Elon Musk’s new contradiction: he advocates for electric cars and solar energy, but AI is pushing Space X into the natural gas business
Elon Musk has decided to tackle one of the biggest bottlenecks in artificial intelligence with a solution that conflicts with much of his energy-related rhetoric. SpaceX is building its own foundry in Texas to produce turbine blades and vanes, essential components for generating the electricity needed by its data centers.
The goal is to buy time. Musk claims that producing these components within SpaceX can accelerate the deployment of new turbines by up to 18 months and help avoid some of the delays that hinder the expansion of electricity capacity for AI. The paradox is clear: while Tesla has been investing in electric vehicles, batteries, and solar energy for years, Musk’s artificial intelligence initiatives require more natural gas at least in the coming years.
The Bottleneck That Has Changed the Rules
The problem isn’t just obtaining chips to train artificial intelligence models. Electricity has become another major bottleneck for data centers.
The electricity demand associated with these facilities is growing rapidly, and power grids don’t always can provide the necessary capacity within the timeframes required by the industry. The International Energy Agency predicts that data center electricity consumption will nearly double by 2030.
The result is a shift in strategy. Major technology companies are turning to on-site power generation near data centers, often using gas turbines, to avoid relying solely on new grid connections.
SpaceX is at the heart of this race. The company invested nearly $16 billion in artificial intelligence infrastructure during the second quarter of 2026.
This decision puts the entrepreneur in a difficult dilemma to ignore. While Tesla focuses on electric cars, batteries, and solar energy, and Musk claims that both Tesla and SpaceX are working toward achieving an annual solar manufacturing capacity of 100 GW each, its artificial intelligence infrastructure increasingly depends on natural gas.
A secret factory in Texas
The new industrial strategy is being developed in Bastrop, Texas, near the facilities where SpaceX manufactures Starlink terminals. The company acquired approximately 830 acres in that area between March and June 2026.
The goal is to manufacture gas turbine blades and vanes in-house. These components are precisely one of the main bottlenecks in the industry.
The manufacturing process is especially complex. The blades in the hottest parts of a turbine operate at temperatures ranging from about 1,650 to 1,980 degrees Celsius. To withstand such temperatures, nickel superalloys, internal cooling channels, thermal coatings, and highly precise casting processes are used.
Musk claims that manufacturing these components within SpaceX could accelerate the delivery of new turbines by up to 18 months.
The market is practically booked up until 2030
There is an industrial reason behind this move. The production of large gas turbines is concentrated among a small number of manufacturers, and their order books are under immense pressure due to the growth of data centers.
GE Vernova, one of the world’s leading manufacturers, has warned that its production capacity is practically fully committed until 2030.
SpaceX aims to eliminate one of the key components that most restrict the availability of these machines from its supply chain. It’s not just a matter of purchasing a turbine; the company wants to take control of part of the technology and manufacturing needed to produce it.
This focus on gas power didn’t start recently either. SpaceX has relied on mobile gas and diesel turbines to supply electricity to its artificial intelligence facilities and has invested around $1 billion in APR Energy, a company specializing in mobile power generation with a fleet capable of over 1 GW of capacity.
In Memphis, the Colossus data centers have used dozens of gas turbines as a solution for electricity generation independent of the grid. SpaceX also plans to replace part of this temporary power supply with a permanent natural gas power plant capable of producing about 1.2 GW.
Here lies the most controversial element of Musk’s strategy. Tesla has become one of the world’s leading symbols of the transition to electric vehicles. The company has helped accelerate the electrification of cars and has also developed stationary batteries and solar systems.
Musk himself now advocates for massive expansion of solar power generation. Tesla is working to achieve 100 GW of annual solar manufacturing capacity in the United States, while SpaceX has set a similar goal.
But the reality of data centers is different. Solar energy requires storage or backup to provide continuous power, while a gas turbine can generate electricity when needed and be deployed much more quickly than new electrical infrastructure.
That’s why Musk argues that natural gas will still be necessary for several years to complement and enable the expansion of solar energy.
The argument may be technically coherent from a supply security perspective, but it creates an obvious environmental tension: to develop an industry aimed at advancing a new generation of artificial intelligence, SpaceX is increasing its reliance on fossil fuels.
The problem doesn’t end with CO₂ emissions. Gas turbines also produce nitrogen oxides and other air pollutants.
The use of on-site power generation in data centers has already caused environmental and regulatory conflicts in the United States. In Memphis, organizations such as the NAACP have raised concerns about the operation of turbines linked to xAI facilities.
And the risks aren’t merely theoretical. An analysis conducted in Virginia using the U.S. Environmental Protection Agency’s COBRA health model estimated that emissions from a system consisting of eight gas turbines and 51 diesel generators could be associated with between 3.4 and 6.5 additional premature deaths per year, as well as $53 million to $99 million in annual health-related costs. This is a modeled estimate, not directly observed deaths.