Elon Musk has finally unveiled one of his most ambitious projects yet, as with TeraFab, Tesla's CEO decides the "future of civilization."
Earth Only Meets 2% of Compute Demand That Musk Intends to Achieve, Which is Where TeraFab Comes In
Musk's entry into the chip industry has been talked about several times in the past as well, since a few months ago at Tesla's shareholder meeting, he envisioned the bold concept of a 'TeraFab', where the idea was to cater to the compute demand generated by Tesla, SpaceX, and xAI. And now, in a dedicated showcase, Tesla's CEO has given a rundown of how the TeraFab operation would look like, and, based on the details disclosed, it appears that the output scale of the facility planned is a figure that no one could have ever imagined before, and probably makes the likes of TSMC, Samsung and Intel look like 'tiny' players.
Diving into the details of TeraFab, Musk says the chip output is expected to reach a whopping 1 terawatt of compute power per year, including logic chips, memory, and advanced packaging. TeraFab was reported to focus on 2nm production, meaning the fab will jumpstart with cutting-edge production from the very beginning. The chip fab also integrates the concept of "rapid, recursive improvement", where the facility will have chip design, fabrication, testing, packaging, and masking all in unified production lines, allowing TeraFab to have recursive local self-improvement.
To the best of my knowledge, this doesn't exist anywhere in the world where you've got everything necessary to build logic, memory, and do packaging, and test it, and then do masks, improve the masks, and keep looping it. So in a single building, we can create a mask, make the chip, test the chip, make another mask, and have an incredibly fast recursive loop for improving chip design.
- Elon Musk
The facility will be built in Austin, Texas, where Tesla is headquartered, and will be a joint effort among all of Musk's companies. As for how fab production would be managed, given the gigantic scale, it is disclosed that 80% of the compute would go to space, since he believes Earth's power constraints make deployment almost impossible. Using Starship delivery channels and the power of the Sun, Musk believes he could deploy "terawatts" of computing power in space, which is why he says TeraFab is vital to leading us to a galactic civilization.
TeraFab's first major product would be the AI5 chips used in Tesla's FSD, Robotaxi, and Optimus robots. The AI5 will account for the 20% ground-based production split, and the remaining 80% of space-based compute will come from the newly unveiled D3 chips. These are custom-designed units for orbital AI satellites, built to run in harsh environments and likely to address thermal constraints when deployed in space. Since AI5 is Tesla's immediate priority, TeraFab's operations would be dedicated to bringing production lines online for those chips.
Talking about TSMC and other chip partners, Musk said he's "very grateful" to those companies, yet he also claimed the production numbers they have achieved are nominal compared to what TeraFab intends to do.
We’re very grateful to our existing supply chain, to Samsung, TSMC, Micron, and others… but there’s a maximum rate at which they’re comfortable expanding. That rate is much less than we would like… and we need the chips, so we’re going to build the Terafab.
If you add up all the fabs on Earth combined, they’re only about 2% of what TeraFab will produce.
- Elon Musk
TeraFab is centered on the narrative that Musk has done "impossible" things before. Yet, many experts have argued that the constraints of a semiconductor project make it almost impossible to achieve the scale Musk intends, at least within the timeline he has specified. When you factor in the limitations of ASML's pace of building lithography equipment and the global demand for it alone, one will realize that TeraFab would need to integrate several manufacturing elements vertically to achieve its objective.
We won't go into the debate over whether Musk would achieve the goals he specified above, but the more interesting question is what type of process nodes we're looking at. It is rumored that Tesla could enter into a licensing agreement with Samsung for 2nm production, but that remains to be seen.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
