Tesla is weighing the construction of its own semiconductor fabrication plant, a move that would mark one of the most capital-intensive bets in the automaker’s history. CEO Elon Musk has spoken publicly about the need for a “gigantic chip fab” that could eventually reach a scale rivaling the world’s largest chipmakers. The ambition arrives even as Tesla locks in a multibillion-dollar chip supply deal with Samsung, raising a sharp question: why would a company that just secured outside production capacity spend billions to replicate it in-house?
Musk’s Vision for a “TeraFab”
At Tesla’s shareholder meeting in 2025, Musk told investors the company needs what he described as a “gigantic chip fab,” according to a Reuters report carried by Yahoo Finance. The initial target, per that reporting, would be around 100,000 wafer starts per month, with a longer-term goal of scaling to 1,000,000 wafer starts per month. For context, a facility producing one million wafer starts monthly would place Tesla in the same production tier as the world’s largest contract chipmakers and demand capital spending on par with leading-edge semiconductor plants.
Musk later expanded on the idea. According to Bloomberg coverage from January 2026, he said Tesla needs to build and operate a “TeraFab” to manufacture semiconductors, including logic, memory, and packaging, domestically. The branding echoes Tesla’s existing naming convention for its vehicle assembly plants, called Gigafactories, but signals a leap into an entirely different industrial domain. Chip fabrication requires ultra-clean environments, extreme ultraviolet lithography tools, sophisticated metrology, and years of process tuning before yields become commercially viable.
One timeline note: Musk’s chip-fab comments were first reported in connection with the 2025 shareholder meeting, while the “TeraFab” language surfaced in Bloomberg’s January 2026 reporting. Whether these represent a single evolving plan or distinct announcements is not entirely clear from available accounts, and Tesla has not published an SEC filing or detailed investor presentation laying out the project’s scope, location, or budget. For now, the TeraFab remains an ambition articulated by Musk rather than a formally approved capital project with disclosed milestones.
The $16.5 Billion Samsung Contract
Tesla’s fab ambitions exist alongside a massive external supply agreement. Samsung Electronics disclosed a foundry contract valued at approximately $16.54 billion, equivalent to KRW 22.76 trillion, with a “large global company.” The contract runs through 2033 and covers the production of advanced chips at Samsung’s foundry business. Samsung’s disclosure did not name the counterparty, but Bloomberg reporting and Musk’s own confirmation identified Tesla as the customer in that $16.5 billion deal.
That figure alone tells a story about how much silicon Tesla expects to consume over the next decade. The company’s appetite for chips spans its electric vehicles, its Full Self-Driving compute hardware, its Dojo training supercomputer, and the Optimus humanoid robot program. Each of those product lines demands custom or semi-custom silicon, and demand is growing as Tesla pushes deeper into autonomous driving, on-board AI inference, and robotics. A long-term commitment at this scale suggests Tesla anticipates not just steady growth, but a substantial ramp in AI-related workloads.
So the Samsung deal is not a small hedge. It is a long-duration, multibillion-dollar commitment that locks Tesla into external foundry capacity through the early 2030s. That makes the simultaneous push to build an in-house fab all the more striking, because it implies Tesla wants both guaranteed access to outside capacity and the option to manufacture critical chips itself.
Why Build When You Can Buy?
The standard playbook for companies that design their own chips, like Apple, Qualcomm, or Nvidia, is to outsource manufacturing to dedicated foundries such as TSMC or Samsung. Designing chips is expensive. Manufacturing them is far more so. A single advanced-node fab can cost $20 billion or more to build and equip, and the learning curve to reach competitive yields takes years. Even for companies with deep pockets, the risk of missing process targets or lagging behind leading foundries is significant.
Tesla’s logic, as best as can be inferred from Musk’s public statements, centers on supply chain control. The global chip shortage of 2021 through 2023 forced automakers to idle assembly lines for months, costing the industry tens of billions in lost revenue. Tesla weathered that crisis better than most legacy automakers, partly by rewriting vehicle software to accommodate available chips and redesigning boards to use alternative components. But the experience appears to have sharpened Musk’s conviction that vertical integration in semiconductors is worth the cost and complexity.
There is also a competitive dimension. Tesla’s AI training and inference workloads are growing rapidly as it develops autonomous driving systems and trains neural networks on vast amounts of driving data. If the company’s robotaxi fleet and humanoid robots scale as planned, each vehicle and robot will need onboard compute hardware refreshed on a regular cycle. Controlling fab capacity would let Tesla iterate on chip designs faster, potentially co-optimizing hardware and software without waiting in line behind other foundry customers or competing for scarce advanced-node slots.
Beyond speed, in-house manufacturing could give Tesla more control over security, intellectual property, and long-term cost structure. Owning a fab would not eliminate the need for external partners, but it could allow Tesla to reserve its own lines for the most strategically sensitive or performance-critical chips, while outsourcing less differentiated components.
The Intel Connection and Open Questions
Adding another layer, Reuters reporting indicates that Tesla has explored a potential tie-up with Intel’s foundry services. Intel has been aggressively courting outside customers for its manufacturing plants as part of a broader turnaround strategy, seeking to position itself as a major contract manufacturer alongside Asian rivals. A partnership could give Tesla access to existing fab infrastructure and process technology without bearing the full burden of building from scratch.
Such an arrangement might take several forms: Tesla could become a large anchor customer for new Intel capacity, collaborate on custom process nodes tuned for its AI workloads, or even co-invest in specific facilities. However, the terms and status of any discussions have not been publicly confirmed by either company, and there is no indication yet that a binding agreement has been signed. Until more details emerge, Intel remains one of several possible paths Tesla could take to secure advanced manufacturing.
This raises a practical question about the Samsung contract’s future. If Tesla begins producing a meaningful share of its chips internally by the late 2020s, the volume commitments in the Samsung deal could become a financial burden rather than a strategic asset. No public statements from Samsung address how in-house Tesla production might affect the agreement, and the contract’s confidentiality provisions limit visibility into renegotiation clauses or flexibility around product mix.
The most likely outcome, based on how other tech companies have managed similar transitions, is a hybrid model. Tesla could continue to rely on Samsung for a baseline of production, particularly for mature designs and high-volume automotive components, while gradually shifting its most cutting-edge AI accelerators or safety-critical chips to either its own TeraFab or a partner like Intel. That would mirror patterns in other industries where companies diversify manufacturing across multiple suppliers and, where feasible, bring a subset of production in-house.
Risks, Rewards, and Strategic Trade-offs
Even if the strategic rationale is clear, the risks are substantial. Building a TeraFab would demand not only enormous capital spending but also a sustained effort to recruit semiconductor process engineers, equipment specialists, and operations staff in a highly competitive labor market. Tesla would be entering a field dominated by companies whose entire business is chip manufacturing, and which have spent decades refining their processes.
There is also execution risk on timing. If Tesla invests heavily in a fab that comes online years after leading foundries move to more advanced nodes, it could end up with expensive but second-tier capacity. Conversely, moving too aggressively could strain Tesla’s balance sheet or divert management attention from core automotive and energy businesses. Balancing the TeraFab vision with ongoing commitments to vehicle production, energy storage, and software development will test the company’s ability to manage multiple large-scale initiatives at once.
On the reward side, success would give Tesla an unusual combination of capabilities: a vertically integrated stack from AI training supercomputers to custom inference chips, manufactured in facilities it controls, and deployed into vehicles and robots it designs. That could reinforce Musk’s broader narrative of Tesla as an AI and robotics company as much as an automaker, and potentially justify the kind of long-term investment horizons usually reserved for leading semiconductor and cloud-computing firms.
For now, however, the TeraFab remains an ambitious concept rather than a fully fleshed-out project. The confirmed facts are Musk’s public comments about needing a gigantic fab, the long-term Samsung contract that secures external supply, and the reported discussions with Intel about leveraging its foundry services. How those pieces ultimately fit together will determine whether Tesla becomes a new kind of hybrid between an automaker, an AI company, and a chip manufacturer, or whether it continues to rely primarily on partners while keeping fabrication as a strategic option rather than a core competency.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
