1 Terawatt: The Massive Chip Output Musk Promises by 2028

Elon Musk is betting everything on a single Texas facility to reshape the global chip supply chain.

The joint venture between Tesla and SpaceX, known as Terafab, targets a staggering annual output of 1 terawatt of AI computing power. Musk outlined these goals during a live broadcast of SpaceX's Q3 2026 earnings call on October 20. He stated that the project aims for volume chip production at its primary site in Gibbons Creek, Texas, by the third or fourth quarter of 2028. While ambitious, Musk described the timeline as highly aspirational, noting it carries significant uncertainty. He emphasized that the dates are aggressive but supported by accelerated delivery schedules from key partners.

The financial and industrial scale of the project is immense. Once fully operational, the commercial value of Terafab's advanced chip output is projected to rival the GDP of a top-20 country. For perspective, Switzerland currently ranks 20th with an economic output of approximately 1.14 trillion dollars. Musk claimed that a single physical building could generate comparable economic weight. The facility is designed to be vertically integrated, manufacturing logic, memory, and advanced packaging in one location. This approach aims to reduce reliance on external foundries like TSMC, which currently face tight capacity constraints for cutting-edge AI processors.

Musk credited specific suppliers for helping to meet these aggressive deadlines. He named ASML as a critical partner in expediting shipments of extreme ultraviolet lithography systems. These machines are essential for leading-edge chip production and cost roughly 400 million dollars each. ASML is the sole commercial supplier for this specific technology. Unofficial sources previously suggested that Terafab's first EUV tool was slated for delivery in mid-2027. It is not yet clear if this earlier date aligns with the expedited timeline Musk referenced in his recent comments. The CEO of both parent companies expressed appreciation for the speed with which equipment is arriving to support the 2028 target.

The supply chain extends far beyond lithography. American manufacturers Applied Materials, Lam Research, and KLA Corporation are providing front-end processing, metrology, and process control machinery. Tokyo Electron from Japan is also a key partner in this ecosystem. Tesla and SpaceX are currently in discussions with approximately 10 South Korean equipment makers. These include Hanmi Semiconductor, known for advanced packaging, and HPSP, a manufacturer of high-pressure hydrogen annealing equipment. Other firms are being engaged for specialized piping components and process valves. This broad network of collaborations is necessary to handle the massive volume of production planned for the Texas site.

Intel plays a distinct role in the technical architecture of the project. The company officially joined the Terafab venture in April 2026. Its contribution involves co-developing and refactoring silicon fabrication technology to provide access to its 14A manufacturing node. The facility will utilize traditional EUV lithography systems built upon this specific Intel process technology. This partnership is central to the plan of fabricating high-performance chips internally rather than outsourcing to existing foundries. The integration of Intel's node with ASML's equipment forms the technical backbone of the new manufacturing line.

Beyond the chip factory, Musk outlined immediate goals for his AI infrastructure. He expects his AI companies to have between 2.5 and 3 gigawatts of computing capacity active by the end of 2026. For 2027, he projects adding an incremental 10 gigawatts to that base. This would bring the total AI compute to 13 gigawatts by the end of next year. He expressed high confidence in hitting this figure, stating he would be surprised if they did not achieve it. This growth is driven by aggressive procurement of next-generation hardware from NVIDIA. Specifically, plans are in place to bring a training cluster of 440,000 NVIDIA Vera Rubin GPUs online by the end of the second quarter of 2027.

These capacity targets place Musk's projects in a competitive field with other major industry initiatives. Project Stargate, which has been restructured to a 10 gigawatt capacity, is not expected to reach full operation until the 2030s. Meta's Hyperion project, with a 5 gigawatt target, also shares a similar long-term timeline. In contrast, Musk's plan for 13 gigawatts by late 2027 is significantly faster. The chips produced at Terafab will serve multiple internal needs, including xAI clusters, self-driving vehicles, Optimus humanoid robots, and SpaceX's orbital AI data center. The output is also intended to supply the broader AI industry, potentially alleviating current shortages of advanced processors.

Skepticism remains regarding the feasibility of the 1 terawatt annual output goal. Research firm Gartner estimates that the total global AI-optimized data center capacity is roughly 44 gigawatts as of late 2026. This existing global capacity represents only 4.4 percent of Terafab's planned annual production. While EpochAI notes that global AI computing capacity has grown by a factor of 3.3 to 3.4 per year since 2022, the jump required to reach terawatt levels is substantial. SpaceX plans to manufacture its first experimental test chip at its research fabrication facility in Austin, Texas, around the second quarter of 2027. This step is a precursor to the full-scale production targeted for the later part of the decade.