GlobalFoundries to make AI chip link component for TSMC in New York
Rising demand for advanced AI packaging is pushing chipmakers to add capacity for components that connect processors and memory at high speed. GlobalFoundries said on Thursday it will manufacture silicon interposers for TSMC at its Malta, New York, facility under a five-year agreement.
Highlights
- GlobalFoundries will expand its Malta, New York plant to produce silicon interposers for TSMC, targeting advanced AI chip-packaging applications.
- The Malta site will become the first U.S. source for silicon interposers, with volume production ramping up in the first half of 2028.
- GlobalFoundries shares rose about 4% in premarket trading after the announcement, reflecting investor optimism about AI packaging capacity expansion.
New York capacity expansion plan
As reported by Reuters, citing GlobalFoundries, the company will expand manufacturing capacity at its Malta, New York, plant to produce silicon interposers for TSMC. The component sits beneath processors and memory chips inside advanced AI packages and helps them communicate at high speeds.The company said the Malta site is expected to become the first U.S.-based source of silicon interposers used in advanced chip-packaging technologies. Under the agreement, volume production is expected to begin ramping up in the first half of 2028.
AI packaging bottleneck and market impact
Advanced packaging has become a major constraint on AI chip production as demand for the technology exceeds available manufacturing capacity. The deal also creates a framework for future capacity expansion as customers seek more supply for AI-related chip packaging.Investors react positively to the announcement, with GlobalFoundries shares up about 4% in premarket trading.
In our earlier article on AI data-center buildouts, we highlighted how the race to add capacity is being slowed by equipment shortages, labor constraints, permitting delays, and grid-connection timelines. We also noted that limited access to reliable power is becoming a central bottleneck, raising the risk that AI companies may not be able to scale server-farm infrastructure as fast as demand grows.