Intel Officially Surrendered To Musk? Company Joins Terafab With A Vague Role
Intel confirmed its participation in the Terafab project, which is being developed by SpaceX, xAI, and Tesla. The chipmaker announced this via its official account.
Intel will focus on rethinking silicon manufacturing processes. The company plans to leverage its own expertise in chip design, fabrication, and final assembly to help reach Terafab’s goal — delivering up to 1 TW of computing power every year. Those kinds of volumes are targeted at future AI and robotics workloads.
Over the weekend, Intel hosted Elon Musk, with the visit happening shortly before the official partnership announcement. A month earlier, the entrepreneur had mentioned possibly working with Intel, but at the time the parties hadn’t signed any agreement. Musk explained the need to build their own production capacity by saying that existing contract manufacturers simply can’t scale up fast enough to meet Tesla’s future demand for AI hardware.
Previously, Tesla and SpaceX planned to build Terafab facilities in Austin, Texas. According to Musk, those fabs will produce chips for Tesla cars, Optimus robots, as well as equipment for SpaceX satellites and space data centers. He didn’t give any specific construction timelines. Intel hasn’t yet disclosed details about the process node, packaging methods, or production model within Terafab. Still, the mere fact that Intel Foundry — which is actively pushing its 14A process — is involved makes the project feel more real and puts it among the notable AI chip initiatives to watch this year.
It’s still unclear how far along the Texas plants are and when exactly Intel will start handing over its technology. One thing’s for sure: without a major manufacturer capable of handling cutting-edge nodes and massive scale, hitting 1 TW of compute per year would be practically impossible.
What do you think — will Terafab actually manage to deliver 1 TW a year, or is that way too optimistic even with Intel on board? Drop your take in the comments.
-
Intel Ditching Regular Buyers? CPUs to Get a Third More Expensive Due to AI Demand -
Intel vs. NVIDIA: who compresses textures better with a neural network? Intel hits 18x and 0.194 nanoseconds per pixel -
No “Ultra” and 4.8 GHz: Why Intel is Passing Off Weaker Chips as Core 7 -
Intel Arc B390 Vs. RTX 5060: Integrated Graphics Gets Absolutely Crushed By Discrete GPU Again -
"We need chips, so we'll make our own." Musk unveils ambitious Terafab plan. Will it work this time?

