Japan's telecommunications giant NTT is facing challenges to its next-generation communication strategy, known as the 'IOWN' (Innovative Optical and Wireless Network). IOWN aims to enhance data transmission speeds and reduce power consumption by utilizing light instead of electrical signals. However, the rise of artificial intelligence (AI) data centers has prompted major U.S. semiconductor companies like Nvidia and Broadcom to lead the development of related technologies, undermining NTT's strategy to establish international standards.
The Nihon Keizai Shimbun reported on June 17 that NTT's IOWN strategy is at a turning point. NTT introduced the IOWN concept in 2019, envisioning a system that connects data centers through long-distance communication and facilitates short-distance data transfer between semiconductors within servers using light, aiming for high-capacity, low-latency, and low-power communication.
Central to this initiative is 'photonic integration' technology, which allows for computation and control through traditional electronic circuits while using light for data transmission instead of electrical wiring. This approach aims to send vast amounts of data between AI semiconductors and servers using light, thereby reducing power consumption and heat generation. NTT's subsidiary, NTT Innovative Devices, is developing photonic integration components for server use.
The growing prominence of photonic integration is driven by the expansion of generative AI. In AI data centers, semiconductors like graphics processing units (GPUs) exchange massive amounts of data, leading to increased power consumption and heat generation, which traditional electrical wiring cannot efficiently handle. Consequently, photonic technology is emerging as a key next-generation technology for AI infrastructure.
Masahito Domizawa, CTO of NTT Innovative Devices, stated in an interview with Nikkei, "The direction is not wrong" regarding research into photonic technology. As the AI data center market grows, opportunities are opening for NTT to sell its photonic integration components to data centers. NTT announced plans to partner with Broadcom in 2025 to adopt Broadcom's components, aiming for sales in the photonic integration sector to reach hundreds of billions of yen by the 2030s.
However, Broadcom is not merely a partner for NTT; it has also begun to outpace NTT in the development of photonic integration technology. After announcing its entry into the photonic integration product market in 2021, Broadcom has already entered mass production for major U.S. tech company clients. While collaborating with NTT, it has also positioned itself as a key player in the U.S.-led standardization efforts from which NTT is absent.
Nvidia is also leveraging its substantial profits from AI semiconductors to secure a supply chain for photonic components. According to Nikkei, Nvidia has invested approximately 1 trillion yen in U.S. photonic component companies such as Coherent and Lumentum Holdings since March. The company, which dominates the AI semiconductor market, is rapidly expanding its influence over internal communication technologies in data centers.
The competition for standards is increasingly dominated by U.S. companies. In 2020, NTT, along with Intel and Sony Group, established the IOWN Global Forum to discuss standards with over 170 companies and organizations. However, six major U.S. tech companies announced in March that they would form a separate organization to create standards for photonic integration. This group includes Broadcom and Nvidia, as well as customer companies like OpenAI and Microsoft that purchase semiconductors and components. NTT's name was notably absent from this initiative.
In the AI infrastructure market, a small number of big tech and semiconductor companies effectively set the standards. The choice of technology and components is determined by the system designs of these companies. Even if NTT possesses the technology, it risks losing control over standards if it cannot integrate into the actual supply chain. Although Nvidia and Broadcom participate in the IOWN Global Forum, a U.S. corporate source told Nikkei, "Collaboration is broadly done for information gathering, and there is no need to be bound to a framework led by NTT."
NTT acknowledges its limitations. Domizawa stated regarding consultations on photonic integration, "At this point, it is difficult to say that the IOWN Global Forum is yielding tangible results." While NTT has made early strides in research and development, it is lagging behind in the development of components and devices that will actually be used in AI data center servers, as noted by Nikkei.
In search of a breakthrough, NTT is working to establish new alliances. This month, it plans to create a $500 million fund in collaboration with prominent venture investors in Silicon Valley to invest in promising startups and expand the photonic integration ecosystem. NTT is also pursuing economic collaboration in the photonic integration sector with South Korea's SK Group and Taiwan's Chunghwa Telecom.
Simultaneously, NTT is exploring collaboration with competing U.S. semiconductor companies. Domizawa remarked, "It is difficult to realize photonic integration with just one company," adding, "Sometimes it may involve shaking hands with one hand while striking with the other." Akira Shimada, president of NTT, also mentioned that discussions with Nvidia are ongoing, indicating potential collaboration opportunities.
Seven years after introducing the IOWN concept, NTT finds itself at a crossroads. It must decide whether to lead the standards for next-generation AI infrastructure using light-based communication technology or remain a supplier of components and technologies to U.S. big tech and semiconductor companies. The speed at which NTT can connect its research achievements to actual business and supply chains will likely determine its future standing.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.