Urgent.News

What's breaking now, across thousands of outlets.

AI

DeepSeek plans big Huawei AI chip order to power new data centre

The startup doesn’t currently plan to use the 950DT chips for training, one of the people said, even though Huawei designed and marketed these processors to serve that more demanding function. While DeepSeek tried to train its previous models on Huawei chips, it has so far relied on Nvidia accelerators for this crucial step.

DeepSeek plans big Huawei AI chip order to power new data centre

DeepSeek, an AI startup, has announced plans to purchase at least 160,000 of Huawei's latest AI chips for its new data center in Inner Mongolia. This could potentially create one of the largest clusters of Huawei AI chips in China, helping the country wean off its reliance on US chipmaker Nvidia. Huawei's next-generation Ascend 950DT chips will be used for operating the models, though they won't be used for training at the moment.

The installation timeline depends on Huawei's production capabilities, and the shortage of components like top-end memory may limit the output of the 950DT chips to a few hundred thousand this year. The Inner Mongolia hub is crucial for DeepSeek's AI infrastructure goals, which aim to consume around 1 gigawatt of energy, enough to power approximately 750,000 homes.

Written by urgent.news from Economic Times Tech's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at economictimes.indiatimes.com →

More in AI

The Part Of Your RAG Pipeline That Decides Everything

Most teams debugging a disappointing RAG system start at the end of the pipeline. They swap the model, tune the prompt, raise top k.

  • Retrieval-augmented generation pipeline often fails due to ingestion stage issues.
  • Parsing non-text formats like PDFs is challenging and crucial for RAG quality.
  • Layout-aware parsers and metadata attachment are essential for effective chunking.

Everything I Didn't Build

For a few weeks I ran a proper multi-agent orchestrator. The concept was right, and I still think the people building those tools are pointed the correct way.

  • Author builds complex agent orchestrator with multiple components
  • System inefficient due to extensive communication overhead
  • Author simplifies using markdown, Git, and agent board

More from Friday 4 September →