Key Points
Artificial intelligence companies are increasingly turning to photonics to address the inefficiency of data transfer between AI chips and systems. Photonics uses light signals instead of traditional copper cable electrical signals for data transmission and can, in some scenarios, perform computations directly. Analysts told CNBC that NVIDIA has invested billions in photonics-related firms, but the technology still faces significant challenges for widespread adoption.
The AI boom is generating an unprecedented wave of industry activity. While it shares similarities with the late-1990s dot-com bubble and the early 2000s mobile internet revolution, the scale of capital investment and the optimistic forecasts for its profound societal impact suggest the AI wave may be even more powerful. Rapid industry growth brings its own set of challenges. AI practitioners face multiple constraints: strained power supplies for large data centers, shortages of memory chips, and the increasingly prominent issue of low-efficiency data transfer between AI chips and systems. Photonics, an emerging technology, offers a new approach to solving these transmission problems. When applied to AI infrastructure, photonics uses light signals to transfer data between graphics processors, memory, network chips, servers, and data centers, moving away from reliance on copper cable electrical signals. Some photonic technologies, like fiber optic communication, are already in practical use. However, currently, most data transfer inside AI servers and racks still relies on copper wiring, which not only limits transmission speeds but also increases energy consumption. "Communication speed between chips and between server clusters is one of the main bottlenecks limiting the performance of AI models," said Gil Luria, Director of Technology Research at D.A. Davidson. "The faster the communication, the quicker users get results and complete tasks. Upgrading the connection between chips and servers to optical interconnects can significantly improve the operational performance of AI models." On March 16, 2026, in San Jose, California, NVIDIA CEO Jensen Huang spoke next to the NVIDIA Vera Rubin system at the company's global AI technology conference. Massive Investment Bets on Future Given its broad application prospects, photonics has attracted NVIDIA to invest billions of dollars in related companies. Since early March, NVIDIA has invested a total of $20 billion in three photonics R&D firms—Lumentum, Cree, and Marvell Technology. It also announced a $5 billion investment in Corning for joint development of high-end optical interconnect solutions and participated in a $500 million Series E funding round for optical startup Ayar Labs. NVIDIA CEO Jensen Huang revealed at the March global AI technology conference that the company is working to expand its silicon photonics manufacturing capacity. "The world's existing silicon photonics capacity is far from meeting our needs," he said, adding that NVIDIA has begun implementing photonics in its own networking platforms and graphics card interconnect platforms. Challenges to Widespread Adoption Scaling new technologies for mass deployment is never straightforward. Alan Weckel, Chief Analyst at market research firm 650 Group, pointed out that manufacturing is the first major hurdle: "The industry has never seen such explosive demand growth. With existing capacity constraints, it's difficult to get the supply chain to keep pace with demand." Luria believes that adapting existing AI systems is another significant challenge. "To widely adopt optical components on a large scale, you must significantly adjust existing product plans, replace copper wires with fiber optics, and redesign hardware architectures. Companies like NVIDIA will likely need one or two more product generations before optical interconnect technology becomes fully widespread." Industry Latest Developments
SAP's cloud data platform raised its performance forecast on Thursday and announced an AI computing partnership with Amazon, leading to its largest single-day stock price gain on record. Arthur Mensch, CEO of French startup Mistral AI, stated the company is exploring developing its own chips and may officially launch related products in the future. Driven by continued investor enthusiasm for the AI chip sector, South Korea's SK Hynix saw its stock price surge 11% on Wednesday, pushing its market capitalization above 1 trillion won for the first time. NVIDIA announced on Wednesday it will invest $150 billion to develop its business in the Taiwan region, lifting the local chip sector. Blue Origin's New Glenn rocket experienced an explosion during a hot-fire test on Thursday evening at the Space Force launch site at Cape Canaveral, Florida. Jeff Bezos stated there were no casualties.
This Week's Hot Stock: Micron Technology Over the past year, Micron Technology's stock price has soared. The U.S. memory chip maker's shares jumped 19% on Tuesday, pushing its market capitalization above $1 trillion for the first time. Demand for its products has surged due to a global memory chip shortage and the booming AI industry. Since the beginning of 2026, its stock price has risen nearly 200%.
Comments