Elon Musk announces Terafab. This is the name for a chip factory tailored for advanced AI chips, built by Tesla in collaboration with SpaceX. The plan comes with a hefty price tag and several question marks.
Elon Musk announces Terafab. This will be the name for a new chip factory where Elon Musk intends to bring the entire microchip production process together. Under the Terafab roof, Musk wants to house chip design along with lithography, memory production, packaging, and testing.
The project Musk has in mind will essentially consist of two fabs. Integration will be key: each fab will produce a single chip. One chip will serve autonomous vehicles and robots, while the other will be for AI and AI satellites in space.
World’s largest chip factory
The budget for the entire venture is estimated at twenty to 25 billion dollars. If successful, the proposed Terafab would become the largest chip factory in the world by a significant margin. The long-term ambition is to process one million wafers per month, accounting for one terawatt of created computing power annually. For comparison: the proposed production volume corresponds to 70 percent of TSMC’s current production capacity across all its factories.
The Terafab is being built by Tesla and SpaceX, with xAI also participating. The project will be located next to Musk’s Gigafactory in the US state of Texas (pictured).
Galactic ambitions
Musk claims that today’s global chip output only accounts for three percent of his companies’ future needs. He envisions humanity eventually heading into space, with cities on the Moon and Mars, and believes this is only possible if his companies receive enough AI chips. Musk himself describes the project as: “the next step toward (becoming) a galactic civilization.”
There are some gaps in this ambitious or perhaps even megalomaniacal plan. After all, SpaceX, Tesla, and xAI have never built their own chips before. Yet, Terafab is expected to deliver 2 nm chips immediately. Currently, only TSMC is capable of building microchips with such tiny components.
More than just buying machines
Building advanced microchips is not easy. Even with decades of experience and billions in research and development, it is a monumental feat for specialized companies like TSMC and Intel to evolve from one production node to the next. The step from 5 nm to 3 nm, for example, is incredibly complex, even when you have all the right machines in-house.
Those machines partly come from ASML. These are lithography systems. These specialized pieces of equipment cost hundreds of millions each. Extensive research is required just to use them correctly. Imec in Leuven specializes in this, and is therefore the first to work with ASML’s latest High NA EUV machine. Chip manufacturers sponsor that research because it is too complex to carry out alone.
From the sidelines to the frontrunners
With the Terafab, Musk must demonstrate that all that complexity is actually fairly trivial for existing and experienced chip manufacturers. Without any experience in the field, Musk wants to catapult himself to the absolute frontrunners of chip production. It is therefore no surprise that a concrete timeline for the project is missing.
Elon Musk does have experience announcing absurd-looking projects. They almost never succeed within the planned timeframe, and sometimes they disappear from the schedule entirely.
In other cases, Musk surprises friend and foe by ultimately succeeding in his goal. For instance, SpaceX also came from nowhere to eventually become more successful than established rocket companies. However, 2 nm chip production isn’t rocket science: it’s more complex. If Musk actually succeeds in building his Terafab according to the proposed plans, he will transform the entire chip sector in one fell swoop.