Intel sold its soul to Musk

On April 7, 2026, Intel announced it was joining Elon Musk's Terafab project. The company posted a photo of CEO Lip-Bu Tan shaking hands with Musk and declared that its ability to "design, fabricate, and package ultra-high-performance chips at scale" would help Terafab hit its goal of producing one terawatt per year of compute for AI and robotics. Read that again. Intel, the company that invented the x86 processor and dominated computing for four decades, is now pitching itself as a manufacturing partner for Musk's empire. Not a competitor. Not an equal. A supplier. The framing was triumphant. Intel's stock jumped nearly 5%. But strip away the press release language and what you see is a company that has run out of options embracing the one person willing to write a big enough check.

The timeline of decline

Intel's fall didn't happen overnight. It was a slow erosion masked by quarterly earnings calls and roadmap promises. For years, Intel held a near-monopoly on high-performance processors. It designed the chips and manufactured them in its own fabs, a vertically integrated model that gave it enormous advantages in cost, speed, and optimization. Then the world shifted. Mobile ate into PC sales. Cloud providers started designing their own silicon. And the AI revolution, powered by GPUs rather than CPUs, caught Intel flat-footed. Nvidia, once a niche graphics card company, became the most important chipmaker on the planet. By 2025, Nvidia held roughly 80-85% of the AI accelerator market and was generating operating margins above 70%. Intel, meanwhile, was losing money. Its foundry business, the unit responsible for manufacturing chips for external customers, lost $10.3 billion in 2025. External foundry revenue was a rounding error, just $307 million for the full year, with one quarter showing only $8 million from outside clients. When Lip-Bu Tan took over as CEO in March 2025, he inherited a company that had been investing "too much, too soon, without adequate demand," as he put it in an internal memo. The factory footprint was fragmented and underutilized. Two European plants were halted. Twenty percent of the workforce was laid off. The strategy shifted from Pat Gelsinger's ambitious "five nodes in four years" to something more pragmatic: don't build capacity until customers commit to using it. The problem was that the customers weren't coming fast enough.

What Terafab actually is

Musk announced Terafab in March 2026 as a joint venture between Tesla, SpaceX, and xAI. The pitch was audacious: a 100 million-square-foot semiconductor manufacturing complex in Austin, Texas, capable of producing more compute than all the world's chip manufacturers combined. The price tag was estimated between $20 and $25 billion, with Bernstein analysts calculating that the full vision could eventually cost $5 trillion. The stated purpose was to build two types of chips. One for terrestrial applications, powering Tesla's autonomous vehicles, humanoid robots, and xAI's models. The other for space, supporting SpaceX's satellite constellation and its planned orbital data centers. The reaction was skeptical, and reasonably so. Leading-edge chip fabs take a decade and tens of billions of dollars to stand up. TSMC, Samsung, and Intel have spent generations perfecting these processes. Tesla has never made a wafer in its life. As Electrek noted at the time, "The idea that a car company, a rocket company, and an AI startup were going to stand up a competitive sub-2nm process from scratch was, to put it charitably, ambitious." Then Intel walked in, and the skepticism shifted. Because now Terafab has someone who actually knows how to run a fab. The question is no longer whether the project is technically feasible. It's what Intel gave up to be here.

A foundry in search of a customer

Intel's foundry business has been the company's most expensive bet and its most persistent failure. The idea, championed by Gelsinger and continued under Tan, was to transform Intel from a company that only makes its own chips into a contract manufacturer for anyone, competing directly with TSMC. On paper, the logic was sound. The U.S. government wanted domestic chip manufacturing. The CHIPS Act provided $7.86 billion in grants. Geopolitical tension around Taiwan made TSMC's dominance a strategic vulnerability. Intel had the fabs, the engineering talent, and the home-field advantage. In practice, winning foundry customers proved brutally difficult. TSMC's manufacturing processes are years ahead in yield and reliability. Intel's 18A node, its most advanced process, has been plagued by supply constraints. Analysts repeatedly expressed doubt about the turnaround timeline. One Seeking Alpha report called the foundry "a lost cause" and recommended Intel abandon the strategy entirely. There were bright spots. Nvidia invested $5 billion in Intel in September 2025 and began exploring using Intel's foundry for packaging and manufacturing starting around 2028. Microsoft and Amazon signed long-term fabrication contracts for custom chips on the 18A process. But these were future commitments, not current revenue. Intel needed a marquee anchor customer now. Musk was that customer. The Terafab deal gives Intel something it has desperately needed: a large, long-term, external demand signal for its foundry. 247 Wall Street estimated the partnership could push Intel's external foundry revenue into the billions annually. For a business that generated just $307 million from outside clients in all of 2025, that's transformational. But transformational for whom?

Vertical integration on steroids

Step back and look at what Musk now controls, or is building toward controlling. xAI trains large language models. Tesla builds the robots and autonomous vehicles those models power. SpaceX provides the global connectivity infrastructure and is planning orbital data centers. And now, through Terafab with Intel as the foundry partner, Musk is moving into chip fabrication itself. This is vertical integration at a scale that hasn't existed in the tech industry before. The closest historical parallel might be the old-school conglomerates, the zaibatsu of prewar Japan or the chaebol of South Korea. Samsung, for instance, makes everything from memory chips to smartphones to ships. But even Samsung doesn't control the AI models, the robotics platforms, and the launch vehicles that put the data centers in orbit. The strategic logic is clear. Musk's companies collectively consume enormous amounts of compute. Tesla's fifth-generation AI chip, AI5, is designed to power autonomous driving. xAI's Grok models require massive training infrastructure. SpaceX needs radiation-hardened processors for satellites. By controlling the fabrication, Musk can optimize chips for his specific workloads, reduce dependency on external suppliers, and potentially undercut competitors on cost. But there's a difference between building infrastructure to serve your own needs and building infrastructure that gives one person control over the entire stack from silicon to orbit. That distinction matters.

The TSMC question

One generous reading of the Intel-Terafab deal is that it's a hedge against Taiwan risk. TSMC manufactures roughly 90% of the world's most advanced semiconductors. Nearly all of Nvidia's GPUs, Apple's phone processors, and AMD's chips are fabricated in Taiwan. If a conflict in the Taiwan Strait disrupted TSMC's operations, the global technology industry would face a crisis with no easy workaround. The U.S. government has recognized this vulnerability. The CHIPS Act was explicitly designed to incentivize domestic manufacturing. TSMC is building fabs in Arizona, but those facilities won't match the scale or sophistication of its Taiwan operations for years. In this context, Intel partnering with Terafab to build advanced fabs in Austin could be seen as a patriotic act, reducing U.S. dependence on a geopolitically precarious supply chain. Intel's own messaging leans into this framing. Lip-Bu Tan has repeatedly emphasized Intel's role as America's leading-edge manufacturer. But there's a less generous reading. Intel isn't partnering with the U.S. government or a broad consortium of American companies. It's partnering with Musk. If Terafab's output primarily serves Tesla, SpaceX, and xAI, then the diversification benefit is narrow. The U.S. gets domestic chip manufacturing, but the capacity is locked into one person's ecosystem. The question is whether Terafab will eventually open its doors to other customers or remain a captive foundry for Musk's ventures. Intel's history suggests it wants external customers. Musk's history suggests he wants control.

The TurboQuant shadow

Two weeks before Intel's announcement, Google published TurboQuant, a compression algorithm that reportedly shrinks AI memory requirements by sixfold. The paper triggered a $90 billion wipeout in memory chip stocks. Samsung and Micron were hit hardest, but the implications extended across the entire semiconductor supply chain. The significance of TurboQuant isn't just the specific algorithm. It's what it represents: the era of brute-force scaling, throwing more chips at every problem, is giving way to efficiency-driven AI. Google, Anthropic, and other labs are competing not just on model capability but on inference cost. DeepSeek demonstrated this in early 2025 when it showed that dramatically cheaper inference didn't reduce demand for compute but expanded the set of things people were willing to build. For Intel, TurboQuant is a double-edged sword. If AI can be made cheaper to run, total demand might increase (Jevons Paradox). But if the efficiency gains are fast enough, the massive buildout of chip fabrication capacity could become partially redundant before it's even completed. A factory designed to produce a terawatt of compute per year looks different if the industry figures out how to do the same work with a fraction of the silicon. This is the backdrop against which Intel walked into Terafab. The standalone chip business isn't just struggling because of execution failures. The ground itself is shifting. Running toward a guaranteed customer, even one who demands exclusivity, starts to look rational when the alternative is building capacity that nobody wants.

What happens to Intel's other customers?

This is the second-order question that hasn't gotten enough attention. Intel's foundry strategy was built on the premise that it could serve many customers. Microsoft, Amazon, Nvidia, and potentially Apple were all in various stages of discussions or contracts. The foundry's value proposition depended on being a neutral platform, a Switzerland of semiconductor manufacturing. Becoming Musk's foundry partner complicates that positioning. If Intel's best engineers, its most advanced packaging technology, and its newest fab capacity are allocated to Terafab, what's left for everyone else? Nvidia is already exploring Intel's foundry for 2028. Will Nvidia be comfortable sharing manufacturing capacity with a project whose primary beneficiaries, Tesla and xAI, are direct competitors in AI infrastructure? There's also a reputational dimension. Intel's foundry pitch has been: "We're a reliable, independent, American alternative to TSMC." Tying the foundry's future to Musk's ventures introduces a different kind of risk. Musk's companies are polarizing. His political activities are polarizing. For a potential customer evaluating whether to commit billions in fabrication contracts, the question "Do I want my chips made in the same fab as Elon Musk's projects?" is now part of the calculus.

The national security angle

Intel is, in a meaningful sense, a national security asset. It's the only American company capable of manufacturing leading-edge semiconductors at scale. The CHIPS Act invested nearly $8 billion in Intel specifically because of this strategic importance. This creates an uncomfortable dynamic. A company that exists partly because the U.S. government decided domestic chip manufacturing was a matter of national security is now hitching its commercial future to one individual's corporate empire. Musk, through his role as head of the Department of Government Efficiency and his relationships with the current administration, is simultaneously a government actor and a private industrialist. None of this is illegal or even unusual in the abstract. Government contractors have always had close relationships with political figures. But the concentration of influence is striking. Musk's companies receive billions in government contracts (SpaceX), benefit from government incentives (Tesla), and now partner with a company that receives billions in government grants (Intel). The circularity is hard to ignore.

What Intel actually sold

Intel didn't sell its soul. That's the headline, not the reality. What Intel sold is something more mundane but equally significant: its independence. For fifty years, Intel was the company that set the pace for the semiconductor industry. It chose which chips to build, which architectures to pursue, which customers to serve. Even during its years of decline, it maintained the pretense of being a technology leader charting its own course. The Terafab deal is an acknowledgment that those days are over. Intel is now, functionally, a service provider. It will design and fabricate chips to someone else's specifications, for someone else's products, in pursuit of someone else's vision. That someone happens to be the richest person on the planet, building what might be the most vertically integrated technology empire in history. This might be the right strategic move. Intel needed a major foundry customer. Musk needed a foundry. The fit is logical. But let's not pretend it's anything other than what it is: the company that once defined Silicon Valley admitting it can no longer compete at the top of the technology stack and accepting a role as the builder of someone else's future. The chips will still say Intel inside. They just won't be Intel's chips anymore.

References