GlobalFoundries to Make Key AI Chip Component for TSMC in New York

GlobalFoundries is set to manufacture a critical component used in advanced AI chips for Taiwan Semiconductor Manufacturing Co. (TSMC), creating a new U.S.-based supply source for one of the semiconductor industry's biggest packaging bottlenecks.

Under a five-year manufacturing agreement announced on October 8, GlobalFoundries will expand production at its Malta, New York facility to make silicon interposers for TSMC. Volume production is expected to ramp up in the first half of 2028.

The agreement is significant because the AI semiconductor race is no longer only about making faster processors. The ability to connect processors, high-bandwidth memory and other components inside increasingly complex packages has become a major constraint.

For GlobalFoundries, the deal could strengthen its position in AI infrastructure without requiring the company to compete directly with TSMC at the most advanced logic nodes.

What GlobalFoundries will make for TSMC

The key component is a silicon interposer, a thin layer that sits between the chip package substrate and multiple semiconductor dies.

In advanced AI systems, interposers provide high-density connections between processors and memory, allowing large amounts of data to move quickly between components. GlobalFoundries itself describes interposer technology as a way to connect multiple dies and support high-bandwidth, low-latency system integration. 

This is particularly important for AI accelerators that rely on high-bandwidth memory, or HBM.

Instead of treating an AI processor as one large piece of silicon, advanced packaging allows several components to be brought together in a single package. This approach can improve bandwidth and performance while allowing different parts of the system to use different manufacturing technologies.

That makes the interposer a crucial piece of the AI chip supply chain.

Why the TSMC deal matters

TSMC is already the dominant supplier of advanced semiconductor manufacturing and packaging for high-performance computing.

Its CoWoS advanced packaging technology combines processors and HBM memory in sophisticated packages used for AI and high-performance computing. TSMC's 3DFabric platform includes CoWoS technologies designed to integrate multiple system-on-chip dies and HBM stacks. 

But demand for advanced AI packaging has grown so quickly that packaging capacity itself has become a bottleneck.

Counterpoint Research estimates that TSMC's CoWoS supply-demand gap remained around 20% in 2026, although it expects the gap to narrow as TSMC expands capacity. The research firm also says advanced packaging is becoming an important competitive battleground as AI processors become larger and more complicated. 

That puts the GlobalFoundries agreement into perspective.

TSMC is not outsourcing its core leading-edge processor manufacturing to GlobalFoundries. Instead, GlobalFoundries is being positioned as an additional manufacturing source for a specialized part of the advanced packaging ecosystem.

First U.S. source of silicon interposers for this application

GlobalFoundries said its Malta facility is expected to become the first U.S.-based source of silicon interposers supporting these advanced packaging technologies. The agreement also creates a framework for future capacity expansion as demand develops. 

The location matters from a supply-chain perspective.

Much of the semiconductor industry's most advanced manufacturing and packaging capacity is concentrated in Asia. Building additional capabilities in the United States gives customers another geographic source at a time when governments and chip companies are placing greater emphasis on supply-chain resilience.

For TSMC, a U.S.-based interposer source can complement its expanding American manufacturing footprint.

For GlobalFoundries, meanwhile, it provides another opportunity to participate in the rapidly growing AI infrastructure market.

GlobalFoundries is targeting the AI infrastructure bottleneck

GlobalFoundries has been increasingly positioning itself around technologies that support AI systems rather than trying to replicate TSMC's leading-edge logic manufacturing model.

The company announced in July that it had entered a letter of intent for a potential $300 million U.S. Commerce Department CHIPS R&D award focused on silicon photonics, advanced optical materials, wafer technologies and advanced packaging. The company specifically linked these technologies to AI and high-performance computing data centers. 

GF also introduced its SCALE co-packaged optics platform in May, targeting higher-bandwidth connectivity for AI systems.

The strategy is becoming clearer: GlobalFoundries is trying to capture parts of the AI infrastructure chain where specialized manufacturing capabilities can be valuable even without competing with TSMC on 2nm or future leading-edge nodes.

Why advanced packaging is becoming so important

For years, semiconductor competition was largely defined by process nodes — moving from 7nm to 5nm, 3nm and 2nm.

AI is changing that equation.

Modern AI accelerators require enormous amounts of computing power and memory bandwidth. Simply shrinking the transistor size is no longer enough to solve every performance problem.

Advanced packaging allows manufacturers to bring computing, memory and other components closer together.

That reduces the distance data has to travel and can improve bandwidth and power efficiency.

GlobalFoundries' advanced packaging portfolio includes 2.5D and 3D integration, wafer-to-wafer and die-to-wafer bonding, silicon interposers and through-silicon vias. 

The result is a growing market opportunity around the "supporting infrastructure" of AI chips.

What the deal means for GlobalFoundries stock

GlobalFoundries shares rose about 4% in premarket trading following the announcement, according to Reuters-sourced reporting. 

The immediate market reaction reflects the possibility that the agreement could provide longer-term visibility into demand for the company's advanced packaging capabilities.

However, investors should not assume the deal will immediately transform GlobalFoundries' financial results.

Volume production is not expected to ramp until the first half of 2028. The financial contribution will therefore depend on how quickly capacity is built, customer demand develops and production scales.

The agreement is better viewed as a strategic positioning win than as an immediate earnings catalyst.

What it means for TSMC and the AI chip supply chain

For TSMC, additional interposer capacity could help address one of the constraints limiting the expansion of advanced AI packaging.

That matters because demand for AI chips remains extremely strong.

TSMC reported record third-quarter 2026 revenue of NT$1.49 trillion, or about $46.7 billion, up 50% year over year, with AI-related demand a major driver. 

As AI chip demand grows, the industry needs more than additional wafer capacity. It needs memory, packaging, testing, interconnects and other supporting components to expand at a similar pace.

The GlobalFoundries agreement therefore illustrates an important shift in the semiconductor industry: the AI supply chain is becoming broader and more specialized.

What investors should watch next

The most important developments will be the pace of GlobalFoundries' capacity expansion at Malta, customer demand for advanced packaging and the timing of the interposer production ramp.

Investors should also watch TSMC's CoWoS expansion and whether advanced packaging remains a bottleneck through 2027 and into 2028.

For GlobalFoundries, the bigger question is whether its investments in interposers, silicon photonics, advanced packaging and other AI-related technologies translate into sustained revenue growth and stronger margins.

The company is increasingly building an AI infrastructure story around technologies that sit alongside the processor rather than inside it.

The bottom line

GlobalFoundries will manufacture silicon interposers for TSMC under a five-year agreement, with volume production expected to ramp in the first half of 2028. The deal gives TSMC another U.S.-based source for an increasingly important component while giving GlobalFoundries a stronger position in the AI packaging ecosystem.

The key takeaway for investors is that the AI semiconductor opportunity is expanding beyond GPUs and advanced processors. Packaging, memory connectivity, optical links and power technologies are becoming equally important pieces of the AI infrastructure build-out.

Follow the blog for more semiconductor, technology, stock-market and global business updates.

This article is for informational and educational purposes only and should not be considered investment advice.