Musk's Turbine Bet: Casting Blades In-House to Break AI's Power Bottleneck
The natural gas turbine casting shortage is now a bottleneck on par with GPUs — and SpaceX wants to solve it by building its own foundry.
Published: 2026-08-31 Category: Quick Take Sources: TechCrunch
The New Bottleneck Is a Metal Part, Not a Chip
For two years the story of AI infrastructure was simple: not enough GPUs. That's still true — Nvidia's newest Blackwell chips still carry lead times of several months. But a second, less glamorous constraint has muscled its way to the top of the pile: the physical power grid. The International Energy Agency projects global data center electricity use will roughly double by 2030, and GE Vernova, a dominant gas turbine maker, says it is essentially sold out of production capacity through 2030, driven largely by AI demand.
That's why hyperscalers including Amazon, Google, Meta, OpenAI, and Microsoft have stopped waiting on the grid and started building private gas-fired plants next to their data centers. After years of pitching wind and solar, they are all now leaning on natural gas to get capacity online faster. The catch: you need turbines to burn it, and the hardest part of a turbine to make is a single component nobody can easily scale.
Why Casting a Blade Is So Hard
The blades inside a gas turbine's hottest section run at roughly 3,000 to 3,600 degrees Fahrenheit — about 800 degrees hotter than the melting point of the metal alloy they are made from. That only works because of internal cooling channels, thermal-barrier coatings, and the exact way each blade is cast. And here is the kicker: 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.
Only four companies on Earth have mastered this at industrial scale, and all of them are tapped out. It's a hard process even for the small blades in jet engines; power-plant turbine blades are considerably larger, making defect-free production at scale far harder. This is the chokepoint Musk is targeting.
What SpaceX Is Actually Doing
Over the weekend, Musk confirmed what a secret foundry SpaceX has been quietly building in Bastrop, Texas, is for, responding to reporting from The Information that had already pieced together the details — job listings mentioning a "blades and vanes foundry," plus evidence SpaceX bought roughly 830 acres near its existing Starlink factory between March and June.
His framing on X was direct: 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 on nat gas turbine production is casting the blades and vanes. By casting in-house, Musk argues, he can accelerate nat gas turbines coming online by up to 18 months — "a profound game-changer," in his words.
The Strategic Angle That Matters
If SpaceX pulls this off — and it must be said, this is easier said than done — the move goes well beyond electricity. A Musk-controlled entity would hold a manufacturing capability that every other AI infrastructure builder currently depends on a tiny oligopoly for. That would give SpaceXAI an edge that's difficult for any well-funded but non-manufacturing competitor to copy quickly. This is classic Musk vertical integration, applied to the least glamorous layer of the AI stack: the power supply.
The Pollution Pushback Already Building
But turbines in the ground are already drawing lawsuits and peer-reviewed research over what they emit. In Memphis, where SpaceXAI has run gas turbines to power its Colossus data centers since 2024, the NAACP has repeatedly accused the company of operating them without the permits or pollution controls federal law requires. The concern: turbines emit smog-forming compounds and hazardous chemicals like formaldehyde, pollutants linked to asthma, respiratory disease, and certain cancers. University of Memphis researchers said air pollution has grown "slightly worse" near the site.
The same fight is playing out anywhere gas turbines have become the default fix. In Virginia's "Data Center Alley," a study commissioned by the Piedmont Environmental Council used the EPA's COBRA health model to assess a single facility's eight full-time gas turbines. Its estimate: emissions reaching more than 2.5 million people across multiple counties, with heaviest impact on already-marginalized communities, and 3.4 to 6.5 additional premature deaths a year — translating to $53 million to $99 million in annual health-related damages.
The uncomfortable truth is that the fastest path to cheaper AI power is also the fastest path to more local pollution. Whoever solves turbine casting fastest also inherits the accountability for what those turbines emit. Musk is betting he can out-engineer the bottleneck; the harder question is whether he can out-manage the blowback.
Based on reporting by Connie Loizos at TechCrunch.