As artificial intelligence systems continue to grow, the industry’s biggest challenge is beginning to shift.

For years, progress was measured by how much faster individual processors became. Today, however, the performance of large AI systems increasingly depends on how efficiently thousands of chips exchange information across vast computing clusters.

Recognizing this emerging bottleneck, the U.S. Department of Commerce has signed a Letter of Intent (LOI) to award GlobalFoundries $300 million through the CHIPS Research and Development Office. The proposed funding will support the development of silicon photonics and co-packaged optics, technologies designed to move data faster and more efficiently inside AI data centers.

Unlike earlier CHIPS Act investments that focused on expanding semiconductor manufacturing capacity, this initiative targets the technologies that connect AI processors together—an increasingly critical part of next-generation computing infrastructure.

Why the U.S. Is Funding GlobalFoundries

The proposed award is not intended to build a new semiconductor fabrication plant or develop another AI processor.

Instead, it addresses one of artificial intelligence’s fastest-growing engineering challenges: moving enormous volumes of data across thousands of processors working simultaneously.

Modern AI models are trained on massive computing clusters where GPUs and AI accelerators constantly exchange information. As those clusters continue to expand, the speed and efficiency of these interconnects increasingly determine overall system performance. Improving the flow of data has therefore become just as important as improving the processors themselves.

What Is Silicon Photonics?

Most computers today transfer data between chips using electrical signals carried through copper connections. While this approach has supported decades of advances in computing, it becomes less efficient as bandwidth demands continue to increase.

Silicon photonics replaces much of that electrical communication with light.

By transmitting information through microscopic optical waveguides integrated onto silicon, the technology can carry significantly more data while consuming less power and generating less heat.

For AI data centers, where processors exchange enormous amounts of information every second, these advantages can improve performance while reducing energy consumption.

The Commerce Department’s investment reflects growing confidence that silicon photonics will become a foundational technology for future AI infrastructure.

Why Co-Packaged Optics Matters

The funding also supports co-packaged optics (CPO), an emerging technology that places optical communication components much closer to processors and networking chips.

In conventional systems, electrical signals travel from a processor to a separate optical module before entering fiber-optic cables. That extra distance increases power consumption and creates limitations as network speeds continue to rise.

Co-packaged optics shortens that path by integrating optical engines alongside the processor itself, allowing data to enter optical networks more quickly and efficiently.

According to GlobalFoundries, the research aims to enable data-transfer speeds of up to 400 gigabits per second while improving energy efficiency by as much as five times compared with current implementations.

Why GlobalFoundries?

GlobalFoundries occupies a distinctive position within the semiconductor industry.

Unlike companies such as NVIDIA, AMD, or Qualcomm that primarily design chips, GlobalFoundries manufactures semiconductors for a broad range of customers while also developing specialized technologies used in communications, automotive systems, aerospace, and industrial electronics.

The company operates major manufacturing facilities in New York, Vermont, Germany, and Singapore, and in 2024 received approximately $1.5 billion under the CHIPS Act to expand U.S. semiconductor production.

The proposed $300 million award builds on that earlier investment but shifts the focus from manufacturing capacity to research and development in advanced optical interconnect technologies.

The work is expected to take place primarily at the company’s facilities in Malta, New York, and Burlington, Vermont, supporting broader efforts to strengthen a domestic silicon photonics ecosystem.

How This Fits Into the CHIPS Strategy

The proposed award also reflects an evolution in U.S. semiconductor policy.

When the CHIPS Act was introduced, the priority was expanding domestic semiconductor manufacturing after years of concern over reliance on overseas fabrication.

As AI infrastructure has advanced, policymakers have begun focusing on another challenge: ensuring that future computing systems can move data as efficiently as they process it.

That has elevated technologies such as advanced packaging, optical interconnects, silicon photonics, and high-speed networking from supporting components to strategic priorities.

By investing through the CHIPS Research and Development Office, the United States is broadening its semiconductor strategy beyond manufacturing more chips to developing the technologies that allow entire AI systems to operate at larger scales.

What This Means for the Semiconductor Industry

The announcement highlights a broader shift taking place across the semiconductor industry.

For decades, computing performance was driven primarily by improvements in individual processors. AI has changed that equation. Future gains will increasingly depend on how effectively thousands of processors operate as a single system.

That is making optical interconnects, advanced packaging, and high-speed networking as strategically important as processor design itself.

The proposed $300 million award to GlobalFoundries illustrates this shift. Rather than focusing solely on semiconductor production, the United States is investing in technologies that could determine how efficiently future AI infrastructure operates.

As AI systems continue to grow in scale, leadership in the semiconductor industry will depend not only on building faster chips, but also on building faster, more efficient networks that connect them.