Elon Musk says the AI race is accelerating beyond anyone’s control. But every competitor faces the same physical limit: electricity. PJM’s widening power shortfall shows why companies with secured grid capacity could become the next major AI trade.
The post The AI power gap Wall Street still hasn’t priced in appeared first on Curzio Research.
Elon Musk likes to move as quickly as possible. The Tesla (TSLA) and SpaceX (SPCX) CEO has been known to occasionally bypass reality when it comes to his projects and businesses. Recently, Musk has become a major proponent of using natural gas turbines for the artificial-intelligence (“AI”) build-out.
All of that came into focus on August 29, when Musk announced that SpaceX will manufacture its own gas-turbine parts – specifically blades and vanes – to help speed up the process of getting natural gas turbines online to provide power for its data centers.
SpaceX and Tesla are each building 100GW/year of solar production capacity as fast as possible, but natural gas will still be needed to supplement and bootstrap solar for several years.
— Elon Musk (@elonmusk) August 29, 2026
The limiting factor for nat gas turbine production is casting the blades & vanes. By doing…
As you can see in Musk’s X post, he believes casting gas-turbine blades and vanes in-house will cut production time by up to 18 months and be a “profound game-changer” for SpaceX.
And he’s probably right. If SpaceX can actually expedite the process by up to 18 months, it will be a huge win for Musk and company. As he stated in his post, natural gas is still a critical energy source needed to supplement solar power.
But some questions remain unanswered:
If you’re unfamiliar with the concept of “Dark Energy,” let me take you back to a June 5 article by my colleague Steven Longenecker.
In that piece, he wrote:
Elon Musk will be one of the world’s top buyers of gas turbines over the next few years…
SpaceX has committed to spending more than $2.8 billion on these turbines for its artificial intelligence (“AI”) build-out, the company recently revealed in its initial public offering (“IPO”) filing…
In March, SpaceX agreed to buy $805 million worth of turbines from an unnamed supplier, with deliveries running through 2029. Then in late April, the company struck a separate, still-pending deal for roughly $2 billion worth of mobile gas turbines and related equipment from another vendor.
Steven’s article continued:
And it confirms something my colleague Professor Joel Litman has been saying for months…
This kind of off-grid “Dark Energy” is the next stage of the AI boom. As I explained in our Dark Energy deep-dive last month:
Dark Energy is a clean, reliable power source. Specifically, it’s a class of natural-gas turbines… essentially a jet engine adapted into an electric generator. They start in minutes, run on natural gas instead of jet fuel, and can produce tens of megawatts of power per unit. Stack enough of them together, and you have a private power plant…
But until recently, no one was using them at scale to power the world’s most advanced AI labs. Now that they are, the suppliers building them are sold out for years.
Since then, SpaceX has also purchased APR Energy, a Jacksonville-based power company specializing in rapidly deployable, modular power generation.
I covered that transaction in a July 21 article, in which I wrote:
The deal, which is thought to be worth at least $1 billion, gives Musk what APR Energy’s website describes as “one of the world’s largest mobile gas turbine fleets,” with solutions that scale from 20 megawatts (“MW”) to 500 MW. APR Energy has delivered more than 50 terawatt-hours of power to more than 35 countries.
Most importantly, the company claims it can “deliver reliable energy in 30 to 90 days.”
And that’s exactly what Musk – and every other data-center operator and hyperscaler in the world – desperately wants and needs. Plenty of power… and fast.
Musk is clearly a big believer in the fast power that natural gas turbines can provide to SpaceX’s data centers. But, based on his recent X post, that power can’t be delivered fast enough. So, the next logical step after buying up natural gas turbines and then buying an energy company for its fleet of mobile gas turbines is to help build them yourself, right?
Regardless of how he procures the turbines, Musk is in the midst of taking Dark Energy to the next level.
Bastrop, Texas, a small town of roughly 14,000 residents located around 30 miles outside of Austin, was once known as a filming location for the iconic 1974 horror classic, The Texas Chain Saw Massacre. Today, it’s Elon Musk’s playground.
Along with Musk’s huge SpaceX Starlink factory, Bastrop will soon be home to the new foundry that will cast blades and vanes for natural gas turbines. Musk, through his recent dealings with gas-turbine production, learned that the primary bottleneck is the casting process.
TechCrunch, via the Information, explained why:
The blades inside a gas turbine’s hottest section run at temperatures around 3,000 to 3,600 degrees Fahrenheit, which is roughly 800 degrees hotter than the melting point of the very metal alloy they’re made from. That’s only possible because of the blades’ internal cooling channels and thermal-barrier coatings, plus the specific way each blade is cast. Just four companies worldwide have mastered the casting process well enough to produce them at industrial scale, and all of them are tapped out right now.
To repeat, only four companies on the planet know how to do this on a large scale. But they’re completely booked through 2029 and into 2030. And, as we’ve learned, patience is not a virtue of Musk’s. So, he’s opting to do it in-house rather than wait.
The problem is that the entire operation is obviously very complex and delicate.
As TechCrunch noted:
What makes the whole thing especially difficult is that each blade has to be cast as a single, unbroken crystal, grown slowly inside a vacuum furnace, without the microscopic seams that let ordinary cast metal crack under stress. It’s a tricky process even for the smaller blades used in jet engines; the blades in power-plant turbines are considerably larger, which makes producing them at that scale and without defects even harder.
Which takes us back to some of those unanswered questions I posed at the start of this article…
The answer to all three questions is, essentially, a giant shrug.
We do know that SpaceX bought roughly 830 acres of land in Bastrop between March and June. But that’s pretty much all we know. As of August, the state of Texas had no record of any environmental or building permits being filed by SpaceX for construction of the facility.
Let’s also not forget that these types of gas turbines release smog-forming compounds and dangerous chemicals like formaldehyde, as well as pollutants linked to asthma, respiratory disease, and certain cancers into the air. That’s why significant industrial air-quality permits are required before construction can even begin.
Those permits haven’t appeared yet.
The next part to consider, assuming permits are granted and foundry construction goes smoothly, is the time it takes to create the product.
Peter Perri III, an energy investor with more than 20 years of experience in building and financing power infrastructure, noted in an August 30 Substack post:
A vacuum casting foundry capable of single-crystal work is 18 to 24 months from groundbreaking to commissioning, and the job listings say they’re still hiring the people who will build it. Process development on single-crystal yield typically runs another two to three years.
At this point, no ground has been broken, and no team has been assembled. So, realistically, we’re probably looking at a minimum of five years before the foundry yields any product. History tells us that Musk won’t want to wait that long.
Global research company Gartner projects that the amount of energy consumed by data centers around the world will reach 565 terawatt-hours (“TWh”) by the end of this year. To put that into perspective, here are some fun facts about how much 565 TWh is worth in real-world contexts:
By 2030, that number will hit nearly 1,000 TWh, according to the International Energy Agency.
The current electrical-grid infrastructure simply can’t handle this type of load. Something needs to power the rapidly growing number of data centers, which shows no signs of tapering off anytime soon. This has hyperscalers like Amazon (AMZN), Meta Platforms (META), Alphabet (GOOGL), Microsoft (MSFT), OpenAI, and SpaceX looking for quick energy sources.
Most of those companies have already gone the renewable, clean-energy route, like solar and wind. But they’re all turning – or have already turned – to natural gas because it has typically been a faster way to get their data centers online.
But the AI-driven demand has overwhelmed the supply. GE Vernova (GEV), the world’s largest manufacturer of natural gas turbines, is effectively sold out of production capacity through 2030.
That’s the reason Musk turned to APR Energy and its fleet of mobile gas turbines. And that’s why he decided to build his own parts to break the bottleneck holding up gas-turbine production.
But can SpaceX pull this off? As I mentioned, there is no foundry yet… permits haven’t been secured, and ground hasn’t been broken. And no team of experts is in place to oversee and take part in the extremely complicated and sensitive process of casting turbine blades.
Could it theoretically happen? Sure. Musk has the financial means to make virtually anything happen. He should be able to lure top talent to Texas to participate in the project. Although, considering the recent news that SpaceX completely restructured its data-center team because of engineering and reliability concerns at its Mississippi and Tennessee factories, is there any guarantee that SpaceX will even be able to bring in the right people?
If SpaceX does pull this off, it will position the company as one of only five Western companies capable of this type of manufacturing. SpaceX would have every AI-infrastructure builder across the globe lining up for its turbine blades and vanes.
But the lack of any progress and the daunting production timelines for both foundry construction and large-scale production make this feel closer to fantasy than reality.
And, let’s face it, Musk has a long history of grandiose ideas and projects that sound amazing at conception but barely get off the ground. The fully self-driving Tesla car (on a large scale), the Hyperloop, the mass rollout of Tesla robotaxis, personal rocket-travel across Earth, the Tesla Roadster, the city on Mars… you get the idea.
The man is undeniably brilliant and has conceptualized some incredible technological breakthroughs. But he also owns a long list of broken promises. Time will tell which category the turbine-blade foundry in Texas falls into.
Regards,
David Engle
Editor’s Note: A new “dark energy” is being rolled out as we speak… It completely bypasses our need for foreign oil. And it’s not nuclear or solar or wind or anything you would expect.
This new potential $10 trillion technology is already seeing investment by early backers of Microsoft, Google, Amazon, and more… Most important, these dozens of billionaires can’t make this tech themselves. They have to go through obscure “supplier” companies.
And in the coming months, this new “dark energy” could send three little-known stocks soaring… while wiping out 10 of the most popular stocks on the market. Learn more about this dark energy opportunity by clicking right here.
The post SpaceX Wants to Build Natural Gas-Turbine Parts for AI Data Centers. Can Elon Musk Pull It Off? appeared first on MarketWise.
It seems like just about everyone is mad about AI data centers right now.
They’re thought of as big, loud nuisances… and it’s hard to find people who actually want them where they live.
At the same time, we’re using artificial intelligence (“AI”) more than ever for searches, coding, writing, problem-solving, and even just chatting. So, demand for data centers is real.
Nonetheless, the AI data-center backlash we’re seeing has become big enough to matter to investors. Towns are pushing back over water, power bills, tax breaks, noise, diesel generators, and the simple reality that nobody wants a windowless server farm dropped into their neighborhood.
That’s a problem for the AI trade because Nvidia (NVDA) stock may be a proxy for AI growth, but chips still need to be installed in hardware which needs to be deployed into buildings, powered by transformers, and receive municipal permits before they become revenue. Nvidia is the top bellwether for AI, but it still relies on data centers for a majority of its chip sales.
It’s also creating a strange side effect in bitcoin mining…
See, while investors have been staring at Nvidia, the bitcoin mining network has been doing something odd.
Crypto-tracking site CoinWarz put bitcoin difficulty at 125.81 trillion today. About a year ago, on August 24, 2025, according to Hashrate Index, the difficulty was at 129.70 trillion… then an all-time high.
In plain English, the work required to mine a block today is roughly the same as it was last summer. That gives miners more time to make a return on their hardware investments without having to upgrade.
So thank you, AI data centers. And thank you to everyone saying they don’t want one in their neighborhood. It’s helping make certain miners a lot of money.
Today, we’ll look further at how the AI-infrastructure build-out and rising bitcoin prices are setting up a profitable scenario in the mining space.
Bitcoin mining difficulty is the network’s thermostat.
Every two weeks, bitcoin adjusts how hard miners must work to keep blocks arriving about every 10 minutes. When more machines mine for the same reward, bitcoin difficulty rises. Fewer machines, or machines pointed somewhere else, means difficulty stabilizes or falls.
For a miner, that math is brutal and simple. Revenue comes from the bitcoin price, block rewards, and transaction fees. Costs come from power, machines, debt, and maintenance. If bitcoin jumps while difficulty stays flat, the same machines earn more dollars without facing the usual penalty of a harder network.
That’s exactly why the recent bitcoin rally matters. Bitcoin gained 23% in a week– its biggest move since last year’s sell-off – after Treasury-bond buybacks pushed long yields lower, the dollar weakened, and a short squeeze forced buyers back into the market. ETF inflows are the next test, because forced buying burns out quickly, while institutional flows can keep a rally alive.
If bitcoin holds even part of that move higher, miners get operating leverage when the selling price rises, and the production hurdle does not.
In previous bitcoin cycles, miner capitulation looked familiar. The price fell, inefficient miners shut off, mining rigs hit the secondary market, and the network waited for difficulty to adjust down.
This cycle has a different escape hatch. Miners don’t have to disappear. Some have become AI-infrastructure companies.
That sounds like a branding trick until you remember what miners can offer to the AI space: power access, land, cooling, substations, operating teams, and a tolerance for absurd capital intensity. Those are the main things that data-center developers are fighting to secure.
CoinDesk reported in March that public miners had announced more than $70 billion in AI and high-performance-computing (“HPC”) contracts. S&P Global listed Iren, Riot Platforms, Core Scientific, Hive Digital Technologies, Cipher Digital, and TeraWulf among the miners shifting capacity away from pure bitcoin mining, with analysts expecting HPC and AI to drive much of their growth in 2026.
That matters because every megawatt chasing AI is a megawatt that’s not automatically chasing the next bitcoin block. While the machines and customers are different, the scarce input is the same: energized infrastructure.
The result is a softer version of miner capitulation. Instead of “we shut off because bitcoin mining stopped being profitable,” the message becomes “we found a better-paying customer for this power.”
Now for the catch…
The same local revolt slowing the data-center land rush and helping bitcoin miners also threatens the miners trying to become AI landlords.
Governors and lawmakers are moving to force data centers to pay for their own electricity supply, limit water use, disclose more about operations, and win local support, according to the Associated Press.
Microsoft (MSFT) is already working to unwind multimillion-dollar tax breaks on several Atlanta data-center projects as the politics get hotter.
So, the trade is not “buy every miner with AI in the press release.” Please don’t do that. That’s how people end up owning a construction project with a bitcoin ticker.
Look for mining stocks that already control power, have credible tenants, and can survive construction delays while still making money mining bitcoin if their AI customer takes longer than promised to come online.
Global asset manager VanEck has already warned that miners chasing AI face a roughly $50 billion near-term funding gap and that only about a quarter of leased AI and HPC capacity has been delivered so far.
That means there’s still a long way to go.
Chips are no longer the bottleneck in the AI trade. Power is. Permits are. Local sentiment is. The boring parts are now in charge.
For bitcoin miners, that means three numbers are worth watching…
That’s one thing many bitcoin investors are missing. While the AI data-center backlash is a risk to the AI build-out, it’s also putting a ceiling on one of bitcoin mining’s worst enemies: endless hash-rate growth.
If bitcoin’s price keeps firming while mining difficulty stays flat, it presents a solid opportunity for bitcoin mining stocks to profit from.
For once, the local zoning board is indirectly doing bitcoin miners a favor.
Good investing,
Eric Wade
Editor’s Note: Trillions of dollars — the largest flood of spending in U.S. history — are pouring into AI and energy. And one little-known company sits right in the path of it, collecting a toll on every dollar that flows past. It doesn’t make a single chip or pump a drop of oil. Whitney Tilson calls it “America’s Greatest Retirement Stock” right now — and today, it’s trading at a rare discount. He’s giving away the name for free. Click here…
The post Bitcoin Miners: A Surprising Winner From the Growing AI Data-Center Backlash appeared first on MarketWise.
Bitcoin miners built their businesses around cheap power and mining rigs. Now, many are realizing their power assets may be worth far more as AI data center infrastructure. Here's why the shift matters—and which names are profiting.
The post Bitcoin miners just found a better business model appeared first on Curzio Research.
Login