The Tech Download: How chip companies are looking to use light to solve this major AI bottleneck

The surge in artificial intelligence (AI) technology is unlike any other, often compared to the dotcom boom and the mobile evolution, yet it stands out in terms of the sheer volume of investment and ambitious forecasts regarding its societal impact. However, this rapid advancement brings significant challenges, particularly in terms of energy access for extensive data centers, a shortage of memory chips, and notably, the efficiency of data transfer between AI chips and systems. A promising solution is emerging through photonics, a technology that utilizes light to facilitate data movement between graphics processing units (GPUs), memory components, networking chips, servers, and data centers. This method presents a departure from the conventional reliance on electrical signals transmitted through copper wires, which not only restricts speed but also escalates energy expenses. Gil Luria, head of technology research at D.A. Davidson, pointed out that one of the primary bottlenecks in AI performance lies in the communication speed between chips and servers. He emphasized, "The faster the communication, the quicker users receive their responses or have their tasks completed," adding that transitioning to optical connections could significantly enhance model performance. In an effort to harness this potential, Nvidia has invested billions into photonics technology developers. Since March, the company has committed $2 billion to firms like Lumentum, Coherent, and Marvell, which are all engaged in photonics advancements. Moreover, Nvidia plans to invest $500 million in Corning for the development of cutting-edge optical connectivity solutions and has participated in a $500 million Series E funding round for optics startup Ayer Labs. During the GTC event in March, Nvidia CEO Jensen Huang announced that the company is beginning to scale up its silicon photonics technology, stating, "The amount of silicon photonics technology capacity that we need is substantially higher than what currently exists globally." He also mentioned that Nvidia is initiating the integration of photonics into its networking and GPU interconnect platforms. However, launching new technology at scale presents its own challenges. Manufacturing capabilities must align with the unprecedented demand, as highlighted by Alan Weckel, a principal analyst at 650 Group. He noted that the industry has never before encountered such growth, making it difficult to adjust the supply chain. Additionally, Luria remarked on the complexities involved in adapting active AI systems to incorporate photonics, indicating that significant redesigns of product roadmaps will be necessary to replace copper wiring with optical fiber, possibly requiring additional product generations from companies like Nvidia. In other developments, Snowflake experienced record gains after raising guidance and securing an AI computing partnership with Amazon. French startup Mistral AI is contemplating the design of its own chips, as reported by CEO Arthur Mensch. Meanwhile, SK Hynix saw its market cap surpass $1 trillion, with shares soaring by 11% due to investor enthusiasm for AI-related semiconductor stocks. Taiwanese chip stocks also rallied following Nvidia’s announcement of a $150 billion investment in the region. Notably, Micron, a U.S. memory chip manufacturer, reached a $1 trillion market cap for the first time, driven by heightened demand amid the memory shortage and AI boom, with its stock climbing nearly 200% in 2026.