Musk’s latest moonshot just became less speculative and more industrial. By bringing Intel into the Terafab project, the plan to build a massive chip factory in Texas jumps from PowerPoint fantasy to something the semiconductor industry has seen before: a classic foundry deal with very demanding anchor customers. For AI, robotics and space infrastructure, this is not a side story – it is about who controls tomorrow’s compute supply.
In this piece, we’ll unpack what Intel actually gains, what Musk really wants, why Nvidia and TSMC should pay attention, and what this all means for Europe’s own chip ambitions.
The news in brief
According to TechCrunch, Intel has agreed to participate in Elon Musk’s Terafab initiative, a proposed U.S. semiconductor manufacturing facility in Texas tied to SpaceX and Tesla. Intel posted on X that its design, fabrication and packaging capabilities will support Terafab’s stated goal of ultimately enabling 1 terawatt per year of compute for future AI and robotics workloads.
Musk first revealed in March that Tesla and SpaceX would collaborate on custom chips for AI compute, satellites and a possible space-based data center, as well as for autonomous vehicles and robots. Until now, it was unclear how two non‑semiconductor companies could realistically build and operate a cutting‑edge fabrication plant, a type of project that typically costs more than $20 billion and takes years to complete.
Intel, which is trying to grow its contract manufacturing (foundry) business and has been seeking major anchor customers, now appears to be the industrial partner behind Terafab. The company did not provide further details beyond the social media post, but its stock rose a few percent on the news.
Why this matters
On the surface, this looks like a win for everyone involved. Musk gets what he craves most: control over the supply of high‑end compute for AI and autonomy, instead of relying on a volatile GPU market dominated by Nvidia and a manufacturing ecosystem centered on TSMC and Samsung. Intel gets something it desperately needs: a marquee, long‑term customer that can justify the billions it is already pouring into new fabs.
But the deeper story is about strategic dependence. AI companies have learned the hard way that innovation is constrained not only by algorithms and data, but also by access to silicon and power. A promise of “1 TW/year of compute” may be marketing shorthand, but the direction is clear: Terafab is positioned as industrial infrastructure for AI and robotics, not just another chip line.
For Intel, attaching itself to Musk’s ecosystem is a calculated risk. Tesla and SpaceX move fast, tolerate high technical risk and accept brutal iteration cycles. Intel, by contrast, has struggled for years with delays and execution issues on advanced process nodes. If it can satisfy the performance and volume demands of two of the most aggressive customers in tech, it sends a strong signal to the rest of the market that Intel Foundry is back in the game.
The losers, at least on paper, are the incumbents who currently own most of Musk’s compute wallet: Nvidia on the chip side and TSMC on the manufacturing side. Even if Terafab takes years to fully materialise, its mere existence strengthens Musk’s hand in negotiations and could nudge other hyperscalers to diversify their silicon strategies.
The bigger picture
This deal is part of a broader trend: hyperscale buyers and platform companies are no longer content to be just customers of the semiconductor industry; they want to shape it. Amazon designs its own Graviton CPUs and Trainium/Inferentia AI accelerators, now attracting customers like Uber, as recently reported. Google has TPUs. Apple vertically integrated its silicon across devices. Meta has its own AI chip efforts. Musk was always going to follow.
Historically, attempts by system companies to build fabs themselves have not ended well – the capital intensity and process complexity are brutal. The more sustainable pattern is what we see emerging here: a deep partnership between a demanding platform company and an established foundry that provides the manufacturing muscle.
For Intel, Terafab fits neatly into its IDM 2.0 strategy announced earlier this decade: keep designing its own CPUs while also becoming a world‑class foundry for others. It has been racing to catch up to TSMC on process technology and has actively courted so‑called “anchor tenants” for its new facilities in the U.S. and Europe. If Tesla and SpaceX become core customers, that helps de‑risk those investments and makes it easier to justify further government subsidies under U.S. industrial policy.
At the industry level, this underscores how AI compute is evolving into a form of critical infrastructure, akin to energy or telecom networks. Reports about orbital data centers and space‑based compute, once dismissed as science fiction, now sit in the same slide decks as power‑hungry data centers in Texas and Arizona. Musk’s space data center concept, mentioned in the TechCrunch piece, bridges those narratives: AI workloads may eventually be distributed across Earth and orbit, but they will still run on silicon made somewhere – and Terafab aims to be that “somewhere.”
The European / regional angle
From a European perspective, the timing is revealing. Brussels has spent the last years drafting the EU Chips Act and financing fabs from Intel in Magdeburg and TSMC in Dresden to reduce dependence on Asian manufacturing. Yet the most aggressive, vertically integrated AI infrastructure plays are still coming from the U.S., and now even the “new” fabs tied to those plays are U.S.‑based.
For European automakers and robotics companies, this should ring alarm bells. Tesla is not just a competitor in EVs; it is building its own stack of compute, data and manufacturing, from self‑driving training clusters to in‑house chips. If Musk secures preferential access to future Terafab capacity, European OEMs risk becoming second‑tier customers in the global queue for advanced AI silicon.
Regulators will also pay attention. The EU’s Digital Markets Act and AI Act are focused on gatekeepers and foundation models, but they largely ignore the underlying compute infrastructure. If control over high‑end chips concentrates in a few U.S.‑centric ecosystems – Nvidia, Intel–Musk, perhaps Amazon – Europe’s sovereignty narrative starts to look thin, even with local fabs on the continent.
On the flip side, the Intel–Terafab alliance could indirectly benefit Europe if it stabilises Intel’s balance sheet and accelerates process development that also flows into its European plants. Slovenian, German or Croatian startups building AI products care less where the wafer was processed and more whether they can get capacity at predictable prices. If this deal makes Intel a more credible alternative to TSMC for European chip designers, that is a net positive.
Looking ahead
Do not expect Terafab to flip on overnight. Leading‑edge fabs typically require several years from groundbreaking to volume production, and that is for companies that have done it many times before. Even if Intel shoulders most of the operational burden, aligning Tesla’s and SpaceX’s custom requirements with Intel’s process roadmaps will be non‑trivial.
The key questions to watch over the next 12–24 months are:
- How exclusive is the relationship? If Terafab capacity is heavily earmarked for Musk’s companies, other AI players may have to look elsewhere – or push for similar co‑investment deals.
- What process node and packaging tech will be used? This will determine whether Terafab can compete with Nvidia‑class accelerators built at TSMC.
- Does this accelerate or sideline Intel’s European fab plans? A big U.S. anchor customer might shift management focus back across the Atlantic.
There are also execution risks. Intel still needs to prove it can hit its process milestones; Musk’s timelines are notoriously optimistic; and macro factors like interest rates, energy prices and U.S.–China tensions could all reshape the economics of the project.
The opportunity, however, is enormous. If Intel executes, it gains a flagship AI account. If Musk executes, he gains a partial escape hatch from Nvidia dependency. If both execute, we get a new center of gravity in the global compute supply chain.
The bottom line
Terafab with Intel onboard is less of a romantic “Tesla builds a fab” story and more of a hard‑nosed industrial alliance: Musk brings insatiable demand and political leverage; Intel brings lithography, clean rooms and yield engineers. Together, they are betting that AI and robotics will justify yet another mega‑fab in Texas.
For Europe, the message is uncomfortable but clear: the race for AI power is moving down the stack to who owns the fabs and the power plants. The real question is whether European industry and policymakers are willing to play at that level – or content to remain premium customers of someone else’s infrastructure.
