In short
Elon Musk says SpaceX is building a Texas foundry to cast turbine parts that could speed gas-turbine deployment for AI data centers by up to 18 months. The plan may ease an infrastructure bottleneck, but it also raises fresh pollution and public-health concerns.
- Musk says SpaceX is building a foundry in Bastrop, Texas, to cast turbine blades and vanes in-house.
- The goal is to relieve a key bottleneck in gas-turbine production for AI data center power.
- Hyperscalers are increasingly using natural gas to bypass slow grid connections and get AI facilities online faster.
- Environmental groups and public-health advocates warn that more turbines mean more local pollution and health risk.
- The project could give Musk a rare manufacturing advantage in a critical AI infrastructure supply chain.
Elon Musk says SpaceX has begun building a foundry in Bastrop, Texas, to cast the difficult turbine parts that have become a major bottleneck for gas-fired power plants serving AI data centers. The move could shorten turbine delivery times by as much as 18 months, but it also threatens to accelerate the spread of local air pollution from the backup power source many tech giants are now leaning on.
On Saturday, Musk confirmed the project after new reporting narrowed in on the site’s purpose. The foundry appears tied to a broader effort to help supply natural-gas turbines at a time when electricity demand from artificial intelligence is straining both the chip supply chain and the power grid. The plan, if successful, would give Musk a rare manufacturing edge in one of the most constrained parts of AI infrastructure.
But the same strategy that could make new data centers easier to power also risks worsening an already heated public-health fight. Gas turbines are increasingly being used to get large computing campuses online sooner, and communities near those facilities are already challenging emissions, permits and the pace of industrial expansion.
What Musk says SpaceX is building in Bastrop
SpaceX is building a foundry in Bastrop to cast turbine blades and vanes, according to Musk’s own description of the project and reporting that identified job listings referring to a “blades and vanes foundry.” The site is near SpaceX’s existing Starlink manufacturing footprint and sits on land the company acquired in recent months.
Musk said the in-house casting operation is meant to remove a major choke point in turbine production. In his view, making the parts internally could speed the arrival of gas turbines by up to a year and a half, helping bridge the gap while solar capacity scales up.
Musk said SpaceX and Tesla are moving to expand solar production rapidly, but argued that natural gas will still be needed for several years as a supplement and starting point for that buildout. He described blade-and-vane casting as the limiting factor in turbine output.
The logic is straightforward: if one of the hardest-to-make parts can be produced under Musk’s own roof, the company no longer has to wait in line behind the handful of suppliers that currently dominate the market.
Why AI data centers are driving a new race for gas power
The immediate reason for this investment is not consumer electricity demand. It is AI.
Large-scale AI systems require enormous amounts of power, and the industry has run into two simultaneous constraints: not enough advanced chips and not enough electricity delivered fast enough by the grid. Even as GPU supply improves, the physical buildout of data centers is running into permitting delays, transmission bottlenecks and utility timelines that can stretch for years.
That is why gas-fired generation has become the quickest practical workaround for many hyperscalers. Instead of waiting for a utility interconnect, companies can place power equipment near a data center and bring capacity online much faster.
- GPU lead times remain long for cutting-edge chips.
- Data center power demand is rising faster than grid expansion in many regions.
- On-site gas turbines offer a faster path to operation than waiting for utility upgrades.
- Major cloud and AI companies are increasingly using natural gas as a bridge fuel.
That shift is visible across the industry. Amazon, Google, Meta, OpenAI and Microsoft have all moved toward gas as a practical way to launch or expand AI facilities sooner, even after years of emphasizing wind and solar procurement.
How does turbine casting create a bottleneck?
It does because the hottest turbine components are among the most difficult industrial parts to make at scale. The blades and vanes in the high-temperature section of a gas turbine must survive conditions that push well beyond the melting point of the alloys they are made from.
According to the underlying reporting Musk was responding to, those parts can operate in temperatures around 3,000 to 3,600 degrees Fahrenheit. They survive only because of internal cooling channels, advanced coatings and a casting process that has to be extremely precise.
Why is the process so hard?
It is hard because the part must be cast as a single crystal, without the microscopic grain boundaries that make ordinary metal more vulnerable to cracking under stress. The process happens slowly inside a vacuum furnace and demands exact control over temperature, chemistry and shape.
That is difficult even for smaller turbine blades used in jet engines. Power-plant turbine blades are larger and more demanding, which makes industrial-scale production even harder.
Only a small number of companies worldwide have mastered this at the necessary scale, and they are reportedly sold out. That scarcity has become a real choke point as AI developers rush to secure more power.
| Issue | What it means | Why it matters for AI |
|---|---|---|
| Turbine blade casting | Specialized single-crystal parts for gas turbines | Limits how fast new power plants can be built |
| AI electricity demand | Data centers are using far more power | Forces operators to find faster generation options |
| Grid delays | Utility connections can take years | Makes on-site gas power attractive to hyperscalers |
| SpaceX foundry plan | In-house manufacturing of hard-to-source parts | Could shorten turbine deployment timelines |
Why this could give Musk an unusual advantage
If SpaceX can manufacture the parts reliably, it would control a capability that most AI infrastructure builders do not have. In practical terms, that could allow Musk-linked companies to move faster than competitors that still depend on a narrow supplier base for the same components.
That matters because speed is now one of the biggest competitive advantages in AI infrastructure. Whoever can secure power, land, equipment and permits first is better positioned to deploy more models, more compute and more revenue-generating capacity.
A manufacturing edge in a sector crowded with well-funded rivals can be difficult to copy. Unlike software, industrial capacity cannot usually be scaled overnight. Tooling, expertise, quality control and supply-chain relationships all take time to build.
That is why the Bastrop project is notable beyond SpaceX itself. It may be one of the clearest examples of a tech founder trying to intervene directly in a physical supply bottleneck that is shaping the entire AI buildout.
What are the pollution concerns?
The concern is that faster turbine deployment also means more combustion-based power at a time when communities and regulators are already questioning where that power is placed and who bears the health burden.
Gas turbines emit pollutants associated with smog, respiratory illness and other health risks. In addition to nitrogen oxides and fine particulates, the facilities can release hazardous compounds including formaldehyde. Public-health advocates argue that those emissions are especially harmful when turbines are built near neighborhoods that already face industrial pollution.
That is the main criticism facing the Memphis site associated with SpaceXAI, where gas turbines have been used since 2024 to power Colossus data centers. The NAACP has repeatedly said the company operated turbines without the permits and pollution controls required under federal law.
Advocates say the Memphis turbines are part of a broader pattern in which communities near industrial corridors absorb the health costs of rapid AI expansion while companies capture the benefits of faster compute deployment.
University of Memphis researchers, in a limited analysis, said air pollution around the site became slightly worse after the data center arrived. Even that modest finding has fueled concern in a city that already lives with heavy industrial exposure.
How big is the health and environmental dispute?
It is larger than one facility in Tennessee. Similar fights are breaking out wherever gas turbines have become the default answer to missing grid capacity.
In Virginia’s Data Center Alley, a study commissioned by the Piedmont Environmental Council and based on the EPA’s COBRA health-impact model found that emissions from a single facility’s eight full-time gas turbines could affect more than 2.5 million people across multiple counties. The analysis estimated 3.4 to 6.5 premature deaths a year and $53 million to $99 million in annual health-related damages.
Those figures have become a flashpoint because they show how a local power workaround can create regional consequences. The burden does not stop at a fence line. It can spread across counties, affecting neighborhoods that may have had little say in the project’s approval.
Environmental groups and some residents argue the AI industry is externalizing its speed problem onto nearby communities. Developers, by contrast, say they need temporary solutions if they want to bring capacity online in time to meet demand.
Why natural gas is becoming the bridge fuel of the AI boom
Natural gas is attractive because it is available, dispatchable and quicker to deploy than many alternatives. Solar and wind remain central to long-term decarbonization plans, but they do not always solve the immediate need for round-the-clock power at an AI campus that must run at full load.
For that reason, companies are increasingly pairing renewable commitments with gas-powered backup or primary generation. In Musk’s telling, gas is a temporary bridge that helps bootstrap solar at the scale he wants.
Critics say the bridge could become a permanent road. Once a gas plant or turbine array is in place, it is often hard politically and economically to replace it quickly with cleaner infrastructure.
- Gas can be deployed faster than many alternatives.
- It provides firm power for data centers that cannot tolerate downtime.
- It can be positioned as a short-term measure while renewables expand.
- It creates immediate emissions and health concerns where turbines are installed.
What happens next for SpaceX and the AI infrastructure market?
The next stage will be whether SpaceX can actually execute the casting process at industrial scale. That is far from guaranteed. Making a turbine blade is one thing; producing a steady stream of defect-free components that meet utility-grade standards is another.
If the project works, it could make Musk one of the most vertically integrated figures in AI infrastructure, with influence not just over chips and software-adjacent systems but over the industrial hardware needed to power them.
If it fails, it will still highlight how brittle the current AI buildout has become. The industry is no longer constrained only by data and compute. It is constrained by factories, furnaces, permits, transmission lines and the communities caught in between.
Either way, the Bastrop foundry is a reminder that the AI boom is becoming deeply physical. The race is no longer just to build better models. It is to secure the metals, machines and megawatts needed to run them.
Timeline of the turbine and AI power story
| Period | Development | Why it matters |
|---|---|---|
| 2024 | Gas turbines begin powering SpaceXAI’s Colossus data centers in Memphis | Shows how quickly AI operators are turning to on-site combustion power |
| March to June 2026 | SpaceX acquires about 830 acres near its Bastrop, Texas, operations | Signals expansion of industrial capacity around the foundry site |
| August 2026 | Reporting identifies the foundry’s likely purpose and Musk confirms it publicly | Brings the project into the open and frames it as an AI infrastructure play |
| Coming years | SpaceX aims to speed turbine output while solar capacity scales | Could reshape who controls a critical bottleneck in AI power development |
Bottom line
Musk’s foundry plan is a bid to attack a supply-chain bottleneck that is slowing down new gas turbine deployments for AI data centers. It may help companies bring power online faster, but it also deepens a separate fight over pollution, permitting and the public-health costs of making AI infrastructure move at startup speed.
Frequently asked questions
Why is SpaceX building a foundry in Texas?
SpaceX is building the foundry to cast turbine blades and vanes, which Musk says are a major bottleneck in gas-turbine production. The goal is to speed up power equipment needed for AI data centers and reduce reliance on a small group of outside suppliers.
How could the foundry affect AI data center power builds?
It could shorten turbine delivery timelines by as much as 18 months, according to Musk. That matters because many AI operators are using gas turbines to get new data centers online faster while grid connections and utility upgrades lag behind demand.
Why are gas turbines controversial near data centers?
Gas turbines are controversial because they emit pollutants linked to smog and health problems, including respiratory disease and cancer risks. Communities near these facilities often argue that the rush to power AI infrastructure shifts environmental and health costs onto nearby residents.
What is the main technical challenge in making turbine blades?
The main challenge is producing single-crystal blades and vanes that can withstand extreme heat without cracking. The parts require advanced casting, internal cooling channels and protective coatings, and only a handful of companies can make them at industrial scale.
Are other AI companies using natural gas for power too?
Yes. Amazon, Google, Meta, OpenAI and Microsoft have all increasingly turned to natural gas as a faster way to secure power for data centers. The approach is often framed as temporary, but critics worry it could lock in more fossil-fuel infrastructure.
