The article discusses the current challenges faced by AI developers, particularly regarding data transfer speeds between chips and servers, which are currently limited by copper wiring. Nvidia is responding to these challenges by investing billions into photonics technology, which uses light for data transmission, potentially improving AI model performance.
Since March, Nvidia has invested $2 billion in companies like Lumentum, Coherent, and Marvell, and an additional $500 million in Corning for optical connectivity solutions. CEO Jensen Huang emphasized the need for increased silicon photonics capacity, indicating that Nvidia is beginning to implement this technology in its networking and GPU interconnect platforms.
However, analysts like Alan Weckel and Gil Luria highlight significant hurdles, including manufacturing challenges and the need for redesigning existing products to accommodate optical components.
The article also notes broader market trends, such as Snowflake's stock surge following positive guidance and AI deals, and Micron's stock reaching a $1 trillion market cap amid high demand for memory chips linked to the AI boom