SpaceX Is Building Its Own Turbine-Blade Foundry and the Casting Oligopoly Felt It Monday Morning
Elon Musk confirmed over the weekend that SpaceX is constructing a dedicated industrial gas-turbine blades-and-vanes foundry in Bastrop, Texas, targeting the tightest bottleneck in natural-gas turbine production. Musk said in-house casting could pull turbine commissioning forward by up to 18 months. The market responded Monday with precision: Doncasters (DPC) dropped roughly 12%, Howmet (HWM) fell about 7%, and GE Vernova (GEV) dipped 1.6% before recovering.
The Bastrop Foundry Is Already Being Staffed
The Information traced the project through SpaceX's Bastrop hiring activity and land purchases – approximately 830 acres acquired between March and June – before Musk said a word publicly. Job postings, some visible for months, seek engineers for:
- Single-crystal, directionally solidified, and equiaxed nickel-superalloy castings
- Wax-injection and ceramic-core molds
- Controlled-solidification furnaces
- Complex internal cooling geometries
That reads like a checklist of the production capabilities behind Howmet's and Doncasters' margins.
One critical distinction: this is advanced investment casting, the classical process incumbents have spent decades perfecting. SpaceX has not unveiled a secret alloy or an additive-manufacturing shortcut. It is recreating the supplier's process stack inside a captive operation.
Who Bleeds, and How Much
DPC is the cleanest casualty. Industrial gas turbines accounted for 39.4% of its Q2 2026 revenue – $106 million, up 42% year-on-year.
Howmet's gas-turbine exposure is meaningful (roughly $1 billion in 2025, expected to exceed $2 billion within five years) but represents only about 11% of total company sales, cushioned by aerospace, defense, and aftermarket.
GEV, as the OEM sitting on 116 GW of Gas Power backlog, could benefit if additional blade supply accelerates deliveries. Monday's price action matched this hierarchy almost perfectly.
Howmet's Q2 Engine Products adjusted EBITDA margin of 37.7% – up 470 basis points year-on-year, with pricing increases exceeding inflation pass-through – is precisely the kind of scarcity rent that attracts vertical integration. DPC's comparable margin was 23.5%. The gap reflects scale, yield history, and aerospace mix.
The Demand Buffer Incumbents Are Counting On
GE Vernova is ramping annualized turbine output from roughly 20 GW today to 30 GW by 2030. Howmet's newest Japanese casting facility is already allocated. Management describes industrywide capacity additions as a 2028–2030 event. In a market this undersupplied, SpaceX's blades could be entirely additive – absorbed by excess demand without displacing a single incumbent shipment.
A calibrated expectation: Howmet Engine Products adjusted EBITDA margin dips below 36% in at least one quarter by Q3 2027, at roughly 55% probability.
The Moat Is Penetrated, Not Gone
SpaceX brings credible metallurgical experience from years of solving blade-cracking problems in Raptor turbopumps, where thermal and mechanical stress is ferocious. That learning transfers. Where SpaceX has not demonstrated competence is at industrial scale: repeatable high yield on large single-crystal airfoils, tens of thousands of hours of creep and oxidation data, OEM qualification, and warranty economics.
When a Customer Rebuilds the Hardest Manufacturing Process on Earth
SpaceX does not need to match Howmet's per-blade cost. Its economic hurdle rate is entirely different: internal foundry expense weighed against the opportunity cost of delayed power and stranded AI compute. If a captive blade runs at worse yield but pulls a multi-hundred-megawatt installation online months earlier, the program generates an exceptional enterprise-level return. That asymmetric calculus – where the customer's alternative cost dwarfs the supplier's margin – is what makes captive integration corrosive to scarcity pricing.
SpaceX may only be the first mover. Amazon, Microsoft, Meta, and sovereign AI programs face the same arithmetic. If any of them start funding their own casting capacity, the erosion of bargaining power compounds. The mechanism of value destruction is not volume displacement; it is the demonstrated willingness of customers to manufacture around the bottleneck, compressing the present value of future scarcity rents even while current order books remain full. Industry revenue can keep growing. Terminal margins cannot be capitalized at 40x if the largest buyers are building their own foundries.
Not investment advice.