U.S. Allocates $300M to Help GlobalFoundries Develop AI Chip Optical Interconnects
Alina Collins
The U.S. Commerce Department is awarding $300 million to GlobalFoundries to develop silicon photonics and co-packaged optics for AI data centers — the first major CHIPS Act research allocation targeting optical interconnects, signaling Washington's push to reshore a supply chain currently dominated by overseas manufacturers.
What problem does this funding address?
Inside AI data centers, chips pass data to each other via electrical signals — a method hitting hard limits on speed and power consumption. Silicon photonics — moving data between chips with light instead of electricity — raises bandwidth and cuts energy use at the same time.
The $300 million comes from the Commerce Department, part of the Trump administration's effort to channel CHIPS Act research funds into targeted semiconductor technologies.
This means → Washington has concluded that optical interconnects are not a future concept but a present bottleneck for AI infrastructure, worth a standalone funding line.
What are the technical targets?
GlobalFoundries set two hard benchmarks: data transfer rates up to 400 gigabits per second and energy efficiency up to five times better than current solutions.
The project also covers co-packaged optics, or CPO — integrating silicon-photonics modules directly next to the AI processor so data travels a shorter path with less loss.
In plain terms = today, chips talk to each other through copper wiring — slow and power-hungry. Replace that with fiber-optic links soldered right next to the chip, and both speed and energy use jump a level.
Why does it have to be built in the U.S.?
Manufacturing capacity for silicon photonics and advanced optical packaging currently sits mostly outside the U.S. — TSMC and Israel's Tower Semiconductor are leading producers.
This means → the U.S. leads in AI chip design, but the manufacturing step that lets chips "talk with light" relies heavily on offshore facilities, creating a supply-chain gap.
The strategic intent is explicit: bring optical-interconnect manufacturing home. R&D will take place at GlobalFoundries' fabs in Malta, New York and Burlington, Vermont.
Why GlobalFoundries?
CEO Tim Breen put it directly: "GlobalFoundries has spent more than a decade building the technology, capacity, and ecosystem needed to lead this transformation."
This reflects GlobalFoundries' positioning strategy: rather than competing with TSMC on cutting-edge process nodes, it builds a moat in specialty processes like silicon photonics.
In plain terms = TSMC makes the smallest chips. GlobalFoundries chose a different lane — building the "optical plumbing" that lets chips communicate at high speed.
How does this fit the broader funding picture?
Washington has already committed $150 million for semiconductor manufacturing equipment and $2 billion for quantum computing. This $300 million for silicon photonics sits in the middle by size but targets the lane closest to AI's immediate supply-chain needs.
This means → three funding lines sketch three pillars of U.S. chip strategy: the machines that make chips, the light that connects chips, and the quantum that computes next.
The real test is whether a silicon-photonics supply chain can reach production scale on U.S. soil. If the work stays in the lab, $300 million is a check, not a production line.
Content is for reference only, not financial advice.