Elon Musk claims to have overcome a significant AI constraint by personally producing a complex turbine component.
Musk verified on Saturday the purpose of a covert foundry SpaceX has been constructing in Bastrop, Texas, seemingly in anticipation of a story revealing its specifics. Earlier that day, The Information reported on the matter, citing job postings for a “blades and vanes foundry” and research by Corey Trinetti, an AI infrastructure specialist, who noted SpaceX’s acquisition of around 830 acres near its Bastrop Starlink facility from March to June.
Musk stated on X Saturday that “SpaceX and Tesla are both rapidly developing 100GW/year of solar production capacity, but natural gas will remain essential for supplementing and initiating solar power for several years. The primary bottleneck in natural gas turbine manufacturing is the casting of blades and vanes. By undertaking in-house casting at SpaceX, we can expedite the deployment of natural gas turbines by up to 18 months, representing a significant advancement.”
The rationale behind this initiative addresses a core obstacle currently confronting the AI sector. While GPU scarcities persist—Nvidia’s latest Blackwell chips, for instance, still have lead times extending several months—a parallel limitation has surfaced: the physical power infrastructure. The International Energy Agency forecasts that worldwide data center electricity consumption will approximately double by 2030, and GE Vernova, a gas turbine manufacturer, reports being virtually fully booked for production capacity until 2030, largely driven by AI infrastructure requirements.
This scarcity explains why constructing dedicated gas-fired power plants adjacent to data centers, rather than relying on the main grid, has become a prevalent approach for major cloud providers like Amazon, Google, Meta, OpenAI, and Microsoft. Following years of emphasis on wind and solar, these companies are now turning to natural gas to expedite the operational readiness of their data centers.
Regarding the casting challenge itself, The Information reports that gas turbine blades in the hottest zones operate at temperatures ranging from 3,000 to 3,600 degrees Fahrenheit—approximately 800 degrees above the melting point of their constituent metal alloy. This feat is achievable solely due to the blades’ internal cooling systems, thermal-barrier coatings, and the specialized casting method used for each blade. Only four global firms have perfected this casting process sufficiently for industrial-scale production, and all are currently at maximum capacity.
The particular difficulty stems from the requirement that each blade must be cast as a single, continuous crystal, meticulously grown within a vacuum furnace. This prevents the tiny seams that typically cause stress-induced cracking in conventionally cast metals. This process is challenging even for smaller jet engine blades; for the substantially larger power-plant turbine blades, producing them flawlessly at scale presents an even greater hurdle.
Should SpaceX succeed in this endeavor—a feat more easily conceptualized than executed, naturally—a Musk-led enterprise would possess a manufacturing capacity that all other AI infrastructure developers currently rely on a small group of suppliers for. This would provide SpaceXAI with a distinct advantage, challenging for any financially robust, non-manufacturing competitor to replicate swiftly.
However, this also implies a rapid increase in gas turbine deployment, and these turbines are already facing federal lawsuits and critical health research due to their emissions.
In Memphis, where SpaceXAI has operated gas turbines for its Colossus data centers since 2024, the NAACP has repeatedly alleged that the company runs these turbines without the necessary federal permits or pollution controls. The organization is concerned that such turbines release smog-generating compounds and dangerous substances like formaldehyde, which are associated with asthma, respiratory illnesses, and specific cancers. (The facility is located near communities already burdened by significant industrial pollution, and University of Memphis researchers, in their preliminary analysis, observed a “slightly worse” air quality due to the data center.)
However, Memphis merely represents the most prominent instance. This identical conflict is unfolding wherever gas turbines are adopted as the primary solution for data center power deficits. Within Virginia’s “Data Center Alley,” a report commissioned by the Piedmont Environmental Council, employing the EPA’s COBRA health-impact model, revealed that emissions from a single facility’s eight continuously operating gas turbines could affect over 2.5 million individuals across various counties—disproportionately impacting already vulnerable communities—and lead to an estimated 3.4 to 6.5 additional premature fatalities annually, equating to $53 million to $99 million in yearly health-related costs.
Similar instances and grievances are extensive.