TLDR
- DeepSeek released open-source programming tools built with Huawei for its Ascend AI chips.
- The toolkit includes TileLang, a rival to Nvidia’s CUDA software platform.
- DeepSeek plans to use over 160,000 Huawei Ascend chips at a new data center in Inner Mongolia.
- The company is finishing a funding round valuing it near $74 billion ahead of a possible public listing.
- Huawei recently unveiled new Ascend chips it expects to be widely used for AI training by 2027.
DeepSeek has released a set of software tools built with Huawei Technologies to program AI chips. The Chinese AI lab announced the move Wednesday on its WeChat account.
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 tools are open-source and free for anyone to download. They are designed to work with Huawei’s Ascend line of AI accelerators.
What the New Software Does
One of the key tools is called TileLang. It serves as a coding platform similar to Nvidia’s CUDA, which is widely used across the AI industry to build and train models.
DeepSeek said a simple and effective coding process is needed to build a strong AI ecosystem. The company worked directly with Huawei to fine tune performance for the Ascend 950 chip.
Huawei said it gave full support during the development of the software. Both companies plan to keep working together on future projects.
The two firms also advanced a project known as a supernode. This system links together 128 Ascend 950 chips to work as one unit.
DeepSeek has used a mix of hardware from both Nvidia and Huawei in the past. Early on, the company relied heavily on Nvidia’s H800 chip for training its AI models.
That reliance has shifted over the course of 2026. DeepSeek has moved more of its work toward Huawei’s domestic hardware instead.
Part of this shift comes from supply limits. China has largely been blocked from buying Nvidia’s most advanced chips.
DeepSeek’s V4-Flash AI model was built with this shift in mind. It was optimized specifically to run on Huawei chips rather than Nvidia hardware.
Scaling Up With a New Data Center
DeepSeek is also building a large data center in Inner Mongolia. Bloomberg News reported last month that the company plans to install more than 160,000 Huawei Ascend chips there.
This would mark one of the largest deployments of Chinese-made AI chips to date. It reflects how far DeepSeek has moved away from foreign hardware.
Alongside this, DeepSeek is finishing a new funding round. The round is expected to value the company at close to $74 billion.
That valuation comes ahead of a possible public listing. Reports suggest the listing could happen before the end of this year.
DeepSeek rarely comments on internal technical work. This week’s announcement is a rare public look at how closely it works with Huawei.
Huawei, for its part, has been central to China’s push to build its own AI chip supply chain. The country’s government has rolled out economic and regulatory support for this effort over the past year.
Huawei recently announced a new line of Ascend chips built for large computing clusters and data centers. The company has said it expects these chips to be used widely for AI model training by 2027.
Nvidia shares dipped slightly following the news, trading down less than one percent. The stock move reflects ongoing questions about Nvidia’s position in the Chinese market as local alternatives grow.
Stop guessing and start investing with confidence. KnockoutStocks gives you the AI insights, market intelligence, and stock research you need to spot opportunities, cut through the noise, and make smarter investment decisions — all in one powerful platform.
Sign up today and get 50% OFF full access to our premium stock picks.
Simply use coupon code SPECIAL50 at checkout to claim your exclusive discount.







