Recently, SpaceX in partnership with Tesla announced plans to build a $55bn semiconductor foundry in Texas so as to secure their fabrication needs going into the future. What challenges has AI introduced in semiconductor supply chains, what exactly has SpaceX and Tesla announced, and could we see a new era of self-built semiconductors?

The Need For More Semiconductors in the Age of AI

The COVID pandemic of 2019-2020 saw massive changes across the world in almost every aspect of life; electronic engineers especially felt these changes. The massive semiconductor supply chains were so badly hit that they went through massive shortages and delays which resulted in the semiconductor industry having to change its practices.

In fact, the massive disruption to the semiconductor industry as a result of COVID and the resulting government intervention in the semiconductor supply chain has taught engineers, and governments alike, that the global mechanism for supporting and manufacturing semiconductors was (and still is) flawed and vulnerable.

However, the problems faced by semiconductor supply chains after COVID are nothing compared to the issues introduced by the rapid growth of AI businesses. As the world moves towards an AI-driven society, the need for more powerful AI has exponentially increased, and this has seen AI businesses purchase up all the available semiconductor manufacturing capabilities.

The net result of this growth has been consumers now faced with far greater prices for hardware, and this is now spilling over into everyday items including RAM, CPUs, memory, and even PCBs. As such, it comes as no surprise that some businesses are starting to rethink their semiconductor strategies and considering whether they should create their own semiconductors to ensure their ability to develop next-generation products.

SpaceX and Tesla To Build $55bn Semiconductor Foundry

Recognising the challenges faced by semiconductor supply chains, SpaceX recently announced that it will be building a $55 billion semiconductor manufacturing and advanced computing facility called Terafab. This facility will be constructed in Grimes County, Texas, and will be developed in partnership with Tesla.

This announcement alone clearly suggests that Elon Musk has his eye on the AI prize, and public filings have indicated that the facility could expand to a total investment of $119 billion if future phases are approved. Furthermore, the facility is described as a major investment in domestic semiconductor manufacturing capacity which suggests that the facility will be used to produce semiconductors for use in the US.

Elon Musk has repeatedly stated his desire to bring semiconductor manufacturing back to the US, and this move, with Terafab would realise this goal. As Elon Musk is well known for his ambitious projects, it is likely that Terafab will be more than just a facility that produces semiconductors; it could also be used to develop next-generation computing systems.

The announcement by SpaceX and Tesla comes at a time when the US is trying to shore up its semiconductor abilities, with concerns over semiconductor dependence after it was revealed that US wafer fabrication capacity saw a decline from 37% of global production in 1990 to under 10% in 2024, with most manufacturing now concentrated in East Asia.

Could We See A New Era in Self-Made Semiconductors?

It comes as no surprise to see companies such as Tesla and SpaceX looking to build their own semiconductors, but what is of interest is what this could mean for the semiconductor industry in general. If the cost of equipment continues to fall, and older nodes such as 250nm become more available and easier to manufacture, then we could enter a new era of self-made semiconductors. Companies that house their own fabrication unit could secure their own supplies, and this could even help to create a more adaptive technology. 

For example, fabs that are designed to produce semiconductors for the masses will have to focus on high-volume low-cost solutions, but those that are tied to a specific customer may not need to worry too much about price. As such, these facilities could be more open to using more exotic technologies or features that could help to produce more efficient designs.

Furthermore, the use of older nodes would significantly reduce the requirements for clean room and other complex infrastructure needed to make cutting-edge semiconductors. Thus, it becomes far easier to establish a foundry, and could even see new foundries being jointly run by multiple companies to all secure their own supplies.

This could then potentially lead to a new era of “communalism” whereby large groups of small companies pool their resources together to ensure that they can all get access to the latest semiconductors. Considering that some YouTubers have already demonstrated crude chips using simple equipment, if serious investment was put into this area it could become a lifeline for future custom semiconductors.

Of course, it should be understood that not every design needs to be the latest node, and moving away from the latest node does come with some advantages. Thus, it is clear that Elon Musk is determined to bring semiconductor manufacturing back to the US, and there is a good reason why he wants to do this. But whether this project succeeds or not remains unknown, and we can only hope that Elon Musk doesn’t bite off more than he can chew.