If you haven’t yet heard of Terafab, we’d like to give you the lowdown. Elon Musk has been looking to get into chip manufacturing for some time now. At one point, he was even thought to be interested in buying out Intel. Last month, he announced plans for a giant semiconductor factory, to be housed in a joint venture between Tesla, SpaceX and xAI. Musk’s starting point is as simplistic as it is implacable: the need for AI chips (and his own in priority), cannot be met by existing foundries(TSMC, Samsung, Intel…). Simply because they can’t scale fast enough.
Terafab: will Elon Musk also shake up the world of chips?
Musk wants to produce 1 terawatt (TW) of computing power per year, far more than the world currently produces AI chips. These chips will power tons of humanoid robots, satellites, autonomous vehicles and xAI infrastructures. To put it bluntly, Elon Musk’s production target is, depending on the analyst, either “very ambitious” or “totally insane”. It’s a project that will require an enormous mobilization of resources (and not just financial ones). Better (or worse, depending on opinion), the boss wants to vertically integrate everything (design, chip manufacturing, HBM memory and advanced packaging all under one roof). But whatever one may think of him, the animal still has a certain capacity to disrupt industries and make the experts lie. And at least he makes us dream and sometimes laugh. Bringing everything together within a single physical structure could accelerate development by enabling engineers to design, test and modify chips more rapidly. Once again, it’s a recipe the man knows well.
Is Intel an opportunist or a visionary?
Where things really start to take off is when Intel announces its involvement in the project. The same tech company that some had said was dead, the same company that had left the keys to the company to accountants for years, is now a partner in this crazy project! An announcement that adds yet another layer of uncertainty. Will Terafab develop its infrastructure from scratch or use existing Intel facilities such as processes 18A and 14A? The Intel boss not only posed for the photo, he also shared his enthusiasm in a way. And for the record, he’s not an accountant, but a veteran of the semiconductor industry.
Elon has a proven track record of re-imagining entire industries. This is exactly what semiconductor manufacturing needs today. Terafab represents a radical change in the way silicon management, memory and packaging will be built in the future (Lip-Bu Tan Intel CEO).
It’s also an important signal for Intel to get involved in such a project, confirming its return to the forefront. We can also see that this is yet another milestone in Intel’s ambition to relocate high-tech industries to the US. Frankly, we’re curious and enthusiastic, because one thing’s for sure: if the world of chips is boring as hell, with Musk it won’t be!
