Loading page…
The Bastrop facility targets a scarce component inside gas turbines, a manufacturing constraint distinct from the grid, permitting and pollution fights that can still hold up AI power projects.
Listen to this story
SpaceX is vertically integrating a rare turbine-manufacturing step as AI data-center developers struggle to secure power equipment. Its Bastrop, Texas, facility will make single-crystal blades and vanes for natural-gas turbines, with Elon Musk claiming up to 18 months could be saved. The opportunity is substantial because GE Vernova’s turbine capacity is reportedly booked through 2030, but the project does not solve permitting, emissions, or siting constraints.
The foundry follows SpaceX’s acquisition of roughly 830 acres near its Bastrop Starlink factory and job listings referencing the operation.
Turbine hot-section blades face 3,000–3,600°F temperatures and require cooling channels, coatings, and single-crystal casting.
TechCrunch reports only four companies have mastered industrial-scale casting; GE Vernova says production is effectively committed through 2030.
SpaceX is building a foundry in Bastrop, Texas, to cast the blades and vanes used in natural-gas turbines. Elon Musk says bringing that work in-house could move turbines into service up to 18 months sooner, targeting an increasingly important constraint on the power supply for AI data centers.
The project focuses on a narrow but consequential piece of the power stack. Musk said blade-and-vane casting is the limiting factor in natural-gas turbine production. He also said SpaceX and Tesla are each building 100 gigawatts of annual solar-production capacity, while gas will be needed to supplement and bootstrap solar for several years.
The site emerged after job listings referenced a blades-and-vanes foundry and land-purchase findings showed SpaceX had acquired roughly 830 acres near its existing Bastrop Starlink factory between March and June. Musk confirmed the purpose on August 30.
The blades in a turbine’s hottest section operate at roughly 3,000 to 3,600 degrees Fahrenheit, hotter than the melting point of their metal alloy. They depend on internal cooling channels, thermal-barrier coatings and single-crystal casting: each blade must be formed as one crystal without microscopic seams that can crack under stress. Larger power-plant blades make that exacting process harder still.
That technical difficulty explains why a foundry could matter beyond SpaceX’s own operations. TechCrunch reported that only four companies have mastered the casting process at industrial scale and that all are fully committed. GE Vernova has said its turbine-production capacity is effectively booked through 2030, largely because of AI-infrastructure demand.
AI builders are seeking gas generation because it can add power without waiting for grid upgrades. Global data-center electricity use is projected to roughly double by 2030, and Amazon, Google, Meta, OpenAI and Microsoft have pursued private gas-fired plants alongside data centers as a faster route to power, according to TechCrunch.
But a turbine delivered sooner is not automatically a turbine allowed to run. At the SpaceX-linked xAI Colossus data centers in Memphis, gas turbines have operated since 2024. The NAACP has repeatedly alleged that SpaceX ran turbines without required permits or pollution controls; University of Memphis researchers found, in a limited analysis, that the data center made local air pollution slightly worse.
Bastrop therefore addresses one specific choke point: the capacity to make the turbine’s hardest parts. Whether SpaceX can reliably produce defect-free power-plant blades at scale remains the central execution question. Even if it can, data-center projects will still face the separate questions of where generation is built and how it is permitted.
Loading discussion...
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.