GlobalFoundries agrees $2 billion TSMC interposer deal, with ramp expected in 2028

GlobalFoundries plans to expand its Malta, New York, factory to supply components linking AI processors and memory. Volume production is expected to begin ramping in the first half of 2028.

GlobalFoundries’ semiconductor manufacturing facility in Malta, New York
Context photograph of GlobalFoundries’ Malta, New York facility; date not stated. NIST identifies the site in its GlobalFoundries coverage. GlobalFoundries (resized and converted to WebP). Free use for editorial articles mentioning NIST in accompanying text or a caption; stock art use excluded and may require permission.
LinkedInPostEmail
Save for later

GlobalFoundries said on October 8 it had reached a $2 billion agreement to manufacture silicon interposers for TSMC at its Malta, New York, facility, Reuters reported. The components connect processors and memory inside advanced AI-chip packages, addressing a part of chip production where demand exceeds available manufacturing capacity.

The five-year agreement calls for GlobalFoundries to expand manufacturing capacity in Malta, with volume production expected to begin ramping in the first half of 2028. That timetable makes the announcement a plan for future supply; it does not establish that the additional capacity is already producing components.

According to Reuters’ report, GlobalFoundries expects the facility to become the first US-based source of silicon interposers used in advanced chip-packaging technologies. That remains the company’s expectation, alongside its projected production schedule.

What the $2 billion agreement covers

Reuters reported that the deal also provides a framework for further capacity expansion as demand for advanced packaging grows. The report does not quantify the initial increase in wafer capacity or interposer output, or specify how much additional expansion could follow.

The $2 billion figure is the reported value of the agreement. Reuters’ account does not identify that sum as factory construction spending, a payment already made or revenue already earned. It also does not disclose employment effects or allocations to particular chip customers.

GlobalFoundries shares were up about 4 percent in premarket trading, Reuters reported. That figure describes an early trading observation on the announcement day, rather than a closing share-price return.

Why silicon interposers matter to AI chips

A silicon interposer sits beneath processors and memory chips within an advanced AI package and allows them to communicate at high speeds. Reuters described advanced packaging as a major constraint on AI-chip production because demand for the technology exceeds manufacturing capacity.

TSMC’s technology documentation explains the role of interposers in its CoWoS advanced-packaging process, which combines similar or different types of chips for high-performance computing. Its description places packaging and chip-to-chip connections alongside wafer technology as parts of its computing platform.

The company lists three CoWoS technologies—CoWoS-S, CoWoS-L and CoWoS-R—with features tailored to different applications. Its general technology page does not identify which variant the Malta agreement will serve, so it cannot establish the specific packaging configuration for the new supply.

TSMC says it plans to increase interposer sizes to accommodate more advanced chips and high-bandwidth memory, aiming to increase computing power and bandwidth. It identifies cloud computing, data centres and high-end servers as applications for its advanced stacking and packaging technologies.

AI demand is also reshaping mature-node capacity

The agreement follows a broader shift in manufacturing demand described by TrendForce in research published on June 30. The research firm said AI servers, conventional servers and edge AI devices were drawing a growing share of foundry wafer capacity towards AI-related products.

TrendForce specifically identified rising wafer consumption for silicon interposers using 65- and 55-nanometre processes within the 12-inch mature-node market. Those process figures describe the wider industry in its research; they do not identify the manufacturing process selected for GlobalFoundries’ Malta agreement.

The firm said foundries were shifting limited capacity towards higher-margin AI applications, including interposers and power-management components, while scaling back lower-margin products. Its findings show that the manufacturing demands associated with AI extend beyond production of the processors themselves.

TrendForce forecast that silicon bridges, interposers and other emerging applications would support mature-node capacity utilisation into 2027. That was a June forecast, rather than an October measurement of available capacity or a quantified assessment of the newly announced GlobalFoundries agreement.

Production timing and output remain key unknowns

The concrete production milestone in Reuters’ account is the expected start of the volume ramp in the first half of 2028. The report does not give an eventual output target or establish how much the Malta expansion would reduce the packaging constraint it describes.

For now, the reported commitment is a five-year manufacturing agreement and a planned expansion at a named US facility. Production delivery and its eventual effect on AI-chip availability remain unproven; the announced timetable and first-US-source description are expectations attributed to GlobalFoundries.

Sources and context

AI-assisted article checked against the listed sources. NewsJaws did not conduct interviews or attend the reported events.

About NewsJaws Desk

AI-assisted reporting and explainers reviewed against the linked source documents. No claim of on-scene reporting or original interviews.