The US government will award $300 million to chipmaker GlobalFoundries to bolster research and development of silicon photonics, a technology increasingly viewed as essential for the future of AI infrastructure, according to an announcement on Wednesday. Under the Letter of Intent, the Department of Commerce’s CHIPS Research and Development Office will advance next-generation optical materials, wafer-level optical packaging, and technologies for AI infrastructure. The US government will also acquire approximately a 1% equity stake in the company.
The funding comes as the US seeks to maintain its technological edge in the face of rapid advancements by Chinese semiconductor firms, particularly in areas like advanced packaging and alternative chip architectures. As EastFrontier has previously reported, China has been investing heavily in photonics to power AI with light, bypassing US chip curbs, making the GlobalFoundries award a direct competitive response.
The Promise of Silicon Photonics for AI
Silicon photonics involves using light (photons) instead of electricity (electrons) to transmit data between computer chips. This approach offers significantly higher bandwidth and lower power consumption compared to traditional copper interconnects, addressing a major bottleneck in modern AI data centers.
As AI models grow exponentially in size, the speed at which data can be moved between processors and memory becomes a critical limiting factor. Silicon photonics promises to alleviate this constraint, enabling faster and more efficient AI training and inference. The technology is particularly relevant for the co-packaged optics (CPO) architectures that next-generation AI accelerators are expected to adopt, where optical interconnects are integrated directly into the chip package rather than running through external cables.
GlobalFoundries, which operates fabs in the US, Germany, and Singapore, is well-positioned to commercialize silicon photonics at scale. Unlike leading-edge logic fabs, silicon photonics manufacturing does not require the most advanced lithography nodes, making it a domain where the US can build and maintain a competitive advantage without being entirely dependent on TSMC or Samsung.
Strategic Imperatives
The investment in GlobalFoundries highlights the US government’s recognition that maintaining leadership in AI requires more than just advanced logic chips; it requires innovation across the entire semiconductor stack, including interconnects and packaging. While Chinese firms like Huawei are aggressively pursuing advanced packaging solutions like glass substrates to bypass lithography constraints, the US is betting that breakthroughs in silicon photonics will provide a decisive advantage in data center performance.
The $300 million award is part of the broader CHIPS and Science Act, which aims to revitalize domestic semiconductor manufacturing and research. By targeting silicon photonics, the US is strategically investing in a technology that could define the next era of high-performance computing, ensuring that the infrastructure powering future AI breakthroughs remains firmly rooted in American innovation.
The equity stake component of the deal is also notable. By acquiring approximately 1% of GlobalFoundries, the US government is not merely subsidizing research, it is becoming a stakeholder in the company’s commercial success. This model, more common in defense procurement than civilian semiconductor policy, signals a shift in how Washington views its relationship with the domestic chip industry. As China continues to pour state capital into its own semiconductor champions, the US appears to be moving toward a more interventionist posture that blurs the traditional line between public funding and private enterprise. The GlobalFoundries deal may be a template for future CHIPS Act awards as the competition for AI infrastructure leadership intensifies.
For the AI industry more broadly, the race to develop silicon photonics is a race to solve one of the most fundamental constraints on future model scaling. As transformer architectures grow larger and more interconnected, the bandwidth between compute nodes becomes a primary bottleneck. Whoever solves the optical interconnect problem at commercial scale, whether through GlobalFoundries, Intel’s co-packaged optics program, or a Chinese alternative, will hold a structural advantage in the next generation of AI infrastructure. The $300 million investment signals that Washington understands this, and is willing to use public capital to ensure the answer is American.
GlobalFoundries’ existing manufacturing footprint, with fabs in Malta (New York), Dresden (Germany), and Singapore, also gives the silicon photonics program a degree of geographic resilience that purely domestic programs lack. If the technology can be transferred across these sites, it will be available to allied nations as well as the US market, potentially forming the backbone of a Western silicon photonics supply chain that can compete with Chinese alternatives on both performance and geopolitical reliability.
The Letter of Intent is expected to be converted into a final award agreement within the coming months, at which point GlobalFoundries will be required to publish detailed milestones and compliance commitments, providing a public benchmark against which progress can be measured. For the broader CHIPS Act program, the GlobalFoundries deal represents a strategic diversification beyond leading-edge logic: by funding silicon photonics alongside the advanced fab investments at TSMC Arizona and Intel, the US is building a more complete domestic AI semiconductor ecosystem that addresses not just chip manufacturing but the interconnect infrastructure on which next-generation AI systems will depend.