GlobalFoundries will produce a key component used in AI chips for TSMC, Reuters reports, marking a rare manufacturing tie-up between two of the world's largest contract semiconductor makers.

The arrangement directly links the second- and third-largest foundries by revenue — behind only Samsung's contract business — in a segment where they normally compete head-to-head for customer orders. Neither company has disclosed financial terms, volumes, or a production start date in the initial report.

What does the partnership cover?

According to Reuters, GlobalFoundries will manufacture a component that forms part of AI chips whose primary production sits with TSMC. The world's largest contract chipmaker supplies the overwhelming majority of advanced AI accelerators, including the designs behind the current generation of large-scale machine-learning systems.

The specific component, the process node involved, and the end customers were not detailed in the report. What the deal signals, however, is that TSMC is willing to route part of the AI supply chain through an external foundry rather than keep every element in-house.

GlobalFoundries, headquartered in Malta, New York, exited the leading-edge race years ago. It stopped developing 7nm-and-below logic processes in 2018 and has since concentrated on mature and specialty nodes: radio-frequency silicon, embedded memory, power management, and photonics. Those niches are exactly where AI systems generate demand that does not require bleeding-edge transistors.

Why would TSMC outsource anything?

AI chips are not single pieces of silicon. A finished accelerator module combines the leading-edge logic die with surrounding elements — interposers, power delivery, interface silicon, RF components — produced on older nodes. Capacity on those mature processes tightened sharply as AI datacenter buildouts accelerated.

TSMC's advanced packaging and front-end wafer capacity for AI products has been effectively sold out, and the company has repeatedly raised capital spending to add more. Moving a supporting component to GlobalFoundries frees TSMC's own mature-node lines for work only it can perform.

For GlobalFoundries, the deal provides volume tied to the fastest-growing end market in semiconductors without requiring the multi-billion-dollar investment that leading-edge EUV lithography tools demand. It also deepens the company's role in the US and European supply chains, where it operates fabs in New York, Vermont, Dresden, and Singapore and has committed to government-supported expansion projects.

What remains unknown?

Several questions stay open until the companies confirm terms:

• Which component GlobalFoundries will produce, and on which process
• Whether production runs in US, European, or Singapore facilities
• Contract volume and revenue impact for GlobalFoundries
• Whether the arrangement covers multiple generations of AI products

How does this fit the broader picture?

The foundry industry has spent the past three years hedging against concentration risk. Customers and governments have pushed for second sources after pandemic-era shortages exposed how few suppliers control critical chipmaking steps.

A TSMC-to-GlobalFoundries flow does something quieter but potentially more significant: it turns a direct competitor into a capacity partner inside the AI supply chain. If the component deal scales, it could establish a template for splitting AI module production across foundries — advanced logic at the leader, supporting silicon distributed to specialists.

For now, the reported agreement stands as a single, concrete data point: the world's dominant AI chipmaker has decided that at least one key component is better built somewhere else.