[News] Marvell Reportedly Taps TSMC’s Sub-3nm, Photonics Tech for Next-Gen ASICs

According to TechNews, citing South Korean outlet Chosun Biz, Marvell is reportedly working with TSMC to strengthen its position in the custom ASIC space by leveraging TSMC’s sub-3nm process technology. In addition, the report notes that Marvell is set to adopt TSMC’s silicon photonics technology, boosting signal processing speeds by more than tenfold.

In March, Marvell unveiled its first 2nm silicon IP for next-generation AI and cloud infrastructure. Built on TSMC’s 2nm process, the chip serves as a key component of Marvell’s platform for developing custom AI accelerators, CPUs, and networking solutions, according to its press release.

Marvell’s Rise in the Growing ASIC Market

Marvell has emerged as a key player in the ASIC market amid the rapid growth of AI. According to the report, the company secured its first custom chip order from Amazon Web Services (AWS) in 2021. In October 2024, it announced a design collaboration with Meta for application-specific semiconductors.

At its 2025 AI Investor Day held on June 17, Marvell revealed that custom chips accounted for over 25% of its Q4 FY25 data center revenue, with expectations that this share will exceed 50% in the future.

U.S. Cloud Giants Push Ahead with In-House ASICs

According to TrendForce, major U.S. cloud service providers (CSPs) are accelerating the development of in-house ASICs. Among them, Google leads with its TPU v6 Trillium and has shifted from relying solely on Broadcom to a dual-sourcing strategy that now includes MediaTek.

TrendForce also notes that AWS is continuing to develop its Trainium v2 chip in partnership with Marvell. Meta, following the deployment of its first in-house AI accelerator MTIA, is now co-developing MTIA v2 with Broadcom. Meanwhile, Microsoft is collaborating with Marvell on an enhanced version of its Maia v2 chip.

Read more

• [News] ASIC Boom by 2027? CSPs Aim to Leapfrog NVIDIA with Custom Chips — Key Moves & Partners
• Geopolitical Tensions Fuel a Wave of AI Chip Independence as US and Chinese CSPs Race to Develop In-House ASICs, Redefining the Market Landscape, Says TrendForce

(Photo credit: Marvell)