Musk Claims Accelerated Shipments for Texas Terafab Amidst Skepticism

One terawatt. That is the staggering amount of artificial intelligence computing power Elon Musk claims his new Texas-based semiconductor facility will produce annually once fully operational. This figure represents the centerpiece of Terafab, a joint megaproject between Tesla and SpaceX that aims to manufacture logic, memory, and advanced packaging architectures under one roof. The facility is designed to be vertically integrated, meaning it will handle every stage of chip creation internally rather than relying on external foundries.

On October 20, during a live broadcast of SpaceX's Q3 2026 earnings call, Musk stated that critical equipment suppliers, including ASML, have agreed to expedite shipments to support the project's aggressive schedule. He emphasized that the company is targeting volume chip production at the primary site in Gibbons Creek, Texas, by the third or fourth quarter of 2028. However, the Tesla and SpaceX CEO tempered these expectations by describing the timeline as highly aspirational and noting that it carries big error bars. He expressed appreciation for the suppliers who are accelerating deliveries to help achieve these aggressive dates.

The reliance on external partners is substantial given the scale of the endeavor. Intel joined the venture in April 2026 to help co-develop silicon fabrication technology, providing access to its 14A manufacturing node. The facility will utilize traditional extreme ultraviolet lithography systems, for which ASML is the world's sole commercial supplier. Each of these lithography systems costs approximately $400 million. Unofficial sources suggest the first EUV tool for Terafab was slated for delivery in mid-2027, though it remains unclear if this aligns with the expedited timeline Musk referenced. Other key partners include Applied Materials, Lam Research, KLA Corporation, and Tokyo Electron, which are providing front-end processing and metrology machinery. Additionally, Tesla and SpaceX are in talks with about 10 South Korean equipment makers, including Hanmi Semiconductor and HPSP.

Musk’s confidence in the hardware pipeline extends beyond the factory floor. He announced plans to bring a massive training cluster containing 440,000 Nvidia Vera Rubin GPUs online by the end of the second quarter of 2027. This move is part of a broader strategy to secure next-generation architectures aggressively. The chips produced at Terafab are intended to power xAI clusters, self-driving vehicles, Optimus humanoid robots, and an orbital AI data center for SpaceX, while also supplying the broader AI industry to mitigate future shortfalls from external foundries like TSMC.

Despite the detailed supply chain plan, significant questions remain about the feasibility of the 1 terawatt goal. Research firm Gartner estimates that the total global capacity for AI-optimized data centers is roughly 44 gigawatts as of late 2026. This means Terafab's planned annual output would exceed the entire current global capacity by a wide margin. For perspective, Gartner’s estimate is only 4.4% of what Terafab aims to produce. While data from EpochAI indicates that global AI computing capacity has grown by roughly 3.3 to 3.4 times per year since 2022, experts have questioned whether this growth rate can support such a dramatic leap in a single facility.

Musk also provided context on his AI computing ambitions during the same broadcast. He expects his AI companies to have between 2.5 and 3 gigawatts of computing capacity active by the end of 2026. He expressed high confidence in adding an incremental 10 gigawatts in 2027, stating he would be surprised if they did not reach a total of 13 gigawatts of AI compute by the end of that year. This target stands in contrast to other major projects, such as the restructured 10 gigawatt Project Stargate and Meta's 5 gigawatt Hyperion project, both of which are expected to reach full capacity in the 2030s.