Musk: $3 Billion Terafab Research Facility to Test Intel 14A New Physics
In the latest earnings call, Tesla CEO Elon Musk stated that the Terafab project is still in the stage of refining deployment details. The "research-grade wafer fab" located in the Texas Gigafactory is expected to invest approximately $3 billion to test "some new physics we want to try," with Tesla responsible for construction and validation. The initial phase of scaling up Terafab will be led by SpaceX, creating a division of labor where "Tesla does R&D and SpaceX does large-scale production." Musk also confirmed that Intel has joined Terafab as a core manufacturing technology partner, with plans to use Intel's 14A advanced process in the project and engage in deep collaboration in key areas such as lithography and packaging. He described this as "one of the most advanced processes available today," which is expected to provide high-performance chips for Tesla's autonomous driving, Optimus robots, and SpaceX satellites.
Source: Public Information
ABAB AI Insight
Musk's latest statement on Terafab outlines a three-phase roadmap: Tesla will build a $3 billion "research wafer fab" in Giga Texas for process and architecture experimentation, SpaceX will be responsible for the initial phase of scaling production, and Intel will provide the 14A process and key manufacturing IP. This elevates the original idea of "making in-house chips" to "co-building a dedicated production line for AI and aerospace with a leading U.S. foundry." In this structure, Musk's companies shift from relying on external suppliers (Taiwanese and Korean foundries) to using orders and equity as leverage to deeply bind Intel's most advanced process nodes, securing priority in computing power and time.
From a global chip landscape perspective, this essentially establishes a "semiconductor corridor tailored for specific ecosystems" in the U.S.: design, validation, and some manufacturing capabilities are concentrated in Texas, while production is completed within Intel's 14A/18A process system, aiming to provide integrated chip supply for FSD, Robotaxi, Optimus, and Starlink satellites. This is not only a supply chain security issue but also a computing power sovereignty issue—whoever controls the vertical integration from architecture to process will have the upper hand in pricing and iteration in the next technological cycle.
The phrase "some new physics to test" also indicates that Terafab is not merely replicating existing 3nm/2nm routes but is innovating in dimensions such as power consumption, heat dissipation, packaging, and radiation protection. Optimus, Cybercab, and Starlink satellites have extreme requirements for energy efficiency, size, and reliability, especially regarding radiation and temperature fluctuations in orbital environments, which impose constraints on processes and packaging that differ from those of smartphones and data centers. Validating "new physics" in the research facility before SpaceX leads large-scale production essentially encapsulates R&D risks within small-scale capital expenditures, bringing mature solutions into large-scale manufacturing through aerospace business, which in turn feeds back into Tesla and xAI's ground and cloud computing needs.
In the longer term, Terafab represents an extreme form of "enterprise self-built infrastructure in the era of large models": not satisfied with buying GPUs or customizing ASICs, but directly delving into wafer fabs and process levels, attempting to master cost curves and performance limits at the lowest level. If this model succeeds, Musk's ecosystem's competitiveness in AI, autonomous driving, and space internet will derive not only from software and data but also from its degree of vertical integration across the entire "silicon-electricity-computing power-application" chain; this will also impact traditional foundries and cloud vendors' pricing power, forcing them to rethink the boundaries between "contract manufacturing" and "joint investment" when facing super-large customers.