Nvidia unveils new AI chip communication tech NVLink Fusion

Nvidia has launched NVLink Fusion, a chip-to-chip interconnect technology designed to improve communication among chips in AI systems.

Developed at Nvidia’s research center in Israel, NVLink Fusion is an upgrade to its existing NVLink technology.

For the first time, Nvidia will license this technology to third-party chipmakers.

This enables them to create advanced AI systems with interconnected chips.

🔗 Source: Calcalist

🧠 Food for thought

1️⃣ Nvidia’s strategic shift from closed ecosystem to platform business model

NVLink Fusion represents a significant evolution in Nvidia’s business strategy, moving from a vertically-integrated approach to a platform model that embraces third-party integration.

The technology builds directly on high-speed interconnect expertise acquired through the $6.9 billion Mellanox acquisition in 2019, showing how Nvidia is leveraging strategic acquisitions to expand its technical capabilities1.

Earlier NVLink iterations were strictly for GPU-to-GPU communication within Nvidia’s ecosystem, as seen in the Turing architecture which used NVLink for multi-GPU setups2.

This shift to licensing core technology resembles the platform strategies that transformed other tech giants, enabling Nvidia to maintain its central position in AI computing while allowing ecosystem partners like Marvell and MediaTek to build complementary products.

By providing the 1.8 TB/s high-speed interconnect that far outpaces PCIe standards, Nvidia ensures its technology remains the connective tissue of next-generation AI systems, even as it allows more hardware diversity1.

2️⃣ Regional diversification as semiconductor nationalism intensifies

Nvidia’s announcement of a new Taipei headquarters, continued investment in Israeli R&D, and opening technology to Asian chipmakers demonstrates a strategic response to increasing semiconductor nationalism.

The timing is particularly significant as it comes during the first major gathering of computing executives in Asia since former U.S. President Trump threatened sweeping tariffs aimed at reshoring tech manufacturing.

Recent analysis shows semiconductor tariffs have created significant market uncertainty, with AMD expecting a $1.5 billion revenue loss due to AI chip export curbs to China3.