跳到正文
Data Center Knowledge· Shane Snider·· 2026-09-03AI 评分77

AWS建设420Tbps Sta’O’Nuk跨太平洋海底光缆支撑AI相关应用

AWS Wires a New US AI Route Across the Pacific

AI 导读

AWS正在建设连接日本与美国华盛顿州的420 Tbps Sta’O’Nuk海底光缆,该光缆采用20对光纤,计划2029年投入服务。项目配套Toptana Technologies在华盛顿州Ocean Shores开发的新登陆站及通往太平洋西北数据中心枢纽的陆地路由,将为分布式大语言模型训练等带宽密集型应用提供新的跨太平洋传输路径。

正文

Shane Snider,

Senior News Writer,Data Center Knowledge

September 2, 2026

4 Min Read

Amazon Web Services is building a 420 Tbps subsea cable between Japan and Washington State, adding a new trans-Pacific route as hyperscalers expand the networks connecting their computing infrastructure.

The Sta’O’Nuk cable will use 20 fiber pairs and is slated to enter service in 2029. AWS says the cable will support bandwidth-intensive applications such as distributed large language model training, financial transactions, edge computing, and video streaming.

The project provides AWS with a new US landing point for trans-Pacific traffic. Toptana Technologies is developing a landing station in Ocean Shores, Washington, designed to initially accommodate up to four subsea cable systems, with the potential for expansion.

As AI data center capacity spreads across regions, the networks connecting that compute are becoming a larger part of the infrastructure buildout. AWS is pairing the 420 Tbps cable with the new landing station and terrestrial routes into the Pacific Northwest.

But AI is adding to an infrastructure market that hyperscalers already dominated before the current buildout.

“Even before the rise of AI investments, hyperscalers were among the primary consumers of subsea cables and capacity,” said Jimmy Yu, vice president at Dell’Oro Group. The surge in AI demand should further drive demand for both terrestrial and submarine networks, he said.

AWS’s choice of Washington reflects the importance of connecting subsea infrastructure directly to major data center hubs, Yu added, improving connectivity while increasing route diversity and network resilience.

AI Puts More Pressure on Network Capacity

Large AI models can require terabit-scale data transport between distributed training clusters and inference locations, noted Ron Westfall, vice president and practice lead for networking and infrastructure at HyperFrame Research. As hyperscalers spread computing across regions, network backhaul and interconnect capacity could become a constraint alongside power availability, he said.

While power availability can determine where operators build large AI facilities, the network determines how effectively those facilities connect to other computing resources, cloud regions, and customers.

Washington Gets a New Cable Landing

Toptana says the Ocean Shores landing station will connect international subsea cables with terrestrial fiber and initially support four cable vaults, with room to expand to 16.

The company plans to connect the landing station to Seattle and Hillsboro, Oregon, through a terrestrial backhaul network along the Interstate 5 corridor. That will carry traffic from the coast into two of the Pacific Northwest’s major data center and network markets.

The company described the station as the first new cable landing station in Washington in more than 25 years. The Quinault Indian Nation owns Toptana, which is developing the project as an Indigenous-owned telecommunications infrastructure facility. Assured Communications serves as the project’s strategic partner, program manager, and operations service provider.

AWS Adds Another Pacific Route

AWS said Sta’O’Nuk will provide geographic diversity between North America and Japan. More than 30 trans-Pacific subsea cables currently operate between the two regions, with existing US landing sites concentrated along the California and Oregon coasts, according to AWS. A Washington landing point gives AWS another path into the Pacific Northwest and reduces reliance on established landing corridors. The inland connection to Toptana’s planned terrestrial network to Seattle and Hillsboro is key because international cable capacity has limited value if operators cannot move traffic efficiently from the coast into data centers and inland network infrastructure.

Hyperscalers Build the Network Too

AWS says its global network spans more than 20 million kilometers of fiber. Sta’O’Nuk extends that strategy into another piece of physical infrastructure under AWS’s direct control.

The cable will use quantum-safe optical encryption at Layer 1, MACsec at Layer 2, and TLS, SSL, or QUIC at Layer 4, according to AWS. The company plans to armor and bury the cable in vulnerable areas, including at depths of up to 1,500 meters, and to use horizontal directional drilling to protect the cable near landing sites.

AWS estimates that roughly 200 subsea cable cuts occur worldwide each year, mostly from fishing activity and dragging anchors, making route diversity and physical protection important design considerations.

Washington’s last two subsea cables were laid in 1999, according to Data Center Dynamics: Pacific Crossing-1 at Harbour Pointe and Alaska United East connecting Washington with Alaska.

Sta’O’Nuk would put Washington back on the trans-Pacific cable map while giving AWS another route into the Pacific Northwest.

The Network Becomes Part of the AI Buildout

AWS’s project shows how hyperscalers are extending AI infrastructure beyond the data center campus. Power generation and transmission, data center capacity, terrestrial fiber, and international connectivity all support the computing behind AI services.

AI is adding pressure across that infrastructure stack. For network operators, the question is how much additional capacity distributed training, inference, and other AI workloads will require as hyperscalers build computing across multiple regions.

AWS is preparing for that demand with a 420 Tbps trans-Pacific cable, a new Washington landing site, and terrestrial routes into the Pacific Northwest. The cable is only one piece of the investment. The landing station’s capacity for additional systems provides a platform for more international connectivity into the region over time.

About the Author

Shane Snider

Senior News Writer, Data Center Knowledge

Shane Snider is Senior News Writer at Data Center Knowledge, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. His reporting focuses on the operational, economic, and environmental forces reshaping digital infrastructure, including AI factories, utility constraints, liquid cooling, renewable energy procurement, and next-generation data center architectures. He has won recent Azbee awards for news series and government reporting. Based in Raleigh, North Carolina, Snider covers how hyperscalers, utilities, chipmakers, and infrastructure providers are responding to the rapid rise of AI workloads and global compute demand.

You can reach Shane at [email protected] or on LinkedIn.

来源:Data Center Knowledge · datacenterknowledge.com