Subscribe to Our Newsletter

Success! Now Check Your Email

To complete Subscribe, click the confirmation link in your inbox. If it doesn’t arrive within 3 minutes, check your spam folder.

Ok, Thanks

Intel joins Musk's Terafab project to build one terawatt of AI compute capacity

The chipmaker's role raises questions given its struggles to scale advanced manufacturing and find external customers

Ian Lyall profile image
by Ian Lyall
Intel joins Musk's Terafab project to build one terawatt of AI compute capacity
Photo by Slejven Djurakovic / Unsplash

Intel, the American chipmaker, is joining Elon Musk's Terafab project, an ambitious scheme to modernise chip manufacturing and build one terawatt of AI compute capacity, a target many multiples greater than the combined leading-edge output currently achievable by Taiwan Semiconductor Manufacturing Company (TSMC) and Samsung.

The project is expected to require trillions of dollars in capital spending over an extended period, though it has not yet been incorporated into Tesla's full-year capital expenditure forecasts, reflecting the considerable uncertainty that remains around how and when the target will be achieved.

Analysts have begun estimating the scale of investment required to reach the desired capacity, but the pathway to one terawatt of compute remains unclear, and the project is still in early stages.

Intel's involvement follows a visit by Musk to the company's Santa Clara campus, after which the chipmaker was brought into the joint venture alongside Tesla and other undisclosed partners.

The rationale for including Intel centres on its experience scaling semiconductor output across multiple process nodes, a capability that the Terafab project will need if it is to move beyond ambition and into production.

However, serious questions surround whether Intel's expertise will prove sufficient or relevant at the scale the project demands.

Intel's recent history in advanced manufacturing has been marked by difficulties in scaling its leading-edge processes and an inability to attract meaningful volumes of external customers to its foundry business, which has positioned itself as a rival to TSMC.

There is currently no established evidence that Intel's next-generation nodes have achieved high-volume throughput, a prerequisite for the kind of contribution the Terafab project would require.

The economics of the project also point away from Intel as a primary financial beneficiary.

A chip moving through a fabrication facility passes through dozens of processes involving equipment from specialist makers, including Lam Research, KLA, Applied Materials, and ASML, the Dutch company that makes the extreme ultraviolet lithography machines needed for cutting-edge production.

If Terafab proceeds, the bulk of the capital spending is expected to flow toward those equipment suppliers rather than to Intel, whose role is likely to remain advisory and operational rather than a direct recipient of large contracts.

The project reflects a broader pattern in the AI industry, where demand for compute is outpacing the infrastructure available to supply it, and where technology leaders are increasingly looking to vertically integrate or commission dedicated capacity rather than rely on the existing foundry ecosystem.

Whether Terafab represents a credible industrial programme or an aspirational target that will take decades to approach remains the central question for analysts tracking it.

For Intel, participation offers an opportunity to demonstrate relevance at a moment when its foundry ambitions have struggled to gain traction with external customers, but the company's ability to translate manufacturing experience into tangible results for a project of this scale and novelty has yet to be tested.

Ian Lyall profile image
by Ian Lyall