Elon Musk says SpaceX has a faster route to one of the AI industry’s hardest infrastructure problems: getting enough electricity to run data centers. The answer, at least for the next several years, involves making a difficult gas turbine component in-house.
The plan could help address a manufacturing bottleneck that has slowed natural gas turbine production. It also brings a more uncomfortable question to the front: what happens when the race to power AI depends on building more gas-fired capacity near communities already worried about air pollution?
Why SpaceX is moving into turbine casting
On Saturday, Musk confirmed the purpose of a secret foundry SpaceX has been building in Bastrop, Texas. The confirmation followed a report from The Information, which cited job listings referring to a “blades and vanes foundry.”
The same report also pointed to findings from Corey Trinetti, a due diligence specialist who writes detailed reviews of AI infrastructure sites. Trinetti had reported that SpaceX bought roughly 830 acres near its existing Starlink factory in Bastrop between March and June.
Musk framed the foundry as a way to remove a specific constraint in natural gas turbine production. In a post on X, he wrote: “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. The limiting factor for nat gas turbine production is casting the blades & vanes. By doing in-house casting at SpaceX, we can accelerate nat gas turbines coming online by up to 18 months, which is a profound game-changer.”
That statement lays out the strategy clearly. Solar remains part of the long-term buildout, but natural gas is being treated as a bridge for AI infrastructure that needs power sooner than the grid can reliably provide it.
The AI power crunch behind the move
The AI industry is facing more than one supply problem. GPU shortages remain an issue, and Nvidia’s newest Blackwell chips are still facing lead times of several months. But computing hardware is no longer the only pressure point.
The physical power grid has become a major constraint. The International Energy Agency projects that global data center electricity use will roughly double by 2030. At the same time, GE Vernova says it is essentially sold out of production capacity through 2030, largely because of demand from AI infrastructure.
That combination helps explain why major hyperscalers are pursuing private gas-fired plants beside data centers instead of waiting for grid upgrades. Amazon, Google, Meta, OpenAI, and Microsoft are all part of that shift. After years of emphasizing wind and solar, the industry is now leaning on natural gas to bring data centers online faster.
For companies building at AI scale, power availability is not a background utility issue. It can determine where data centers are built, how quickly they open, and how much computing capacity can be deployed.
Why turbine blades are so hard to make
The specific component Musk highlighted is not a simple industrial part. The blades inside the hottest section of a gas turbine operate at temperatures around 3,000 to 3,600 degrees Fahrenheit. That is roughly 800 degrees hotter than the melting point of the metal alloy used to make them.
They survive those conditions because of a combination of internal cooling channels, thermal-barrier coatings, and the particular casting process used to make each blade. According to The Information, only four companies worldwide have mastered that process well enough to produce the parts at industrial scale, and all of them are currently tapped out.
The core difficulty is that each blade must be cast as a single, unbroken crystal. It has to be grown slowly inside a vacuum furnace and avoid microscopic seams that can cause ordinary cast metal to crack under stress.
That process is already difficult for the smaller blades used in jet engines. Power-plant turbine blades are considerably larger, which makes defect-free production at scale even harder.
If SpaceX succeeds, a Musk-controlled company would own a manufacturing capability that other AI infrastructure builders currently depend on a tiny group of suppliers to provide. That could give SpaceXAI an advantage that is difficult for competitors without deep manufacturing operations to copy quickly.
The pollution problem does not disappear
The same plan that could speed AI data center power also points toward more natural gas turbines being installed quickly. That is where the infrastructure story becomes a public health story.
Gas turbines already face federal lawsuits and peer-reviewed health research over the pollution they emit. In Memphis, SpaceXAI has used gas turbines to power its Colossus data centers since 2024. The NAACP has repeatedly accused the company of operating turbines without the permits or pollution controls required by federal law.
The group’s concern is that turbines can emit smog-forming compounds and hazardous chemicals like formaldehyde. Those pollutants are linked to asthma, respiratory disease, and certain cancers.
The Memphis site is also near neighborhoods that already face heavy industrial pollution. University of Memphis researchers said that in their own admittedly limited analysis, air pollution grew “slightly worse” because of the data center.
Memphis is the most visible example, but it is not the only place where the conflict is emerging. In Virginia’s “Data Center Alley,” a study commissioned by the Piedmont Environmental Council used the EPA’s own COBRA health-impact model to examine emissions from a single facility’s eight full-time gas turbines.
That study found emissions could reach more than 2.5 million people across multiple counties, with the heaviest impact falling on already-marginalized communities. It estimated 3.4 to 6.5 additional premature deaths a year and $53 million to $99 million in annual health-related damages.
What the turbine race says about AI infrastructure
Musk’s foundry plan highlights a broader shift in how AI infrastructure is being built. The bottleneck is no longer just chips, models, or software talent. It is also land, grid access, specialized manufacturing, and the ability to bring large amounts of electricity online quickly.
In-house casting could shorten one part of that timeline if SpaceX can execute. But faster turbine deployment would not settle the environmental conflict around data center power. It would likely make that conflict more urgent, because more gas turbines would mean more local decisions about permits, emissions controls, health risks, and who bears the burden of keeping AI systems running.
The industry’s near-term answer to the power crunch is becoming clearer. Natural gas is being used to supplement and bootstrap solar while companies wait for broader energy capacity. The harder question is whether moving faster on gas turbines can be reconciled with the pollution concerns already following those turbines into communities.