News

[News] NVIDIA and TSMC Showcase Silicon Photonics Innovations as Samsung Speeds up Development


2024-12-19 Semiconductors editor

Leading GPU manufacturer NVIDIA recently unveiled its vision for future AI chip designs featuring silicon photonics at the IEDM 2024 conference. According to a report from Business Korea, NVIDIA highlighted the potential of silicon photonics for chip-to-chip connections within AI data centers, drawing significant attention from the chip industry.

NVIDIA presented a silicon photonics prototype developed in collaboration with TSMC at the event, the report notes. TSMC also provided a detailed overview of its silicon photonics manufacturing technology, which involves fabricating two advanced devices and seamlessly integrating them as if they were a single chip using the hybrid bonding (SoIC) method.

The report from Business Korea, citing industry sources, highlights that silicon photonics technology delivers data transmission speeds hundreds of times faster than copper. This breakthrough is particularly significant for AI data centers, where rapid and efficient data transfer is essential.

As NVIDIA and TSMC’s collaboration on silicon photonics underscores a strong partnership between the leading fabless company and the most advanced foundry, Samsung is gaining attention to see if it can close the gap through technological partnerships with key customers, the report states.

Samsung is actively planning the development of its own silicon photonics technology, the report notes. The company has named its silicon photonics processes as “I-CubeSo” and “I-CubeEo” and is working on related products.

Some of the clients interested in Samsung’s silicon photonics technology are said to include NVIDIA, Broadcom, and Marvell, as per a senior official from Samsung cited by Business Korea.

Silicon photonics, as described in the report, is an optical semiconductor technology that enables semiconductors in electronic devices to communicate through the use of light instead of electricity. Compared to traditional electrical connections, it offers higher speeds and lower power consumption. As a result, it plays a crucial role in data-intensive applications, such as AI and data centers.

Read more

(Photo credit: NVIDIA)

Please note that this article cites information from Business Korea.

Get in touch with us