Behind NTT’s $10 billion bet on optical networks: what Japan is fighting for in AI infrastructure is the control point for the next generation of industry

Behind NTT’s $1 Billion Optical Network Bet: What Japan Is Fighting For in AI Infrastructure Is the Next Generation of Industrial Control Points

NTT is preparing a fund of about $500 million, to be called the “IOWN AI Fund.” If this news is understood only as a telecom giant making industrial investments, its significance would be underestimated. More accurately, this is a strategic move around control of infrastructure in the AI era: Japan hopes to turn its traditional strengths in communications, materials, manufacturing, and systems integration into structural capabilities for the next round of technological competition.

According to public reports, the fund is expected to be established by the end of this month, with partners including South Korea’s SK Group, Taiwan’s Chunghwa Telecom, and the Development Bank of Japan. In addition, more than a dozen Japanese companies and financial institutions, including Toshiba, Sony, Fujitsu, MUFG, SMBC, and Mizuho, have also expressed interest in participating. The fund will target startups in North America, Asia, and Europe, with a focus on photonics-electronics convergence, AI semiconductors, and next-generation AI models designed for the IOWN architecture.

Behind this set of information, what truly deserves attention is not “another fund has been launched,” but the larger question Japanese industry is trying to answer: when generative AI drives the world’s data centers to unprecedented levels of power, bandwidth, and cooling demand, who can provide AI with a more efficient, scalable, and sustainable underlying network?

When NTT proposed the IOWN concept in 2019, the core idea was not simply to upgrade telecommunications networks, but to move the topology of telecom networks from electronic transmission to photonic transmission. Although today’s backbone networks already run optical signals over fiber, the processing stage still requires converting optical signals into electrical signals before computation and forwarding. What IOWN envisions is an end-to-end all-optical network that eliminates this conversion bottleneck as much as possible, thereby reducing power consumption, improving speed, and expanding capacity.

From an industrial logic perspective, the significance of this approach is amplified in the AI era. The bottlenecks in large-model training and inference are no longer just in chip compute, but also in data movement and energy consumption. In other words, competition in AI infrastructure is shifting from “who has more GPUs” to “who can make compute and networks work together more efficiently.” NTT’s bet is precisely on the most easily overlooked part of this chain—the part most likely to form a durable long-term barrier.

More importantly, IOWN is moving from technical R&D toward commercial validation. Reports note that in 2024, NTT and Chunghwa Telecom successfully launched a 3,000-kilometer subsea IOWN link connecting Japan and Taiwan, with one-way latency of just 17 milliseconds. For a cross-border network, this is not a marketing slogan but a signal with industrial meaning: the all-optical network is not staying in the lab; it is beginning to enter real international communications scenarios.

This is also why the partnership structure of this fund is more worthy of study than the amount itself.This is also why the fund’s partnership structure is more worth studying than the amount itself. NTT is not betting alone; it is building an alliance that spans telecommunications, semiconductors, finance, and corporate R&D. The participation of SK Group, Chunghwa Telecom, and the Japan Bank for International Cooperation means this is not a purely corporate financial investment, but is closer to a transnational experiment in industrial collaboration. For Japan, this kind of collaboration is extremely important, because Japan’s shortcomings in the AI era are not entirely about isolated technical bottlenecks, but about the speed at which it can integrate technology, capital, manufacturing, and standards.

The three directions targeted by the fund are also highly representative.

First, optoelectronic integration. This corresponds to the restructuring of network and computing architectures, rather than simply upgrading bandwidth. Second, AI semiconductors. The focus here is not on chasing the general-purpose chip boom, but on seeking more suitable co-design of computation and transmission around the IOWN architecture. Third, IOWN-native AI models. This direction shows that NTT is not content with providing “high-speed pipelines,” but hopes to further participate in the upper-layer definition of AI systems, so that it does not merely become an infrastructure provider in the future AI ecosystem.

From the perspective of Japan’s industrial strategy, this line of thinking is both familiar and new. It is familiar because Japanese companies have long been skilled at securing key positions in global industrial chains through materials, equipment, precision manufacturing, and systems engineering; it is new because Japan is now trying to transfer this capability to AI infrastructure and network architecture. In other words, Japan does not just want to catch up with the United States and China at the AI application layer; it also wants to build irreplaceability in the “skeleton” that supports AI.

This is also where IOWN connects with Japan’s semiconductor strategy. Over the past few years, Japan’s focus in semiconductors has gradually shifted from the device-level advantages of the consumer electronics era toward materials, equipment, advanced packaging, systems collaboration, and regional supply-chain resilience. The optical networks, AI semiconductors, and cross-border data links targeted by IOWN naturally require coordination among chip design, optical components, communications equipment, data centers, and software stacks. For Japanese companies, this is a path for reorganizing existing industrial assets.

Of course, this path is not easy. For all-optical networks to truly achieve large-scale deployment, technical validation alone is not enough; standards, ecosystems, and business models must also mature in parallel. Whether enterprise customers are willing to pay a premium for networks with lower latency and lower energy consumption, whether operators and cloud service providers will restructure existing architectures, and whether startups can develop genuinely killer applications on this new platform will all affect the speed at which IOWN spreads.

But from a longer-term perspective, Japan’s push for the IOWN fund at this point reflects a very clear practical judgment: AI infrastructure is rapidly becoming part of national-level industrial competition. The United States holds advantages in computing power, cloud platforms, and chips, while China is advancing quickly in scaled application and engineering deployment. If Japan does not want to be locked into the lower end of the next round of competition, it must form its own leadership capabilities in certain underlying standards and key nodes. All-optical networks are precisely a direction that could create differentiated advantages.In addition, this fund also reveals a shift in Japan’s innovation ecosystem: large enterprises are increasingly using capital and alliances to connect external startups and cross-border technology networks. Compared with the past model of relying on an internal R&D loop, Japan now needs to rapidly embed external innovation into its industrial system. For companies like Toshiba, Sony, and Fujitsu, participating in the fund is not just a financial investment, but also a way to plug into the future network of collaborative routes for chips and AI systems.

If the core words in global tech competition over the past decade were “cloud” and “platform,” then the new keywords over the next decade may be “power efficiency,” “photonics networks,” “AI semiconductor synergy,” and “infrastructure sovereignty.” The fund represented by NTT is not merely a strategic extension of Japan’s telecommunications industry, but an attempt by Japan to rewrite its industrial position in the AI era: moving from a traditional manufacturing powerhouse to a definer of infrastructure and a participant in technology standards.

Whether this shift can succeed remains to be seen. But one thing is certain: NTT has already pushed the question far enough ahead. In an era where AI is increasingly dependent on energy consumption and transmission efficiency, the truly important competition may lie not at the surface of the model, but at the very bottom of the network.