What does GlobalFoundries’ $2 billion TSMC deal actually specify?

An interposer is a silicon component used in advanced chip packaging. TSMC describes CoWoS as a way to integrate compute chips with high-bandwidth memory; its CoWoS-S version uses a silicon interposer. The planned component is part of packaging AI hardware, not a promise to manufacture complete AI chips in the United States.

GlobalFoundries expects volume production at Malta to begin ramping in the first half of 2028. That is a forecast, not evidence that the interposers are already qualified or being made at volume. The public announcement gives no output target, minimum TSMC purchase commitment or detailed production schedule. It also does not explain whether the $2 billion represents capital spending, payments or guaranteed purchases.

GlobalFoundries says the Malta site is expected to become the first U.S.-based source of silicon interposers for advanced packaging. That, too, is a future outcome, not a demonstrated one. The public reporting reviewed does not establish that this new supply has reached the market.

Why isn’t the $2 billion figure enough?

The strongest case for treating the agreement as concrete is that it identifies a customer, a manufacturing site, a component and an expected production-ramp period. Those details make it more specific than a general promise to expand semiconductor capacity. A multi-year plan can be a necessary step toward manufacturing output.

But the missing terms are central, not cosmetic. Without disclosed volumes, purchase commitments or capacity figures, readers cannot judge the scale of the planned addition. The $2 billion value cannot answer that question when the public terms do not explain its financial structure.

GlobalFoundries presents the project as a way to create a secure, scalable U.S. source and strengthen the semiconductor supply chain. Those are the company’s goals; the reporting does not establish that the supply chain has become more resilient or that customers will receive a particular quantity of components. The public materials also do not include a separate TSMC statement detailing its commitments.

That is not a reason to dismiss the agreement. It is a reason to distinguish a manufacturing arrangement from factory output. The useful measures will be whether Malta capacity is installed and qualified, when production begins, and how many interposers it makes.

What does the Xanadu partnership show?

A separate GlobalFoundries partnership illustrates why component announcements should be kept distinct. On Tuesday, Oct. 6, Xanadu announced a multi-year collaboration focused initially on industrializing silicon nitride photonics and superconducting nanowire single-photon detectors on GlobalFoundries’ Malta 300 mm production line. The Xanadu partnership concerns components for quantum-computing systems, not TSMC’s AI packaging.

Xanadu says the work could help scale components for its quantum-computing architecture and future demonstrators. The announcement does not establish that the components are in volume production or that the anticipated benefits have been independently verified. This effort may show another attempt to develop specialized manufacturing at Malta; it does not demonstrate the scale or success of the interposer program.

GlobalFoundries and TSMC should make progress against production milestones clear, including when the Malta capacity is qualified and operating. When volume production begins, they should report output in a way that lets customers and the public assess its scale. Until then, the deal is a signed manufacturing arrangement with a forecast—not proof that additional U.S.-made interposers are already supporting AI systems.