📰 Key Summary

NTT Data Group is in final-stage negotiations with US AI company Anthropic to build a dedicated data center in western Japan (Osaka), according to multiple sources who spoke to Nikkei Asia. This would be Anthropic’s first dedicated data center in Japan, with processing capacity expected to rank among the largest in the country. The new facility could reportedly cost up to $6.3 billion to build, with partial operations expected to begin in 2028. The deal is still in its final negotiation stage and hasn’t been officially announced yet, but multiple sources consistently indicate that the NTT Data-Anthropic partnership is wrapping up. This move shows Anthropic continuing to expand its computing infrastructure footprint across the Asia-Pacific region, partnering with Japan’s homegrown tech giants to secure the large-scale compute resources needed to support growing demand for model training and inference. See the original article for full details.

💬 JudyAI Lab Take

NTT Data is in final-stage talks with Anthropic over a dedicated data center in Osaka that’s expected to rank among Japan’s largest, with construction costs up to $6.3 billion and partial launch in 2028 — this is the kind of signal worth tracking to gauge the pace of AI infrastructure expansion.

This partnership reflects a clear trend: major AI companies’ compute needs have outgrown just building or renting cloud facilities — they’re now directly co-building dedicated infrastructure with local tech giants. NTT Data brings local power, land, regulatory know-how, and operational experience in Japan, while Anthropic needs enough compute to support the growing scale of model training and inference. This division of labor lets both sides skip the time cost of building from scratch. For AI builders, this is also a reminder: when you’re evaluating the stability or growth potential of any AI service, the compute infrastructure behind it often says more about that company’s growth ceiling over the next 3-5 years than the model’s technical specs do.

Next time you’re evaluating an AI tool or platform, it’s worth asking: how solid is its compute backbone, and how much usage growth can it actually support?

📅 Original Source

🔗 Further Reading