DeFi Intel

China debuts largest AI industrial park in Ulanqab, betting big

China's AI park in Ulanqab highlights a strategic shift towards leveraging renewable resources and infrastructure scale to counter tech export limits.

TL;DR

China opened its largest AI industrial park in Ulanqab on August 10, 2026, underscoring a push for domestic AI hardware. Huawei claims its Ascend chips have outsold Nvidia in China, reflecting a shift in volume while Nvidia still dominates training workloads.

Key facts

  • China unveiled its largest AI industrial park in Ulanqab on August 10, 2026.
  • Huawei says its Ascend AI chips have outsold Nvidia within China as of October 2026.
  • Domestic AI chipmakers could capture about 50% of China’s AI accelerator market by 2026.
  • Nvidia’s share in China may fall to roughly 8% in 2026, down from about 40% in 2025.
  • DeepSeek plans to deploy at least 160,000 Huawei Ascend 950DT chips for inference.

China’s push for semiconductor self‑sufficiency took a visible step on August 10, 2026, as the nation unveiled its largest AI industrial park in Ulanqab. The development, highlighted by Huawei’s claim that its Ascend AI chips have outsold Nvidia within China, underscores a strategic shift toward domestic silicon. Huawei’s rotating chairman Eric Xu framed the move around reliable supply, while top Chinese AI labs continue to rely on Nvidia hardware for training large language models, exposing a split between inference and training workloads.

Background

The AI hardware landscape in China is defined by a tension between domestic chipmakers and foreign suppliers, especially Nvidia. US export controls have intensified Beijing’s drive for homegrown silicon, making the transition from Nvidia’s CUDA‑based ecosystem to alternatives like Huawei’s CANN a policy priority. Huawei, operating under strict US restrictions, has positioned its Ascend series as a dependable, locally sourced option, emphasizing volume over peak performance. Despite Huawei’s reported sales surge, many leading Chinese AI developers still favor Nvidia for the compute‑intensive training of large language models, as the software stack and performance advantages remain entrenched. The market share projections suggest domestic AI chipmakers could capture about half of China’s AI accelerator market by 2026, while Nvidia’s share may fall to roughly eight percent, down from around forty percent the previous year. Enforcement actions against illegal Nvidia chip smuggling and limited US approvals for certain Nvidia models create a patchwork environment, prompting Chinese firms to plan around regulatory uncertainty. This backdrop frames the significance of the Ulanqab AI park, which aims to consolidate domestic AI hardware production and accelerate inference deployments, while the training segment continues to be dominated by Nvidia’s technology.

What happened

On August 10, 2026, China announced the opening of its largest AI industrial park in Ulanqab, signaling a major investment in domestic AI hardware capacity. The park’s launch coincided with Huawei’s claim, reported by CryptoBriefing, that its Ascend AI chips have outsold Nvidia within China as of October 2026. Huawei’s rotating chairman Eric Xu emphasized the importance of a reliable domestic supply chain over raw performance. Despite this sales milestone, the report notes that many top Chinese AI labs, such as DeepSeek, continue to use Nvidia hardware for training large language models while planning to deploy at least 160,000 Huawei Ascend 950DT chips for inference. The article explains that Nvidia’s CUDA platform remains a deep‑seated software moat, making a shift to domestic training hardware costly and time‑consuming, with projected training‑workload transitions carrying roughly 50% higher costs and time. US export controls have led to arrests related to Nvidia chip smuggling and limited approvals for certain Nvidia models, creating a mixed legal environment for chip imports. Market projections in the source suggest domestic AI chipmakers could capture about 50% of China’s AI accelerator market by 2026, while Nvidia’s share may drop to around 8% that year, down from about 40% in 2025. The park’s establishment is intended to bolster inference deployment agility and support Huawei’s volume growth, even as Nvidia retains a foothold in frontier model training.

Market impact

The source indicates that Huawei’s Ascend chips have outsold Nvidia within China as of October 2026, reflecting a shift in unit volume. Projections suggest domestic AI chipmakers could command roughly 50% of China’s AI accelerator market by 2026, while Nvidia’s share may fall to about 8%, down from around 40% in the prior year. DeepSeek’s plan to deploy at least 160,000 Ascend 950DT chips for inference highlights the scale of domestic hardware adoption. The report also mentions that transitioning training workloads to domestic chips could increase costs and time by roughly 50%. These figures illustrate a reallocation of market share and potential cost implications for Chinese AI developers, while Nvidia’s dominance in training remains despite reduced overall volume.

Why it matters

The development matters because it illustrates China’s concrete steps toward reducing reliance on foreign AI hardware amid tightening US export controls. Huawei’s claim of outselling Nvidia within China signals a growing domestic supply chain that could reshape procurement decisions for AI developers. However, the continued dependence on Nvidia for training large language models shows that software ecosystems like CUDA still lock in foreign technology where performance and cost are critical. The projected market‑share shift to about 50% for domestic chips suggests a long‑term rebalancing of the AI hardware market in China, potentially affecting global chip manufacturers and investors. Moreover, the Ulanqab park aims to accelerate inference deployments, which could speed up AI product rollouts domestically, while the higher cost of training on homegrown chips may slow the pace of frontier model development unless software alternatives narrow the performance gap.

DeFi Intel’s take

Editorial analysis, not a statement of fact.

DeFi Intel's take: the Ulanqab AI park signals that Beijing is moving from rhetoric to tangible infrastructure to back its semiconductor self‑sufficiency goals. We read this as a clear bet that volume‑driven inference markets can be captured with domestic chips, even as training workloads remain tethered to Nvidia’s CUDA ecosystem. The reported 50% domestic market share by 2026 suggests investors should watch for a gradual reallocation of AI‑related capital toward Chinese hardware firms, while still recognizing Nvidia’s lingering leverage in high‑end model training. The higher cost and time associated with switching training workloads may temper the speed of this transition, implying a bifurcated market where inference and training follow different supply‑chain dynamics.

Sources used for this analysis

DeFi Intel wrote this analysis from the reporting above. Any figure or quotation can be checked against those pages.

Frequently asked questions

Why does Huawei claim its Ascend chips have outsold Nvidia in China?

Huawei says its Ascend chips have outsold Nvidia within China as of October 2026, highlighting a focus on dependable domestic supply amid US export restrictions. The company argues that a chip that can be reliably purchased outweighs a faster one that may be harder to obtain.

What role does Nvidia’s CUDA platform play in China’s AI development?

Nvidia’s CUDA platform remains deeply entrenched in the AI industry, providing a software toolkit that developers use to write efficient code for Nvidia hardware. This entrenched software moat makes it costly and time‑consuming for Chinese firms to shift training workloads to domestic chips, with projected cost and time increases of about 50%.

How might the Ulanqab AI industrial park affect AI hardware markets?

The park is intended to boost domestic AI hardware production and inference deployment agility. By consolidating manufacturing and encouraging the use of Huawei’s Ascend chips, it could help domestic chipmakers capture roughly 50% of China’s AI accelerator market by 2026, while Nvidia’s share may fall to around 8%.

Direct answers

Short answers, written to be quoted.

What is the significance of the Ulanqab AI industrial park?
The Ulanqab AI industrial park marks China’s largest investment in AI hardware production, aiming to boost domestic chip usage and reduce reliance on foreign suppliers like Nvidia.
How does Huawei justify its Ascend chip sales success?
Huawei argues that reliable domestic supply of Ascend chips outweighs the higher performance of Nvidia chips that may be harder to obtain due to US export controls.
What challenges remain for Chinese AI labs using domestic chips?
Chinese AI labs still depend on Nvidia’s CUDA platform for training large language models, and switching to domestic chips for training could increase costs and time by about 50%.
Read full article on Crypto Briefing →

DeFi Intel is an entity-graph aggregator: we curate, tag and link crypto news to a typed knowledge graph of protocols, tokens, people and incidents. We do not republish the full article body. Use the link above to read the original report at Crypto Briefing.

Discussion

Comments are reviewed before publication.

No comments yet.

Stay current on entity-tagged crypto news

Get the weekly DeFi Intel brief — entity-graph intelligence covering protocols, tokens and incidents, free to your inbox.