we are thinking about both the procurement that is necessary to serve all customers and the transmission and distribution infrastructure necessary to deliver electricity to them.

Speaking first to procurement, California is seeing load growth on multiple fronts. Data centers are a component of that. We are not the largest data center market in the country by any stretch, but they are a component of our load growth. We are also seeing load growth from vehicle electrification, and we expect further load growth from electrification of both transportation and buildings. More extreme weather patterns are adding to the challenge.

In 2020 and 2022, we had extreme heat events that pushed our grid to the brink. We have been procuring a lot of resources to help with reliability since 2020. We brought online more than 27,000 megawatts of new clean resources from 2020 through March this year. Our utilities have another 22,000 megawatts under contract to come online before 2030.

So while we are building resources at a pretty tremendous pace, we have many needs for those resources. The California Public Utilities Commission, in partnership with our sister agency, the California Energy Commission, and our grid operator, the CAISO, have been able to support the ongoing procurement by utilities to serve data centers and other load growth.

Turning to transmission, what we as regulators really worry about is major infrastructure upgrades that involve 30-, 40-, 50- and 60-year assets where the expectation is that customer demand will cover the cost and put some downward pressure on customer rates, but we worry about the potential for large customers at single sites where we would not otherwise need major electrical infrastructure reducing their demand over time, whether that is because of changes in their business models or the development of behind-the-meter resources that lead them to use less power and pay less for the system.

We have several ongoing regulatory proceedings in which we are addressing cost allocation and interconnection issues for large loads, including data centers. We have an Interim Rule 30, where large-load customers like data centers in the PG&E service territory can pay up front for the infrastructure needed to connect to the grid. They do not have a guarantee that those funds will be returned to them.

Existing customers are protected. We will be doing studies to evaluate how much of that infrastructure will end up being used solely by the data center customer versus used for the system as a whole and, therefore, should have more cost sharing with other customers. We have a couple of other dockets that are also dealing with large-load issues, including both our high-DER docket and a new advanced-rate-design docket, both of which will touch on rate design, protecting existing customers and facilitating greater speed to power for these customers whom we ultimately want on the system, drawing power from the grid and staying on the grid for a long time.

IPP Concerns

MR. GHAZNAVI: Jan Smutny-Jones, what are independent power producers most concerned about when it comes to data centers?

MR. SMUTNY-JONES: We view them as potentially new customers. We are always looking for people who will buy our power. That has not been much of a problem. As President Reynolds indicated, no one is racing to California because we have cheap land, projects are easy to permit and there is an abundant water supply.

California already has a significant number of data centers in the Bay Area and San Diego and a few other places. The California Assembly released a really good paper earlier this year on data center impacts on ratepayers. There are basically four types of data centers.

The type that is most relevant in California is the data centers built around latency. I had to go on AI to figure out what that means. The communities south of San Francisco -- Santa Clara and San Jose, for example -- have aggressive policies to encourage data centers.

LS Power is building 2,000 megawatts of new transmission capacity to bring additional power into that area. We are doing this with our eyes wide open.

We do a lot of planning in California. The California Energy Commission looks forward in terms of what we need, and then that turns into procurements at the California Public Utilities Commission. The data is used for transmission planning all the way out to 20 years. Last year, the California Energy Commission was looking at data centers and basically discounted what looked like a flood of future data centers. The commission decided to revisit the subject on a monthly basis in an effort to avoid an overbuild of transmission.

That is how California is dealing with the potential blowbacks. The CEC actions bind the CPUC. They do not bind the CAISO, so the CAISO is still behaving as if that load may show up. That may not be a problem because we will probably need the transmission anyway.

I was on a call recently with colleagues in Texas and at PJM. Texas believes it will have 400 GWh of data centers come online between now and 2032, which is something like five times the current load of ERCOT. But even if you discount that to 200 GWh, it is a big number. PJM is similar. PJM believes it will have sufficient capacity, and something like 55 GWh have cleared the interconnection queue. Neither seems panicked about being able to meet the coming data center load.

Bankability

MR. GHAZNAVI: Nearly 40% of data center projects this year are expected to be canceled or delayed due to permitting, lack of construction workers, strained grid capacity and shortages of electrical equipment. Blake Nixon, how do you plan for a bankable project with this kind of uncertainty?

MR. NIXON: It is hard. We have some recent experience that illustrated that. I have been doing this for 20 years. In the wind business 20 years ago, you had lots of misinformation that took three to six months to sort out to a point where you could have a fact-based debate.

This has parallels. The pushback on data centers is a phenomenon, and whether or not it is being fanned by China, it has grabbed hold of the popular imagination and makes the job really hard.

Working directly with the data center helps. You get to know the answers to a string of important questions, such as water use, power consumption, how the power supply needs to be structured, and whether it will have to be behind or front of the meter.

Developers then have to talk about how their projects will help the local community. It is a different conversation than we are accustomed to having in the discrete local project business.

The hyperscalers have no shortage of guys like me throwing things at them. There are lots of fake projects. There are lots of real projects. They have small teams trying to work through big piles of proposals. They are putting 10 times the capital inside those data center shells that we are putting into the power outside those shells. We are part of their supply chains.

MR. ANAND: I don't think many people are surprised by the 40% cancellation number. I don't know if it is 40% or 50% or 30%. Projects get cancelled, but what is different this time is the size of the cancellation on a per-unit basis has gone up substantially. This contributes to greater volatility in our sector.

Do we need to get closer to the data center companies? Absolutely. Behaviorally, they are open to it. The other stakeholder we are trying to get closer to, and it is not an easy behavioral change, is the utilities because a decent amount of our business is in regulated markets.

As an anecdote, we have a fantastic project shortlisted by one of the utilities, but it is stuck in the shortlisting phase because the utility is trying to pull together the deal on the load side with a hyperscaler. We are trying to decide what to do with the project. Should we stay the course? Should we allocate more capital to it? Should we switch gears and do a VPPA? That is an example of the challenges facing developers

MR. GHAZNAVI: You mentioned utilities. Are there markets that have become more attractive in the last 12 months, either because of utilities or because there are areas where it is easier to get community support for projects?

MR. ANAND: The window of 2030 is forcing disciplined capital allocation. All else being equal, you would prefer a project that has high odds of success in the next four or five years. That leads you to a certain set of markets, ERCOT being one.

MR. NIXON: There are not many geographic areas that look favorably on new project development. There are a few, but they are few and far between. We used to be able to rely on Texas as one of those areas. Texas sometimes still works, but not always.

These transactions are not meant to move at lightning speed. We are not meant to build brand new transmission lines and spend trillions of dollars on transmission just because somebody says they want it. We are meant to be thoughtful and intentional and slow in making these decisions so that we do not end up with stranded assets and escalating power prices. Utilities are trying to be more nimble, but they are just not built that way. They are not staffed that way. They have 30-year veterans who are smart and care and want to do good jobs, but mostly they do not want to make mistakes.

The intersection between the fastest moving industry in the world, data centers, and our industry, which generally moves slowly, is tough to sort out. I think greater focus is needed on load management and demand response. Storage has a role to play.

Market Dynamics

MR. ANAND: One of the questions this panel was asked to answer is how this is changing the power industry.

There has been a notable shift and one that is happening with a lot of urgency. The hyperscalers are taking an approach that everything is on the table. They are getting very aggressive around bringing your own transmission, peak shaving, and demand response practices and technologies.

Utilities are receptive. That has been a change in the last 12 to 18 months. It has an impact on how we operate as an industry. One approach is to try to make better use of the supply we already have. That does not fully solve the transmission issues, but it may be possible to find 7,500 MW of capacity by taking actions that might not have been taken in the past.

MR. GHAZNAVI: We have seen promising looking load forecasts for energy producers, but is this placing too much power in the hands of a relatively small number of hyperscaler buyers?

MR. ANAND: I don't know that is it too much power, but industrial customers with clean energy goals are not going to be able to afford what others in the market are able to pay.

It is also a big issue for utilities, which is how to manage their customers when a hyperscaler in the territory is willing to pay almost 30% more than its industrial customers can afford. I do not have any great solutions. It is an emerging problem.

MR. NIXON: There is a lot riding on a handful of really large, important companies. The regulators have allocated costs on a but-for basis in other contexts. They know how to do this. It is not like we are dealing with large loads for the first time ever. These tariffs exist.

Mining is big in northern Minnesota. Large load tariffs have been on the books for 50 years. We do have tools. It is the speed at which we are having to address this that is the change. If done well, it can lower prices. When I talk to our utility partners, every single one says publicly that this will help to moderate the growth rate or reduce rates for most if not all of their customers by spreading costs over a much wider base and having the direct costs paid entirely by the hyperscalers.

MR. SMUTNY-JONES: The California legislature is in session now, which is always an exciting time. The message to data centers is go find your own power, develop whatever you need, and then we are okay. No one is saying don't come here. It is just you pay your own way.

California is having to put a lot of money into fire-related costs. We are basically having to rebuild significant parts of our infrastructure. And then you have the data centers on top of that. The state has a lot on its plate.

MR. REYNOLDS: I was going to make a similar point. I agree that in general we do have the tools to connect large loads and protect existing customers. We know how to do this. But there is a bit of an X factor in the pressure on supply chains. The competition for scarce equipment needed to serve large loads is a real challenge.

Local Anger

MR. GHAZNAVI: I want to talk about the backlash that has been building. It is becoming popular politically to oppose data centers. Bloomberg Law reported that about a dozen out of the 38 states offering some sort of tax incentive for data centers are now thinking about rolling them back. OpenAI recently claimed social media accounts that are tied to China are trying to stir up local opposition to data centers. With midterm elections around the corner, are you concerned that the backlash will affect independent generators?

MR. NIXON: Politicians -- even experienced political leaders that get stuff done -- have their heads down. It is a tough environment in which to try to do anything before the midterms. After the midterms, maybe we can start talking about things a little bit more practically again. That said, the concerns are real.

People are really concerned. In my opinion, they are concerned about the wrong things. The propaganda campaign is really strong. We can prove that a couple of chatbots in China have been stirring up opposition to our sites.

It is not only that. The locals stand up in meetings and are quite angry. They are verbally threatening members on boards, commissioners, members of my staff. That happened maybe once in 20 years. Now I get that once a week. This is different. I am hiring security forces to go with my developers. Seriously.

I keep telling my team that I believe we are doing the right thing. But it takes some real gumption to keep doing this. This backlash is not like anything we have seen in the past. It is at an emotional level and not just a fact-based-debate level like we have dealt with as an industry in the past. It is going to take different tools.

MR. SMUTNY-JONES: The AI industry has done a terrible job of promoting itself. I agree that this is a national security issue. I don't necessarily want to learn Mandarin. We have a good edge on AI models. There are really big problems around them, potentially. There are public fights about what AI may do, and some of the rhetoric is dystopian. Pretty much one of the selling points we are hearing from the AI guys is AI will reduce the amount of labor needed.

I watched a recent hearing in Ohio where a member of the public complained to the AI industry, "You told us to have our kids grow up to be computer engineers and programmers, learn code. And now you are telling me there are no jobs?" Somebody needs to hit the reset button to talk about this as an opportunity, not something that will destroy the world as we know it.

MR. REYNOLDS: As an infrastructure regulator, I will say that opposition to big projects is not new. That is something that we experience on a routine basis and have for a very long time.

I wonder if it might be useful to make a comparison to the continued investment in the telecom networks and industry adoption of 5G that involved the deployment of lots of physical infrastructure in our communities in order to deliver better connectivity.

While there definitely was and continues to be real opposition to developing that infrastructure, people also see the value in allowing their phones to receive and send data at higher speeds. There is a real parallel. If the AI industry can show the same kind of everyday value to people using their products, it can do a lot to alleviate some of the fears about developing infrastructure. I am not saying it will be easy.

It was not easy on the telecom side. It is not easy for any kind of large-scale projects. But how the debate is framed makes a difference.

Gas Inroads

MR. GHAZNAVI: One of the statistics that we saw on the screen this morning was 40% of data center projects have behind-the-meter solutions for power. Many of them involve installation of gas generators. Is this a threat to renewables?

MR. REYNOLDS: Historically across the country, we have seen gas and renewables as complements to each other. The intermittency of renewables combined with the firm capacity delivered by gas has produced real value on grids. Obviously, we as a state have climate goals to get to 100% retail sales of clean energy by 2045, and that is what we are building the grid towards. We continue to use gas today as part of our mix for energy, but we are planning for a future where we are relying less on gas for electricity generation than we are today.

MR. NIXON: I agree gas and renewables are complementary. We tried to provide a behind-the-meter solution for a data center with just renewables and storage, and it is not practical. Right now, gas is needed both behind the meter and front of the meter. Gas is needed to help moderate the intermittency of renewables, even with storage.

MR. SMUTNY-JONES: I am the last guy standing in California defending the gas fleet. This is an opportunity to put gas into perspective. The reality is that coal-fired power plants did not disappear in California due to renewables. It was the gas industry that basically displaced them. In 2008, 18.2% of the power in California was from coal, and basically that is now down to zero. The gas fleet at one point was producing about 44% of our electricity. It is now down around 30%, but it is needed on those hot September days. We have about 26,000 MW of gas on line to keep the lights on.

Whether we will be allowed to keep the fleet at that size remains to be seen, but it is a valuable reserve.

MR. GHAZNAVI: We are going to do a quick lightning round and we are going to involve the audience. I will ask a series of rapid-fire questions, and I will ask you to raise your hands.

Will data center growth slow in the next year? Audience, if you think yes, raise your hand. I see one, two, three, four, five, six. Not too many hands.

Will electricity prices increase or decrease next year? If you think prices will go up, raise your hand. Looks unanimous.

Will we see a change in the public sentiment about data centers, which today is not very positive? If you think the sentiment will turn positive, raise your hand. A few optimists in this group, but most have their hands down.

The last question is what are you most concerned about: permitting delays, interconnection delays, not enough construction workers or FEOC? We'll start with permitting delays. Raise your hand. A fair amount of the audience with hands up. Interconnection delays? Almost everybody. Not enough construction workers? Also a fair amount. FEOC? Fewer than I expected.