Terafab Timeline: Musk Promises 1 Terawatt Annual Chip Output by 2028

One terawatt. That is the annual computing power output that Tesla and SpaceX plan to generate from a single physical facility in Texas. This ambitious target sits at the heart of Terafab, a joint venture designed to manufacture the logic, memory, and advanced packaging components required for next-generation artificial intelligence hardware. The project represents a significant shift in strategy for the two companies, which are moving beyond merely designing chips to owning the entire manufacturing pipeline.

The sheer scale of this goal requires a closer look at the numbers to understand its magnitude. To put the 1 terawatt figure in perspective, research firm Gartner estimates that the total global capacity for AI-optimized data centers is currently only about 44 gigawatts as of late 2026. This means Terafab’s planned annual output is roughly 23 times the entire existing global capacity. Furthermore, if the facility operates as intended, its commercial value would place it within the top 20 global country gross domestic products. For context, Switzerland currently holds the 20th spot with a GDP of approximately $1.14 trillion.

Elon Musk, who serves as the CEO for both partners in the venture, announced these details during a live broadcast of SpaceX’s Q3 2026 earnings call on X on October 20. He stated that the primary facility in Gibbons Creek, Texas, targets volume chip production by the third or fourth quarter of 2028. Musk described this schedule as "highly aspirational," noting that it carries "big error bars." He also mentioned that SpaceX plans to produce its first experimental test chip at a research fabrication facility in Austin, Texas, around the second quarter of 2027.

The project involves a complex web of international suppliers and partners. ASML, the world’s sole commercial provider of extreme ultraviolet lithography systems, is considered the most critical partner. Each of these systems costs around $400 million and is essential for leading-edge chip production. Unofficial sources suggest the first EUV tool was slated for delivery in mid-2027, a timeline that may have been accelerated. Other key suppliers include American firms Applied Materials, Lam Research, and KLA Corporation, as well as Japan’s Tokyo Electron, which provide front-end processing and metrology equipment.

Intel joined the Terafab venture in April 2026 to help co-develop and refactor silicon fabrication technology. The facility will utilize Intel’s 14A manufacturing node technology, which relies on traditional extreme ultraviolet lithography. Additionally, Tesla and SpaceX are in discussions with approximately 10 South Korean equipment manufacturers. These include Hanmi Semiconductor, known for advanced packaging, and HPSP, which produces high-pressure hydrogen annealing equipment, along with firms specializing in piping and process valves.

The chips produced at Terafab are intended to power several of Musk’s other technological initiatives. The facility aims to supply processors for xAI’s computing clusters, Tesla’s self-driving vehicles, the Optimus humanoid robots, and SpaceX’s planned orbital AI data center. By manufacturing in-house, the companies hope to mitigate future shortfalls of cutting-edge AI processors that typically come from external foundries like TSMC. This vertical integration is designed to secure a stable supply chain for the broader AI industry.

Beyond the semiconductor work, Musk outlined near-term ambitions for his AI infrastructure during the same broadcast. He expects his AI companies to have 2.5 to 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. To achieve this, SpaceX and xAI are aggressively procuring the next-generation NVIDIA "Vera Rubin" architecture, with plans to bring a cluster of 440,000 GPUs online by the end of the second quarter of 2027.

Despite the detailed planning, many experts have questioned the feasibility of the 1 terawatt ambition. While global AI computing capacity has grown by roughly 3.3 to 3.4 times per year since 2022, according to EpochAI, the leap required for Terafab remains substantial. Other major projects, such as the 10 gigawatt Project Stargate and Meta’s 5 gigawatt Hyperion, are not expected to reach full capacity until the 2030s. Terafab’s 2028 target for volume production is therefore significantly more aggressive than its peers, relying on expedited shipments and rapid scaling of a new manufacturing site.