DeepSeek Launches Open-Source Toolkit for Huawei AI Chips, Moving Away from Nvidia (NVDA)

Key Highlights

Table of Contents

Chinese AI company DeepSeek launched open-source development tools created in partnership with Huawei for Ascend AI accelerators.
The release features TileLang, a programming platform designed to compete with Nvidia’s dominant CUDA ecosystem.
DeepSeek is constructing a massive Inner Mongolia data center equipped with more than 160,000 Huawei Ascend processors.
A funding round nearing completion is set to value the AI startup at approximately $74 billion before a potential IPO.
Huawei has introduced next-generation Ascend processors projected to become mainstream for AI model development by 2027.

DeepSeek has unveiled a comprehensive software development kit created alongside Huawei Technologies for programming artificial intelligence processors. The announcement came Wednesday through the AI laboratory’s official WeChat channel.

DeepSeek and Huawei are building an alternative to Nvidia’s CUDA as China ramps up its domestic AI chip industry. Via Bloomberg:

DeepSeek has released an open-source toolkit for Huawei’s Ascend accelerators, including TileLang support optimized for Ascend 950.

The tools… pic.twitter.com/3C2hvNK6Pn

— Chubby♨️ (@kimmonismus) September 30, 2026

The development suite is distributed as open-source software available for public download at no cost. It’s specifically engineered to support Huawei’s Ascend series of AI computing hardware.

Capabilities of the New Development Platform

Among the core components is TileLang, a programming framework that functions as an alternative to Nvidia’s CUDA ecosystem, which has become the industry standard for artificial intelligence development and model training.

According to DeepSeek, creating a streamlined and efficient programming environment is essential for cultivating a robust AI development ecosystem. The organization collaborated extensively with Huawei to optimize functionality specifically for the Ascend 950 processor.

Huawei confirmed it provided comprehensive technical assistance throughout the software creation process. The two organizations have committed to ongoing collaboration on additional initiatives.

The partnership also produced advances in supernode technology, which interconnects 128 Ascend 950 processors to function as a unified computing system.

Historically, DeepSeek has operated with a combination of Nvidia and Huawei components. In its earlier stages, the company depended significantly on Nvidia’s H800 processors for model training operations.

This dependency has transformed throughout 2026, with DeepSeek transitioning a substantial portion of its infrastructure to domestically-produced Huawei components.

This strategic realignment stems partly from procurement constraints. Chinese companies face significant barriers accessing Nvidia’s cutting-edge chip technology.

DeepSeek’s V4-Flash artificial intelligence model exemplifies this transition, having been engineered from the ground up to leverage Huawei processors instead of Nvidia technology.

Massive Infrastructure Expansion Underway

The company is simultaneously developing an extensive computing facility in Inner Mongolia. According to Bloomberg News coverage from last month, DeepSeek intends to deploy over 160,000 Huawei Ascend chips at this location.

This installation would represent one of the most significant commitments to Chinese-manufactured AI hardware on record, demonstrating DeepSeek’s dramatic pivot away from international suppliers.

Concurrently, DeepSeek is finalizing a capital raising initiative expected to establish the company’s worth at nearly $74 billion.

This assessment precedes a prospective stock market debut, with industry observers suggesting the listing may occur within the current calendar year.

DeepSeek typically maintains discretion regarding its engineering operations and technical partnerships. This week’s disclosure offers an uncommon glimpse into the depth of its collaboration with Huawei.

Huawei has emerged as a cornerstone of China’s initiative to establish autonomous AI semiconductor manufacturing capabilities. Government authorities have implemented substantial economic incentives and policy frameworks supporting this objective throughout the previous twelve months.

The telecommunications giant recently introduced an advanced Ascend processor line optimized for large-scale computing infrastructure and data center deployments. Huawei projects these chips will achieve widespread adoption for AI training applications by 2027.

Nvidia’s stock experienced minor downward movement following the announcement, declining less than one percent in trading. The modest reaction underscores persistent uncertainty regarding Nvidia’s competitive positioning in China as domestic semiconductor alternatives mature.

✨ Limited Time Offer