Why SpaceX’s Single-Crystal Turbine Blade Plan Runs Into the Hardest Wall in Manufacturing

I have spent my career trying to compress the timelines of investment casting. At DDM Systems, we built the Digital Foundry to take castings from CAD to poured metal in days instead of months. So when someone announces they will disrupt casting, I pay attention. I usually cheer.

This time, I want to walk you through why I am skeptical.

SpaceX confirmed plans to build a foundry that will cast single-crystal nickel superalloy blades and vanes for both Raptor engine turbopumps and natural gas power turbines, with Elon Musk claiming the move could accelerate turbine deployment by up to 18 months.

The ambition is real. The demand is real. The physics, the yield curves, and the market structure are also real. Let me show you what sits between the announcement and the outcome.

The Technical Wall. Growing a Perfect Crystal, Every Time

Single-crystal casting is the most demanding process in metallurgy. Each blade must solidify as one continuous nickel superalloy crystal inside a vacuum furnace. Any flaw, even a hairline crack, renders the part unusable.

I have worked in ceramic shells and superalloy solidification for decades. The gap that trips up every newcomer is this.

Demonstrating the process is a different achievement than yielding it at scale.

You can cast your first successful single-crystal blade in a lab within a year or two. Getting from that first blade to robust, repeatable, commercially viable yield takes the rest of the decade. Yield losses at a newly built facility could exceed 50 percent for years, according to SemiAnalysis analyst Nigel Chiang. Reject rates on the most demanding alloys run into the double digits even at the most experienced foundries.

Size Makes Everything Harder

Here is a detail that gets commonly overlooked. Power generation blades scale the difficulty in a nonlinear way.

Power-plant turbine blades are considerably larger than jet engine blades, which makes defect-free solidification harder.
Larger cross-sections cool unevenly, and uneven cooling breaks the single crystal.
Internal serpentine cooling passages and abrupt thickness changes demand precise gradient control across the entire part.

At DDM, we have cast a directionally solidified industrial gas turbine blade with cast-in film cooling holes. I know firsthand what the thermal management on a 12-inch blade demands. Every inch of added length multiplies the ways a crystal can go wrong.

💡 The core economics a single set of 40 single-crystal blades takes 60 to 90 weeks to produce and costs over $600,000. At those stakes, a 50 percent scrap rate destroys the business case before it starts.

The Market Wall. A Duopoly Built on Decades of Know-How

The second barrier gets less press than the physics, and it matters just as much.

Howmet Aerospace and Berkshire Hathaway’s Precision Castparts hold roughly 80 percent of the global market for these components between them. There are four to five companies in the world that can do single-crystal casting at scale to commercial aerospace specifications. All of them are fully booked.

The barrier here is process knowledge, accumulated over decades of vacuum furnace operation, alloy development, and yield improvement. Capital cannot shortcut it. Howmet alone holds over 1,170 patents around this process family, and the patents are the visible part. The invisible part lives in furnace logs, operator judgment, and thirty years of scrapped parts that taught someone something.

Multiple turbine OEMs have tried to break this structure before by funding alternative suppliers and building internal foundries. None succeeded.

China as the Control Experiment

If you want an honest benchmark for how hard this is, look at a nation-state that has thrown enormous resources at the problem.

Only five countries have independently mastered the complete manufacturing chain for these blades. Even so, Chinese single-crystal blades have historically lagged badly. Public data shows the turbine blade in the F119 engine achieves a service life above 10,000 hours, while a comparable Chinese blade has come in under 4,000 hours.

A country with strategic imperative, unlimited patience, and aggressive technology acquisition has spent decades closing this gap and has still struggled. That tells you something about a four-year corporate timeline.

The Demand Story Is Genuine

I want to be fair to the other side of this argument, because the pressure driving SpaceX’s move is completely real.

GE Vernova, one of our own strategic partners, is essentially sold out of gas turbine production capacity through 2030. The company ended the second quarter of 2026 with 116 gigawatts of backlog and slot reservations, taking orders for 2031 delivery. Turbine blade lead times of 60 to 90 weeks are the bottleneck inside that bottleneck.

Meanwhile, the incumbents hesitate to expand. Howmet and PCC got burned in previous cycles when they built capacity ahead of demand commitments and had to take write-downs. For them, allocating capacity to long-tenor, high-margin aerospace programs is safer than short-tenor, higher-scrap gas turbine work.

The industry has a supply crisis, a demand supercycle, and two dominant suppliers with rational reasons to sit still. That is exactly the kind of frozen market that invites a disruptor to make noise.

My Read. The Announcement Is the Strategy

Here is where I land after weighing both walls.

I believe the announcement itself does most of the work, whether or not SpaceX ever ships a qualified power-class blade. Consider what the threat of entry accomplishes.

It pressures Howmet and PCC to expand capacity they have resisted expanding.
It creates leverage for take-or-pay agreements, where power customers and Musk-adjacent buyers guarantee volume, shifting expansion risk from suppliers to customers.
It keeps the acquisition option open. Buying established expertise remains the pragmatic path when building proves too slow.

SpaceX is also making one genuinely smart foundational move. Designing and building its own vacuum induction melting furnaces. Those furnaces cost 3 to 8 times more than standard induction furnaces and come with long lead times from a handful of global suppliers. Controlling that tooling infrastructure removes the first bottleneck and enables faster iteration. I respect that decision. It mirrors the logic that led us to build the Digital Foundry as a vertically integrated system rather than assembling someone else’s toolchain.

What This Means for You

If you buy castings, design turbines, or invest in this supply chain, three takeaways matter.

First, respect the yield curve. Process capability at 90 to 95 percent sounds close to done. In single-crystal work, the last few percent of yield contain most of the difficulty and all of the profit.

Second, watch the contracts, and the headlines will follow. If take-or-pay agreements for blade capacity start appearing over the next 18 months, you will know the announcement worked as leverage.

Third, digital tooling is where the real opening sits. The parts of this problem that can move fastest are the ones upstream of the vacuum furnace. Ceramic molds and cores built directly from CAD, with no wax tooling, remove months from the front of the process. That is the work we do every day at DDM Systems, and it is why I believe the future of casting is digital even where the crystal itself remains stubbornly analog.

⚠️ One caution before you write anyone off Musk has beaten confident skeptics in launch, in EVs, and in satellite broadband. I hold my skepticism with humility. The physics of a single crystal does not care about track records, and it will test his the same way it has tested everyone else’s.

The power supercycle is coming either way. The question worth watching is whether it gets served by a genuine new entrant, by incumbents finally forced to expand, or by a deal that makes the foundry unnecessary. I am watching the contracts.