The Scoop
The breakout success of Chinese AI platform DeepSeek doesn’t mean a dimmer future for the US electric grid, despite what Wall Street seems to think, the billionaire CEO of a company racing to produce low-carbon power from natural gas told Semafor.
NET Power was among the energy companies whose share price plunged this week alongside chipmaker Nvidia’s, when DeepSeek claimed to be able to function as a far cheaper and less energy-intensive alternative to ChatGPT. But CEO Danny Rice — who made a fortune drilling for natural gas in Pennsylvania and whose brother Toby is CEO of the country’s top gas producer, EQT — is still bullish about the opportunity AI presents for energy companies.
Rice sees the pivot by oil majors ExxonMobil and, as of this week, Chevron into the data center power game not as competition, but as a lucrative new batch of customers for NET’s technology capturing CO2 from gas-fired power plants. But he thinks the Trump administration should reconsider how tax credits for CO2 capture are doled out.
In this article:
Tim’s view
Rice is talking his own book, of course. But there’s good reason to think that even if DeepSeek’s energy footprint becomes the norm, there will still be plenty of business for companies banking on a bigger grid.
To be sure, US energy forecasters — and energy company shareholders — had started to take it for granted that AI will drive major growth in electricity demand. Firms like GE Vernova and Constellation Energy lined up to sign deals with Big Tech and, as a result, have seen their share prices pegged more closely to its rise or fall. If the thesis about power demand growth is wrong, then these companies are overvalued, and the power bubble could be ready to burst. The DeepSeek episode implies that the path to AI dominance doesn’t necessarily require replicating the existing energy system at greater scale. Instead, especially given the various financial and bureaucratic obstacles involved in building new power infrastructure, Big Tech needs to focus on finding ways to squeeze more AI power out of fewer electrons. As University of Texas energy researcher Arvind Ravikumar wrote on Bluesky: “This is a good time to take stock of unrealized efficiency gains in AI compute before we end up building a bunch of 30-year gas plants.”
But data centers are just one area of growing demand among many, including the electrification of vehicles and buildings and the expansion of US manufacturing. And until it’s possible to get a clearer look under DeepSeek’s hood, investors should be cautious about granting it a premature victory, said Dan Goldman, managing partner at the climate tech investment firm Clean Energy Ventures: “I think that sell-off may have been a bit overstated, because I don’t think anyone really knows just what’s going on and how impactful it could be.”
The View From Danny Rice
This interview has been lightly edited for length and clarity.
Tim McDonnell: What did you make of the market’s reaction to DeepSeek this week? Have we been wildly overestimating the amount of energy AI is really going to need?
Danny Rice: It was just a knee-jerk reaction from the market. People think if these AI models need less power than originally predicted, that means less demand for utilities and natural gas. But people missed two things. One, they don’t appreciate how power-constrained we are to even hit moderate-to-low growth in AI data center expansion. Things are still going to be pretty tight even if DeepSeek-level efficiency takes over the market, which would certainly be a good thing for end users and for broader adoption of AI.
Two, if AI becomes a lot more affordable, does that mean less expansion of it, or more? It’s hard to find examples in history where lowering the cost of something lowered the consumption of it. We’re already in a very power-constrained world, so if we democratize and expand the potential of AI, it would really increase the absolute [electricity] load that way. So in any scenario, the reality is grid systems across the country have seen a massive under-investment in firm power generation for the last 15 years. We’re going to be in the mode of playing catch-up for the next decade, whether load growth is 1 or 2 or 3%, driven by AI ambitions. Even the lower scenario is growth the US grid hasn’t seen for 15 or 20 years.
And where do you see gas fitting into that picture?
It’s been apparent to us for a long time that natural gas is going to be the presumptive energy source for any meaningful load growth. One thing is just the abundance of natural gas. If we need to add 100 or 200 gigawatts of load to the grid, the only energy source that we can actually scale at that magnitude is natural gas.
There are two alternatives [for data centers], renewables with battery storage or nuclear. Nuclear is just inherently more expensive. And time is not on their side. Everybody’s cautiously optimistic that we’ll see new nuclear by the early 2030s, but my guess is you’re probably not going to see real new nuclear at the scale and price you want to see until the late 2030s, which means they’re really going to miss out on this wave. Then renewables with battery storage, they haven’t proven they can get to the capacity factor you need to where you can bank on sticking it next to a data center. And it’s cost prohibitive.
But gas can scale extremely fast. And if you’re in a place that has access to low-cost gas, like the US, it’s a no-brainer. Certainly people are asking, can we do it even more responsibly, and cleaner? That’s the role that natural gas with CCS can play. We can decarbonize it and still end up with a lower cost of energy than we would from any of the alternatives. That’s the whole thesis.
That’s what the oil majors seem to think — first Exxon and now Chevron are also rolling out gas with CCS power systems for data centers. Can you compete with them?
Well, they’ll become our customers. We’re inventing this technology, we’re standing up the global supply chain for all of the key components. We’re lining up the [construction] firms. So utilities or oil and gas companies will come to us and say, ‘I want to build X number of NET Power plants. Sell me a license, give me your kit, give me the blueprint, and just let me go do it.’ That’s our business model, as a licensor of this technology.
Then we have another revenue stream from the CO2. For each of our power blocks, we’re capturing almost 900,000 tons of CO2 per year. And in the US, with the 45Q tax credit, the government will pay you $85 per ton of CO2 you sequester. And if you think about what industries possess the skill set to understand the subsurface, it’s oil and gas. The value proposition to them is, they’re selling these molecules in gas form at $3-4 per MMBtu. If they vertically integrate and own the power generation facility, you can double the value of your gas because you’re now selling it as an electron instead of a molecule, and you make more money on the back end through the CO2.
And you don’t have any concerns about whether that tax credit will survive under Trump?
The Trump administration’s big thing is domestic energy security, and increasing domestic oil and natural gas production. We create more demand for domestic natural gas, which incentivizes new production. And actually at our first plant in the Permian, the CO2 is going to be used for enhanced oil recovery. So we’re also increasing domestic oil production, while improving the reliability of the grid. And 45Q helps ensure the affordability of that power until we can scale this technology up and get into manufacturing mode and get our capex down. But we certainly don’t see ourselves needing it forever.
There will be other incentives on the chopping block. But for the things that increase domestic production and energy security, I think those are going to be things they’ll want to support.
Right now 45Q pays out more when the CO2 is sequestered than when it’s used for oil recovery. Would you prefer to see that equalized?
I think there’s some hope for an increase in the utilization credit to be at parity with permanent geologic sequestration. If that happens, it will end up promoting more oil production that wouldn’t otherwise happen.
Higher oil and gas production may be good for lowering consumer energy costs, but to the extent that doing so lowers prices, it’s not what most drilling companies and their shareholders want. How do you view the tradeoff between prices and production?
That’s the big question, how do you expand production without spiting yourself? That’s why folks like [my brother] Toby are really pushing for permitting reform to allow us to build more natural gas infrastructure, to get to more markets that have an insatiable demand. When we think about ‘drill, baby drill,’ it’s more like ‘build, baby, build.’ The demand is already there. The capital is already there. All we need is permitting reform to get us back to a place where we can let market forces dictate where things go. That’s what the industry needs to say, ‘alright, let’s start to drill again.’