NTT leads $500M fund to scale optical AI data centers

The transition from electronic to photonic networking is no longer a research project but a commercial imperative as AI power consumption hits sustainable limits. By consolidating capital from leading Asian telcos and Silicon Valley expertise, NTT is attempting to establish the IOWN architecture as the global standard for the next-generation AI stack. For streaming and video providers, this technology promise of ultra-low latency and higher data throughput at lower power is critical for real-time edge processing and distributed cloud rendering. Watch for the first round of portfolio investments in late 2026 to see which semiconductor startups are prioritized for photonics integration.

The launch of the IOWN AI Fund follows several years of operational milestones for NTT’s all-optical vision. In November 2025, per Light Reading, NTT commercially deployed its All-Photonics Network (APN) in Hong Kong, marking one of the first major implementations outside of Japan. This site-to-site connectivity targeted the financial sector's demand for ultra-low latency, achieving 100G bandwidth between data centers. To support this global expansion, NTT Global Data Centers announced in March 2026 that it plans to nearly double its total capacity to five gigawatts over the next five years, with 3 project developments already underway in European and APAC markets.

Technological development reached a critical threshold in early 2026 when NTT demonstrated the ability to integrate end-to-end optical network visualization directly into a coherent DSP chip. This breakthrough, presented at OFC 2026, allows transceivers to detect network anomalies without standalone measurement equipment, a necessary step for maintaining the stability of the high-capacity backbones required by generative AI. Furthermore, in April 2026, per Business Wire, NTT and Chunghwa Telecom demonstrated cross-border real-time translation and intelligent transportation applications using IOWN photonic computing, validating the feasibility of real-time collaboration between distributed GPU clusters in Japan and Taiwan.

Strategic alliances are also formalizing around the IOWN Global Forum, which now includes over 170 members. According to the Forum's May 2026 updates, the board includes executives from Ericsson, Microsoft, and Red Hat, signaling a broad industry effort to standardize photonics-electronics convergence. As noted in NTT's Technical Journal, the commercial provision of second-generation photonics-electronics convergence (PEC-2) switches is scheduled to begin in late 2026, with an initial production target of 30,000 units per month to support the burgeoning AI infrastructure market.