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

Elon Musk has announced Terafab, a chip venture meant to produce specialized AI processors for his companies. Intel is named as the key partner, and that is where the questions begin. The details of the agreement remain vague: it is unclear what role Intel will actually play, what the real timeline is, and how realistic the whole arrangement is, given that Intel itself is going through one of the hardest periods in its history. Wired broke down five key questions about the partnership.

The rationale for Terafab is easy to grasp. xAI, Musk's artificial intelligence startup, has run into an acute shortage of computing power. The giant Colossus cluster in Memphis, with more than 100,000 Nvidia GPUs, cost billions of dollars and still barely covers the needs of training the Grok models. Nvidia holds a monopoly position in the AI accelerator market: prices are high, queues stretch quarters ahead, and dependence on a single supplier is a systemic risk for any AI company.

Terafab is conceived as an answer to that problem. In-house chips would let xAI cut operating costs and follow the path of the industry's biggest players: OpenAI is building chips with Broadcom, Google makes its own TPUs, Amazon develops Trainium, and Microsoft is investing in the Maia series. Custom hardware has become a strategic priority for the entire AI industry.

The central question is what exactly Intel does under the deal. Fundamentally different scenarios are possible. In the minimal version, Intel is a contract manufacturer: Terafab designs the chips and Intel stamps them out in its fabs, without touching architectural decisions. In the maximal version, Intel takes part in designing the processors themselves, turning from contractor into full co-author. The difference is critical, both for the distribution of intellectual property and for the risks to both sides.

For now, the companies are keeping quiet about the details, which is notable in itself: major chip partnerships are usually accompanied by concrete technical announcements and specifications. The lack of specifics raises legitimate doubts about how far the talks have actually gone.

Intel is in a structural crisis, which makes a partnership with Musk both attractive and risky. The company has posted multibillion-dollar losses, cut 15,000 employees in a large-scale layoff, and lost the technological leadership it held for decades. New CEO Lip-Bu Tan is betting on reviving the foundry business, contract chip manufacturing for outside customers. A deal with Musk would be a high-profile anchor contract, able to convince the market that Intel Foundry can compete with TSMC. Intel's 18A process node promises a technological breakthrough, but mass production of complex AI chips with acceptable yields is something Intel has yet to prove at real volumes.

Terafab fits Musk's characteristic strategy of vertical integration: Tesla designs FSD chips for its autopilot, and SpaceX builds rocket components in-house. The principle is the same: control the critical supply chain and don't depend on competitors. The logic is flawless. But execution inevitably runs into time. Developing a custom AI chip takes 18 to 36 months even for deeply experienced teams, followed by debugging of production lines and the software ecosystem. The AI industry does not wait. If Terafab reaches the market in three to four years, competitors will have gone through several generations of architectures.

The main question behind this deal is simple: can Intel, in the midst of painful restructuring, sustain the pace and meet the demands that Musk will inevitably impose?