Key Takeaways
- Elon Musk confirmed SpaceX is building a foundry in Bastrop, Texas, to cast single-crystal nickel superalloy blades and vanes for natural gas power turbines.
- In-house casting aims to shorten turbine deployment timelines by up to 18 months, addressing the energy supply bottleneck slowing down AI data centers.
- The specialized casting sector is dominated by a duopoly: Howmet Aerospace and Berkshire Hathaway's Precision Castparts.
- Single-crystal superalloy casting requires vacuum furnaces where sacrificial ceramic cores and wax molds are destroyed during manufacturing, preventing competitors from reverse-engineering parts.
The 18-Month Energy Wall
AI compute expansion has run into an energy grid bottleneck. Running gigawatt-scale clusters requires dedicated power plants, and building new natural gas capacity is delayed by long equipment lead times. As John Coogan explained, “Over the weekend, Musk confirmed on X that SpaceX will start making blades and veins that go into natural gas power turbines.”
Securing commercial turbines often takes years. Musk claims that “by doing in-house casting at SpaceX, we can accelerate natural gas turbines coming online by up to 18 months.”
When a venture needs gigawatts of electricity to train models immediately, waiting on legacy suppliers costs more than constructing an advanced metallurgical foundry from scratch. SpaceX is bypassing the supplier queue by vertically integrating the single slowest physical component in the electrical supply chain.
The Physics of Single-Crystal Casting
Gas turbine blades operate under extreme mechanical stress and temperatures hot enough to melt ordinary steel. To survive those environments, the metal cannot have normal crystal grain boundaries, which act as structural failure points under heat.
This level of manufacturing has historically been confined to specialized aerospace suppliers. Coogan pointed out that “Howmet and Berkshire Hathaway's Precision Castparts are the two biggest players in the market for these parts, which also include blades and veins for the aerospace and defense sectors.”
Incumbents protect their process through consumable tooling rather than public documentation. “Once the parts are produced, the ceramic cores and wax used to make them are just destroyed, Howmet chief executive John Plant said at an investor meeting earlier this year,” Coogan highlighted. That destruction prevents rival machine shops from scanning or reverse-engineering the internal cooling channels.
Internal Scrap Rates vs. Vendor Queues
Traditional foundry suppliers run high-margin operations optimized for commercial yields and low scrap rates. A high rejection rate on a run of single-crystal blades ruins a public supplier's quarterly margins. Because of that, commercial foundries ramp capacity slowly and conservatively.
SpaceX operates under a different set of constraints. If an internal foundry scraps forty percent of its early casts, the cost of that wasted metal is trivial compared to the value of bringing an AI data center online 18 months ahead of schedule. Internal captive demand allows SpaceX to trade metallurgical scrap for calendar speed.
What to Do With This
Audit your primary operational roadmap tomorrow and pinpoint the single component or vendor with the longest lead time. Calculate what 12 months of delayed deployment actually costs your business in lost revenue. If that delay costs more than building dedicated internal tooling or buying out dedicated vendor capacity, stop waiting in the public supply queue and vertically integrate the bottleneck.