Written by: Rita

The competition for computing power in AI data centers is shifting from chips to electricity. SpaceX has secured chip supply, and the blades and rotors of gas turbines are becoming new bottlenecks.

In a report released on August 16, Morgan Stanley analyzed SpaceX's strategic intentions regarding the new foundry being built in Bastrop, Texas. Only three foundries worldwide can produce the hot end blades and rotors required for large gas turbines, with orders already booked until 2030. SpaceX's solution is vertical integration, building its own foundry while meeting the demand for turbine engines in data centers and the turbo pumps for Starship Raptor engines.

Morgan Stanley maintains an overweight rating for SpaceX with a target price of $300, believing that the decline after the lock-up period provides an entry opportunity.

Building a Foundry to Break Through Power Supply Bottlenecks

SpaceX's recruitment page recently featured four positions related to the Bastrop foundry, involving automation, operations, and material engineering. Morgan Stanley believes that this job information is an important clue about SpaceX's strategic direction.

In a podcast in February, Musk revealed his intentions: to ensure sufficient electricity online, SpaceX and Tesla may need to manufacture turbine blades and rotors in-house. There are only three foundries globally producing these components, and orders are seriously backlogged.

The components of gas turbine hot ends must operate under extremely high temperatures and pressures, involving complex single crystal casting processes. This is the narrowest bottleneck in turbine capacity. The order book for mainstream turbine manufacturers is filled until 2029 to 2030, and new order EBITDA profit margins have risen from the historical average of about 5% to over 30%. To secure capacity, customers are placing orders at about $3,000 per kilowatt (around $1,000 per kilowatt in the first half of the 2020s) and paying about 25% in advance.

The logic behind SpaceX building its own foundry has three layers: breaking free from supplier timelines, optimizing product performance through first principles, and having a single foundry serve both aerospace and AI businesses to spread fixed costs.

The Next Stage of the Computing Power Race is Electricity

SpaceX AI's computing power scale depends on two variables: chips and electricity.

The chip aspect is basically resolved. An exclusive agreement with Nvidia secures at least 10GW of chip supply. The electricity aspect remains unclear. Musk proposed a target of 10GW by the end of 2027 during the second quarter earnings call, while Morgan Stanley is currently modeling only 5GW, with the gap reflecting uncertainty in power supply timeliness.

Morgan Stanley identified that SpaceX has secured about 3 to 4GW of gas turbine supply through procurement agreements and Musk's personal acquisition of APR Energy (a mobile power provider with a fleet of over 11GW), publicly disclosing a total commitment close to 8GW. The actual constraints fall on the blades and rotors, rather than the gas turbine units themselves.

Every GW of computing power represents approximately $50 billion in revenue opportunities each year. Morgan Stanley's SOTP target price breaks down as follows: Space business $18, Connectivity business $118, X and Grok business $8, enterprise AI business $165. The current stock price is about $141, and after deducting $127 for Space and Connectivity, the market is valuing enterprise AI at only about $13, approximately 1 times the 2028 EV/revenue, lower than the valuation levels of new cloud service peers.

How to Monetize Billions of Watts of Computing Power

The Morgan Stanley team estimates that the four major hyperscale vendors will add about 25GW of computing power by 2027 (excluding TPU). If SpaceX AI reaches 5GW by the end of 2027, it would represent about one-fifth of the total new additions from the four major cloud vendors.

Upside risks include Starship reuse progress exceeding expectations, accelerated growth in Starlink capacity, enterprise AI monetization exceeding expectations, and accelerated Cursor ARR. Downside risks include slowing Starship reuse pace, slowing Starlink user growth, enterprise AI monetization not meeting expectations, longer power supply cycles, and higher capital requirements.

SpaceX is replicating its vertical integration experience in the aerospace sector into the AI infrastructure field. Morgan Stanley believes that sell pressure after the lock-up period provides a window for long-term investors to enter. If SpaceX can prove it can scale computing power to 5GW or higher before the end of 2027, the valuation of its enterprise AI business will be significantly revised upward.

Disclaimer

This article is a compilation and interpretation of third-party brokerage research reports (Morgan Stanley, August 16, 2026) by Trend Research, combined with public market information. The ratings, target prices, earnings forecasts, and related judgments quoted in this article are the views of the brokerage's analysts and represent their respective institutions' positions, not the views of Trend Research, and do not constitute any investment advice.

The market carries risks, and decisions should be made independently. This article should not be used as the basis for buying or selling any securities.