September 17, 2026, (Inside AI) — GlobalFoundries and Marvell Technology have expanded their agreement to boost production capacity for a specialized semiconductor used in high-speed optical connections within data centers, the companies announced on Thursday. The move aims to address surging demand for faster data transmission driven by artificial intelligence workloads.
Financial terms of the expanded deal were not disclosed. The announcement sent GlobalFoundries shares up nearly 6% in premarket trading, while Marvell gained 4.5%, signaling investor confidence in the partnership's growth potential.
The semiconductor in question is a key component in optical interconnects, which use light to transmit data between servers and switches. As AI models grow larger and more complex, data centers require faster and more efficient connections to handle massive datasets. Traditional copper connections struggle with bandwidth and power consumption at these scales, making optical technology essential.
GlobalFoundries, a contract chipmaker, will manufacture the chips for Marvell, a fabless semiconductor company that designs them. The expansion likely involves increasing wafer capacity at GlobalFoundries' facilities, though specific locations and timelines were not provided.
Read: Huawei Unveils Ascend 960 SuperPoD with NPO Technology for AI Infrastructure
The deal underscores a broader trend: AI's insatiable appetite for computing power is reshaping the semiconductor supply chain. Data centers are investing heavily in infrastructure to support AI training and inference, driving demand for specialized chips. Optical connectivity is a critical piece of this puzzle, enabling the high-speed, low-latency links needed to stitch together thousands of processors.
Industry analysts have noted that the optical interconnect market is expected to grow significantly over the next decade. According to a report by MarketsandMarkets, the global optical interconnect market size is projected to reach $30.4 billion by 2028, up from $14.6 billion in 2023, at a compound annual growth rate of 15.8%. This growth is largely attributed to the increasing adoption of cloud computing and AI.
The partnership between GlobalFoundries and Marvell is not new. In 2022, the two companies announced a collaboration to produce chips for data center and 5G infrastructure. The expanded agreement suggests that initial volumes were insufficient to meet demand, prompting both firms to scale up.
Marvell has been actively expanding its optical portfolio. In 2023, it acquired Inphi, a leader in high-speed data movement, for $10 billion. That acquisition bolstered Marvell's capabilities in electro-optics and positioned it as a key player in the data center interconnect market. GlobalFoundries, meanwhile, has been investing in its Silicon Photonics platform, which integrates optical components onto silicon chips.
Competitors are also racing to capture this market. Intel, Broadcom, and Cisco are among the major players developing optical interconnect solutions. Intel has its own silicon photonics products, while Broadcom offers a range of optical transceivers. The competition is fierce, but the growing pie means multiple players can thrive.
The AI boom has created a supply crunch for many types of chips, from GPUs to networking components. Companies are scrambling to secure capacity, often signing long-term agreements with foundries. GlobalFoundries, which operates fabs in the United States, Germany, and Singapore, is a key beneficiary of this trend. Its stock has risen over 20% year-to-date, outperforming the broader semiconductor index.
Read: Delos Data Raises $100 Million for AI Data Center Networking Chips
Marvell, too, has seen its shares climb, up about 15% this year. The company's data center segment has been a bright spot, with revenue growing 40% year-over-year in the last quarter. Analysts expect this momentum to continue as AI adoption accelerates.
However, challenges remain. The semiconductor industry is cyclical, and any slowdown in AI spending could impact demand. Geopolitical tensions, particularly between the U.S. and China, could disrupt supply chains. Both companies have exposure to the Chinese market, which is a major consumer of data center chips.
Moreover, the transition to optical interconnects is not without technical hurdles. Integrating optical components with silicon requires precise manufacturing and can be costly. Yield rates and reliability are critical factors. GlobalFoundries and Marvell will need to ensure their solutions meet the stringent requirements of hyperscale data centers.
Despite these challenges, the expanded agreement signals confidence in the long-term demand for AI infrastructure. As AI models become more sophisticated, the need for faster, more efficient data movement will only grow. Optical interconnects are poised to play a central role in this evolution.
The announcement comes amid a flurry of activity in the semiconductor sector. Just last week, TSMC announced plans to build a new fab in Arizona to meet AI demand. Samsung is also ramping up production of high-bandwidth memory. The race is on to build the backbone of the AI era.
For investors, the GlobalFoundries-Marvel deal is a reminder that the AI trade extends beyond obvious names like Nvidia. The infrastructure layer, including foundries and networking chipmakers, offers compelling opportunities. As data centers evolve, companies that enable faster connectivity will be well-positioned.
GlobalFoundries and Marvell did not provide additional details on the expanded agreement. Both companies are expected to discuss their AI strategies at upcoming industry conferences. The deal is a clear bet that the AI-driven demand for optical connectivity will continue to surge.
