In short: Intel has signed on as the primary foundry partner for Elon Musk’s Terafab, a $25 billion joint venture between Tesla, SpaceX and xAI that aims to produce one terawatt of artificial intelligence computing a year, handing the struggling chip giant the customer it was looking for after moving to a foundry-first strategy.
On April 7, 2026, Intel announced that it had joined the Terafab project, becoming the foundry partner for the most ambitious semiconductor device ever proposed in the United States. The announcement comes two weeks after Musk first unveiled Terafab at Giga Texas’ North Campus in Austin, a joint venture between Tesla, SpaceX and xAI that claims to produce one terawatt of AI computing annually. Intel’s role is to contribute its most advanced process node, packaging expertise and manufacturing scale to realize this claim. For Intel Chief Executive Lip-Bu Tan, who spent the past year trying to rebuild Intel around its foreign foundry business, the deal is the company’s most significant foreign customer win since he took over.
What Terafab claims to build
Terafab is designed as a vertically integrated semiconductor complex that includes chip design, lithography, manufacturing, memory fabrication, advanced packaging and testing under one roof with the goal of producing 100 billion to 200 billion individual AI and memory chips per year. Initial construction targets 100,000 wafer starters per month, with ambitions to eventually reach one million wafer starters per month at full capacity. The project spans two separate facilities on the Giga Texas campus: one dedicated to chips for automotive and humanoid robotics applications, including Tesla’s Self-Driving system, the Cybercab robotaxi program, and the Optimus robot line; and the second for high-performance AI data center infrastructure and dedicated processors for orbital deployments.
That orbital component is central to the project rationale. SpaceX, which completed its acquisition of xAI in February 2026, is creating a constellation of space-based artificial intelligence satellites internally designated AI Sat Mini, creating a combined entity valued at approximately $1.25 trillion. Musk said 80% of Terafab’s computing output will be focused on that orbital infrastructure and the remaining 20% on ground applications. The full cost of the project has been put at between $20 billion and $25 billion, although independent analysts are highly skeptical that this figure will be remotely sufficient to meet stated production targets. A note from Bernstein Research estimated the actual capital required to hit one terawatt of annual computing at about $5 trillion, more than 70% of the total annual US federal budget.
What is Intel’s role and the value of the contract?
Intel will contribute the 18A process node, the company’s most advanced logic manufacturing technology, which is currently expanding into high-volume manufacturing at Intel’s manufacturing facilities in Arizona and Oregon. Intel’s 18A node is a 1.8-nanometer class node, placing it on par with the most advanced processes currently in commercial production globally, and it represents the most sophisticated semiconductor capability produced entirely within the United States. Intel’s statement on joining Terafab was direct: “Intel is proud to join the Terafab project with SpaceX, xAI and Tesla to help reinvent silicon fab technology.” The company added: “Our ability to design, fabricate and package ultra-high-performance chips at scale will help accelerate Terafab’s goal of producing 1 TW of computing per year to power future advances in artificial intelligence and robotics.”
Tan’s post on X was more personal in its framing. “Elon has a proven track record of reimagining entire industries,” he said. “This is exactly what is required in semiconductor manufacturing today. Terafab represents a step change in how silicon logic, memory and packaging will be built in the future. Intel is proud to be a partner.” After the announcement, Intel shares rose nearly 4% to close at $52.91. The market reaction reflects how important the deal is to Intel’s foundry ambitions: In the most recent full year, Intel Foundry generated just $307 million in revenue from foreign customers, rivaling Taiwan Semiconductor Manufacturing Co., which also brings in tens of billions of dollars annually from foreign customers. Terafab, if even partially realized, would completely change the commercial profile of Intel Foundry.
Intel recovery and what is required for this bet
Tan inherited Intel during a severe crisis. The company had lost out to TSMC and AMD in nearly every major product category, its manufacturing roadmap had slipped repeatedly, and its foundry business, an effort to produce chips for foreign customers like TSMC, attracted little interest outside of US CHIPS and government-backed contracts under the Science Act. Tan’s restructuring has been aggressive: thousands of layoffs, a sharper focus on Intel’s 18A and 14A process nodes as foundry cores, and a deliberate effort to position Intel’s domestic manufacturing capabilities as a geopolitical differentiator at a time when U.S. policymakers are intensely focused on chipmaking again.
Terafab is the clearest expression of the location of this pitch. The CHIPS Act, the Trump administration’s desire to see advanced semiconductor manufacturing in the United States, and Musk’s companies’ specific demand for high-volume US-made chips on the leading edge, all converge in this partnership. Whether Intel’s 18A can provide the productivity and volumes that Terafab’s targets require is a separate question. The node has been in development for several years and is only now entering volume; the gap between a controlled high-volume production ramp and the scale of production that Terafab envisions remains enormous. Chipmakers building the world’s largest foundries require years of construction and billions of dollars before the first wafer is processed. The scale of capital commitments that now characterize AI infrastructure investments It gives some context as to what strict execution would actually be required on the targets Terafab claims.
Terafab’s reliability problem has not been solved
The skepticism around Terafab is structural, not just financial. According to Tom’s Hardware’s analysis of Bernstein’s research, building a 2-nanometer-class manufacturing facility capable of cranking out 100,000 wafers per month alone costs about $25 billion to $35 billion, meaning the entire Terafab budget is enough to build a factory that runs at a fraction of what is claimed. Dozens of such objects are needed to reach the starting point of one million waffles per month. The figure of $20-25 billion represents the initial construction capital for the first phase, not the cost of the stated ambitions.
There is also a question about companies on the table. SpaceX-xAI’s internal situation has been turbulent: 11 of xAI’s original co-founders have now left the company Since the acquisition of SpaceX, the rate of attrition has raised questions about the technical sustainability of the organization. Musk’s companies have a documented history of announcing timelines for facilities and products that stretch years later. Tesla’s Cybertruck, Optimus and Full Self-Driving have each missed several commitment dates without affecting the company’s willingness to make new commitments. None of this disqualifies Terafab, Musk’s companies have also achieved goals that have been widely dismissed, notably SpaceX’s orbital launch program, but that explains why analysts don’t take the one-terawatt title at face value.
What the partnership means for the chip industry
Intel’s arrival at Terafab comes as the chip industry undergoes a broader restructuring of who does what and for whom. The rise of dedicated AI silicon, Amazon’s Trainium, Google’s TPUs, Microsoft’s Maia, are eating up the share of AI workloads running on Nvidia hardware. Nvidia’s response has been to open up the NVLink Fusion connection to third-party silicon, including Marvell’s proprietary AI accelerators.a strategy designed to keep custom chip buyers in the Nvidia ecosystem even as they move away from pure Nvidia hardware. Terafab represents something different: a vertically integrated effort to produce custom silicon on a scale that has no precedent outside of the established foundry giants. If the project proceeds anywhere near its stated ambitions, it will add a third major domestic US semiconductor manufacturing ecosystem to a landscape currently dominated by TSMC’s Arizona expansion and Samsung’s Texas operations.
The strategic logic for Intel is clear. Hyperscalers and technology companies are increasingly testing non-Nvidia chips for AI training and inferenceThe market for localized, advanced manufacturer foundry services is growing just as Intel is positioned to serve it. Whether Terafab is finally the vehicle to validate that position, or just another ambitious announcement that tests the distance between Musk’s predictions and physical reality, will be clearer as construction begins and waffle starts count. Capital flowing into AI infrastructure at this scale There’s a way to turn improbable timelines into achievable ones, and Intel will benefit if it’s the first time in years.
