Elon Musk's partnership with Intel on Terafab chips raises more questions than answers

Elon Musk has announced the creation of Terafab — a chip venture designed to produce specialized AI processors for his companies' needs. Intel has been named as the key partner — and that's exactly where questions begin. The details of the agreement remain murky: it's unclear what exact role Intel is taking on, what the real timelines are, and how realistic the entire structure is given that Intel itself is going through one of the most difficult periods in its history.

Wired broke down five key questions about this partnership. The reasons for Terafab are easy to understand. xAI, Musk's startup in artificial intelligence, has faced acute shortages of computing power.

The massive Colossus cluster in Memphis, which houses more than 100,000 Nvidia GPUs, cost billions of dollars and still barely covers the needs for training Grok models. Nvidia holds a monopoly on the AI accelerator market: prices are high, waiting times stretch quarters into the future, and dependence on a single supplier is a systemic risk for any AI company. Terafab was conceived as an answer to this problem.

Custom chips would allow xAI to reduce operating expenses and follow the path of the industry's largest players: OpenAI builds chips with Broadcom, Google makes its own TPUs, Amazon develops Trainium, Microsoft invests in the Maia series. Custom hardware has become a strategic priority for the entire AI industry. The central question is what exactly Intel does as part of the deal.

Fundamentally different scenarios are possible here. In the minimalist version, Intel acts as a contract manufacturer: Terafab designs chips, Intel produces them on its factories, without touching architectural decisions. In the maximum version, Intel participates in designing the processors themselves, transforming from a contractor into a full-fledged co-author.

The difference is critical — both in terms of intellectual property distribution and risk allocation for both sides. For now, the companies remain silent about the details, which is itself noteworthy: major chip partnerships are usually accompanied by concrete technical announcements and specifications. The lack of specifics raises legitimate doubts about how far negotiations have actually progressed.

Intel is experiencing a structural crisis, which makes a partnership with Musk simultaneously attractive and risky. The company has recorded multi-billion dollar losses, conducted a massive layoff of 15,000 employees, and lost the technological leadership it had maintained for decades. New CEO Lip-Bu Tan is betting on reviving the foundry business — contract manufacturing of chips for third-party customers.

A deal with Musk would be a headline-grabbing anchor contract, capable of convincing the market that Intel Foundry can compete with TSMC. Intel's 18A manufacturing node promises a technological breakthrough, but mass production of complex AI chips with acceptable yield — a task that Intel still needs to prove it can handle at real volumes. Terafab fits into Musk's characteristic strategy of vertical integration: Tesla designs FSD chips for autopilot, SpaceX creates rocket components itself.

The principle is the same — control the critical supply chain and don't depend on competitors. The logic is flawless. But implementation inevitably runs up against time.

Developing a custom AI chip takes 18 to 36 months even for teams with deep experience — and then there's still debugging of production lines and software ecosystem. The AI industry doesn't wait. If Terafab hits the market in three to four years, competitors will have gone through several generations of architectures.

The key question behind this deal is simple: is Intel, while undergoing painful restructuring, capable of keeping pace with and meeting the demands that Musk will inevitably impose?