Key facts
- Marvell is planning to use TSMC's 1.4nm chip production technology for its next-generation AI data center products.
- The company is also reportedly adopting TSMC's sub-3nm process and silicon photonics technology for custom ASICs.
- Marvell's custom chips already represent over 25% of its data center revenue, with future growth expected to exceed 50%.
- Major cloud providers like AWS, Google, Meta, and Microsoft are increasingly developing in-house ASICs.
- TSMC's 1.4nm technology is anticipated to be available in 2028.
Marvell is strengthening its partnership with TSMC, with plans to utilize the foundry's upcoming 1.4-nanometer chip production technology. This strategic move is aimed at bolstering the company's offerings in high-speed connectivity for artificial intelligence data centers, a critical and rapidly expanding market.
The U.S. chip developer has reportedly engaged with TSMC about adopting its advanced 1.4nm process, which is expected to be available in 2028. This follows Marvell's earlier announcement of its first 2nm silicon IP, built on TSMC's 2nm process, designed for next-generation AI and cloud infrastructure. The company is also said to be leveraging TSMC's sub-3nm technology and silicon photonics to enhance its custom ASIC capabilities, aiming to significantly boost signal processing speeds.
Marvell's focus on custom chips is a key driver of its growth in the burgeoning ASIC market. Custom chips already account for over 25% of its data center revenue, with projections indicating this share will surpass 50% in the future. This strategy aligns with the broader trend of major U.S. cloud service providers, including AWS, Google, Meta, and Microsoft, accelerating their development of in-house ASICs to optimize performance and manage costs, often through dual-sourcing strategies and collaborations with chip designers like Marvell.
Marvell positions itself as a provider of a comprehensive, full-stack custom platform, integrating system architecture, design IP, silicon services, advanced packaging, and manufacturing logistics. The company has deployed numerous AI acceleration chips in production for leading cloud providers. The increasing demand for AI infrastructure is driving significant capital expenditures from these providers, shifting the focus toward accelerated architectures and custom solutions to break the reliance on expensive and scarce GPUs.