We are investors in the public equity markets. We are more generalists than specialists in the electrical sector, so we are trying to ramp up our knowledge in this specific area, focusing on power and supply-demand curves. You have probably noticed, and I am not sure how much you follow the public markets, but those power producer stocks have been rising significantly.

Yes, it impacts my 401k more than anything, so I watch it morning and night.

I would love to cover the power prices and how you expect them to evolve over time, then a few discussions related to data centers, power trends, and other related topics. Could you introduce yourself and talk about what you have historically done in the sector and what topics you are comfortable discussing?

I can talk about markets in as much depth as you want. I have been doing this for two decades, working on the merchant side, the hedge fund side, and now the utility side. I have a 360-degree view of the markets from different perspectives, including both load and generation, as well as the hedge fund side. I trade power and gas markets day in and day out, including futures and options, and I also actively trade equities. I am very familiar with equity valuation models and I am trading daily.

Are you familiar with both ERCOT and PJM, or just one of them?

PJM and gas markets, yes.

The companies we are looking at are Entergy, Vistra, and Constellation. I want to hear your thoughts on the prices for power today. I think my reference is PJM, and I believe they are around $50, maybe that is peak power prices. What do you think power prices are going to be next year, in two years, in three years, and why?

All things being equal, prices should be higher due to load growth. Power prices are about 70% correlated to gas, more so in PJM than in ERCOT. Gas is the marginal unit, so they are mostly correlated. However, ERCOT this summer was a good example. There was 5% to 6% load growth, but power prices were very low. Two factors contributed to this: one was the weather which was really mild, the other is the huge influx of batteries. Whether prices should be $50, $60 or $70 is hugely a function of gas prices and the weather. Cold weather can drive drastic outcomes because the market is becoming tighter so there is less slack in the system. The PJM capacity markets were a good indication of this, with increasing capacity.

In the capacity market, load growth is linear with no weather adjustments, which is why prices have hit the cap. If similar conditions persist and the weather doesn't act as a headwind, prices could easily increase by 25% to 30%, if not 100%.

Are you talking about capacity or energy?

Energy prices. However, there are bearish headwinds from solar, not as much in PJM but in ERCOT. The amount of solar and battery build-out is significant. Starting next year, you will have four hours of battery, which won't make it a one-way bet on prices.

We understand gas prices as the marginal producer and the impact of weather. One thing we are trying to understand more is that you don't have capacity during just one specific time of the day, like maybe from 4 PM to 6 PM.

That is referred to as the evening peaks.

During the rest of the day, do you actually have capacity?

Solar peaks from morning to evening, which is when we have a lot of excess capacity. Prices are very depressed and will continue to be because more solar is being added. The evening ramp-up, as people come home and turn on their TVs, is increasingly the expensive part of the day. That is where the batteries are trying to arbitrage. Earlier this week, we had the highest record net load day, and the price was $60. Without batteries, it could have been $600. If you have a continuous string of hot days, like three or four back-to-back, the batteries will save you for a day or two, but not on the third day. That was similar to five years ago during Hurricane Yuri, when prices went to $5,000 for a week.

To answer your question, in short, the afternoons are going to be depressed by solar. In the next one or two years, they will be eligible for subsidies, leading to a huge influx of solar, especially in ERCOT and some in the Midwest, not much in PJM, which is a good thing. In PJM, there will be pockets of congestion. The Dominion Data Center in Virginia will see prices decouple significantly from the rest of the PJM area due to congestion, disproportionately impacting generators with bigger exposure there.

Can you elaborate on why you think the Dominion Data Center area is going to have a change or decouple in prices?

The amount of load growth and data centers in that area is almost 75% of PJM's load growth within a hundred-mile radius. PJM has about 2,000 pricing locations, and the data center area, even north and south of Virginia, has a $20 spread.

We are trying to explore the data center angle to the prices. My understanding is that data centers run 24/7.

They do have a significant cooling shape. It is not like they are running flat 24/7. They have some shape influenced by cooling. Even non-AI servers and AI servers run about 75% of the time, and there is a cooling load to cool the servers, which has a temperature shape.

What is the shape for the AI data centers? Is it more during one time?

It depends. AI training runs 90% of the time, whereas AI inference doesn't. I haven't seen significant research on this, but there are one or two studies.

The sources and imprints need to be aligned with either business or consumer usage, which means they operate during the day and perhaps into the early evening. One topic we are trying to understand is whether, during the day, when the data center is running and we have excess capacity from solar, prices should not increase. The only time we lack excess capacity is in the early evenings, and that is when prices should rise. For the rest of the day, we shouldn't see much change in prices.

My question is, given that data centers are likely the largest part of load growth moving forward, the major change we will see is a demand curve that is the same, plus about 10% overall. But given the supply curve, we are not going to see price increases during off-peak hours, only during peak hours. Is that correct?

No, it is the opposite. Peak hours are when you have solar. Typically, off-peak is cheaper, which is what the market still reflects. Some hedge funds are going long on off-peak, betting there will be less solar but equal load growth during off-peak versus peak. However, there won't be enough supply during off-peak, except for batteries, which should push off-peak prices higher relative to on-peak. That is a bet some people are making. You are seeing load growth, but you are also building about 50 gigawatts of renewable energy annually, with 5 gigawatts of load growth each year. The peak U.S. load is around 750 to 770 gigawatts, while the average load is 450 gigawatts. Unlike other demand sources like oil or gas, which have more stable demand, power has significant supply and demand variation. So it is not a one-way bet.

It makes sense that off-peak hours should have higher price growth because that is when the data center load hits, and there is not enough power supply.

Yes, correct.

But during the rest of the day, do you have enough supply?

Yes, during the rest of the day, there is significant supply growth from solar.

There is also a discussion about whether operators can limit power to data centers to prevent high price increases, possibly making agreements or mandates that during off-peak hours, data centers can run inference at their highest capacity.

During high load days, like peak summer or peak winter, there are some proposals, although none have been finalized yet, to curtail power either voluntarily or involuntarily. The peak demand is 770 gigawatts, but the regular demand is only 450 gigawatts. You only need to ration power for about one week or 40 hours in a year when the load is high and there is scarcity. Residential customers, like you and me, are more price-sensitive. I cannot tell my wife when to run the dishwasher; she will do it when she wants.

Yes, we don't have that option.

There is no option or incentive. I am charged the same price regardless. But if you are a big customer, you can have variable pricing, like hourly pricing.

You mentioned prices should rise by around 20% over the next few years. How do you reconcile rising prices with excess capacity and these proposals? Do you think we won't reach any proposals that limit this growth during peak times, or do you think the opposite?

Are you asking if prices won't rise during peak times?

No, the question is how you reconcile that energy prices should continue to rise if there are ways to relocate the load.

By relocating the load, do you mean curtailing?

Yes, it is more about curtailing than relocating.

Relocating means putting data centers in places like West Texas, where they can be directly connected to a solar plant and a battery. They will mostly be off-grid but can connect to the grid when it is cheaper and available. These are called behind-the-meter data centers. This trend is likely to increase significantly. Elon Musk did something similar with the Colossus, buying used power plants and putting them behind the meter. Data centers are being pushed to become more flexible. Google has a system where they send computing tasks to parts of the U.S. with more solar or wind power at a given time. If the East Coast has more solar, they will do more computation there. There is definitely an incentive and a voluntary push for more flexibility.

If you are on PJM hours, maybe on the West Coast you are dealing with solar hours, so you can have an interest there without needing to do it.

Or even wind.

People always say prices are going to rise significantly due to capacity constraints, but I think there might be some flexibility around it.

Yes, with the amount of batteries being built and the behind-the-meter developments happening, it is not a one-way bet for sure. For data centers, instead of asking if prices will rise overall, it is better to ask which locations will see price increases. Some of it may rise due to gas prices. Gas is a one-way bet, with almost a 20% growth due to LNG plants, new power plants, and behind-the-meter data centers. However, there is not enough supply coming in.

Do you think gas prices are going to rise?

Yes, there is a symmetrical load growth relative to supply growth in gas, from various demand sources like LNG power plants and data centers. All of this will increase the price of gas.

Everyone will have to pay more because gas is the marginal cost.

Yes, a $1 increase in gas typically results in an $8 to $10 increase in power, so that is leveraged.

Your thought on gas is because the US is not producing enough gas to support new demands?

Yes, primarily due to LNG plants and data centers.

Why have most of the new large data centers opted to go off-grid? Is it because it is faster, avoiding interconnections, or because it offers more flexibility and faster approval?

It is definitely not about price. Grid power is cheaper for commercial users. They are going off-grid due to the speed of interconnection and other processes.

Do you think most gigawatt-based data centers will be off-grid?

There are projections suggesting 30% of the big ones will be off-grid. I think that is a believable number. Currently, it is only about 10%.

What is your thought process on data centers using nuclear as a power source? I understand the rationale for it, but do you think they can get the necessary approvals from regulators to use it?

It depends on how you approach it. If you do it as a physical behind-the-meter setup, like what Talent tried with Amazon, that can lead to trouble, even with a regulatory-friendly administration. However, if you structure it as a virtual PPA, a financial PPA where you just do a financial swap, that is an organic mechanism.

Then you are not doing anything directly; you just buy energy from them.

Exactly. You are essentially doing a financial swap, which is a good way to finance nuclear. Currently, some nuclear plants in ERCOT are getting paid around $40 per megawatt-hour. Some proposals suggest prices more than twice the current rate, which helps finance nuclear, keep it stable, and bring some plants back from retirement, like Microsoft did with the Constellation plant.

Yes, with the Three Mile Island.

Yes.

The financial PPA makes it easier because, as a center, you only need to fix the interconnection. Although it takes time, you don't need approval from PJM or ERCOT to take the plant off-grid. Taking the plant off-grid would be difficult as it could raise prices for everyone.

Exactly.

Do you have any opinions on which independent power producers are best positioned for nuclear plants? I think Constellation has the biggest fleet, but much of it is already priced on the spot. Do you have any ideas about other providers?

As you mentioned, Constellation has the most nuclear assets, and its valuation reflects that with higher multiples compared to others. Nuclear is currently popular with SMR and other developments, but many are still in experimental stages. Existing assets are more valuable than potential ones that may not materialize.

What do you think about on-grid gas assets? Talent, for example, has many on-grid assets, and Entergy also has a lot. Do you think they will be able to sell part of those assets to data centers?

Yes, I think Meta and Entergy, are involved in some deal. They are financing gas assets. Google just did one yesterday where they are doing a gas PPA with carbon sequestration. Gas is definitely the most flexible and responsive solution. It can scale quicker and can be operational in one or two years, unlike nuclear, which might not be ready until 2030, if at all.

My understanding is that these are mostly new sites, not utilizing existing power plants.

I think the Google PPA involves an existing power plant.

I haven't looked into them. I knew the Entergy project was new.

Yes, but I think many people are working on existing gas power plants as well, depending on the site location and other factors.

What locations are the best for this? The same PJM and ERCOT locations?

Yes, and even AEP Ohio is growing. I think 80% of the data center load growth will be in PJM and ERCOT. It is interesting because they still have the cheapest gas.

In ERCOT, yes.

In ERCOT, which actually has negative gas prices, and then in PJM.

A 50% discount on gas prices.

Yes.

We have heard from a few people that building a new gas power plant costs around $75 per megawatt hour. If you are going to build a new one, it is going to cost you to build and run it. Someone has to take that contract at $75 to make it financially viable. Do you agree with that number?

Yes, I would say it is north of $75.

If you think about those gas plants doing deals now, the minimum prices are at $75, is there any reason for this with older or new ones, or only new ones?

Are you asking about both older and new ones?

Can you explain the rationale for older and newer ones?

Older ones are already depreciated, so they will be slightly cheaper. Sometimes they don't have good efficiency, so you can get them for less. For example, a new gas plant's capacity might cost $2,200 or more per kilowatt, which is the capex cost to build a new plant. However, you can buy used plants for around $1,400 to $1,500 per kilowatt. That is what many IPPs are doing. The good thing is that it is accretive. You are trading at a 10 multiple to EBITDA, but you can buy these at a 7 multiple, and you have accretive cash flow.

I have seen a few of those. You think Talented is one of those, but they are supposedly newer plans?

Yes, those Talented plans are the newer ones. They are also in a better location.

Do you think if Google or Facebook would pursue clients, would they target the new ones or the older ones?

I think it is more location-specific. They care about location and speed. They are not going to pursue a 30-year-old asset. They are more time-conscious than price-conscious at the moment.

Because the older ones have more carbon?

Exactly, which is why Google is doing carbon sequestration.

Otherwise, you would have to do a lot of carbon compensation to offset it. Speaking of Talented, do you know their assets? Have you ever looked at them in detail?

Yes, mostly in PJM. They have the Susquehanna nuclear plant and two or three gas plants.

Yes, they have three, maybe four or five gas plants. Do you have any opinion on which one could be involved in a data center contract over the next five years?

I don't remember exactly. There is one near an Amazon data center, so that is a likely target. If you plot big data centers on their plants, the answer would be obvious.

Do you plot them on the map?

Yes.

If it is closer to the Data Center Valley, the e-commerce hub, it is probably that one?

Yes. Enviros is selling a solution where you can visualize that, making it very obvious.

Would you do the same thing for Texas? Look at providers in Texas close to where everyone is building data centers?

Yes, but in Texas, most people are building in West Texas. There, you look at gas pipelines and try to place them near those.

Gas pipelines, because you want to connect to them for a new plant. Do you think it makes sense for existing gas plants to be rented out in West Texas?

There is not much available to rent out now. It has to be new ones.

Why is the data center valley in Virginia such a prime spot for data centers?

A lot of it is inertia. Why are most IT companies in Silicon Valley? About 70% of all Internet connectivity in the U.S. is located in Virginia.

Is that where the fiber optics run through, which is relatively important?

And it also connects to the Atlantic.

I want to talk to you a little bit about capacity prices. We talked a bit about PJM, which is the one that is exploding. What is your view on how capacity prices will move? The next one is already set with the cap, but what happens after the cap?

Fundamentally, there is only one way it can go, which is up. The only question is regulatory. With political pressure from the New Jersey and Virginia governors, the question is what PJM will do to change the cap or make adjustments. ERCOT did something similar by reducing the energy price cap from 9,000 to 5,000. I think the political risk is bigger than the market risk for capacity prices.

What prices do you think it would have cleared at without the cap?

At least 20% higher.

That is a lot.

Yes, definitely. Especially in some regions, it would have cleared much higher.

Do you think there is a chance for solutions in the capacity market? One solution I have heard is new capacity having no cap while existing capacity has a cap. Could that be a solution in the next few years?

Yes, there could be fragmentation. That is a decent proposal, but it is a very messy one. It is possible though.

Why do you say it is messy?

For example, with nuclear, without the capacity market, they say they will retire. So, would you create a separate one for nuclear? It is complicated.

How would you treat different resources?

Then there is the issue of resource discrimination. People could argue you are discriminating against older resources in favor of new ones, which could lead to lawsuits.

There are probably going to be a lot of losses around it. You have been following the market for stocks on the energy side. Is there any topic we haven't discussed that you think is relevant?

Yes, data center growth and PPs are big drivers. It is about who is going to get what PPs and who is bringing on new generation. Those are major drivers. Also, gas price exposure and its impact on revenue is something you are missing. If gas prices increase, how will revenue be affected?

I have a question about gas price exposure. My understanding is that in the PJM market, the price of energy is set by the marginal cost of the marginal producer. So, if the price is at $50 now, it is because the marginal cost to produce is $50, is that correct?

Yes, the spot price is set by the marginal cost. There is a spot price and a forward price.

So the price reaction we see in the financials is a mix of this spot and forward expectations?

Yes, spot and forward expectations.

How does that evolve from different types of gas plants? What is the evolution of marginal prices from a new gas plant to an older one?

New gas plants have a heat rate of 6, meaning $1 of gas would cause them to produce $6 of power. If gas is $4, then the power price is $24. Older, inefficient plants can have a heat rate of 10 or 11, so $4 of gas would result in $40 or $44 of power.

What is the highest it can go if you have to go to the older, less efficient ones?

Sometimes they even run on oil. Power prices can go up to $5,000 for certain hours when the system is tight. It is an artificial price set.

So it is not the marginal cost when it goes to $5,000?

No, it is set by a price cap when the system is tight, which happens for 40 to 100 hours a year.

When it gets tight, the regulator sets a price to reduce the incentive for everyone to buy energy. Is that the rationale?

Yes, people respond to demand.

So you actually get enough buffer? They are like, "Yes, we're running at a 5% capacity limit." They raise the prices so everyone has to pay more, then people will go off-grid.

Yes, especially the industrial kind of response.

And that happens during the 40 hours a year you mentioned?

Yes, during the big summer and winter peaks.

During the rest of the year, the prices are more dependent on the early evenings, which is the time of day when we have the highest prices?

Correct.

It is very complex and niche. Every area has its own plan, so you need to understand more about each area, like what the plants are, the load for that specific area, the market cost, and how the weather will be.

Yes, looking at oil companies and CNP producers is relatively easier.

It's not like I can simply go to my model and input prices. For example, I think ATC prices for PJM are around $44 over the year. I cannot simply say, "Oh, it's going to be $55 now," because it depends on the time of year and whether there will be curtailment from the load. Is that pretty much the rationale for this?

Yes.