Full Breakdown
DeepSeek Unveils V4 Model on Huawei AI Chips Amid U.S.–China Tech Tensions
4/24/2026, 12:58:21 PM
DeepSeek Launches V4 Preview on Huawei Chip Platform
On 24 April 2026, Chinese AI startup DeepSeek released a preview of its V4 model, adapted to run on Huawei’s Ascend AI chip family. The preview includes a high-performance “pro” version and a lower-cost “flash” version. DeepSeek said the pro version surpasses other open-source models on world-knowledge benchmarks, trailing only Google’s Gemini-Pro-3.1. The company did not disclose the specific processors used for training and gave no timeline for a final release.
From Nvidia Dependence to Domestic Chip Strategy
DeepSeek’s earlier V3 model relied heavily on Nvidia GPUs, a reliance that attracted U.S. scrutiny after Washington imposed export controls on advanced AI chips in 2022. In response, Beijing accelerated development of domestic semiconductor capabilities, with Huawei’s Ascend line positioned as a cornerstone of China’s effort to reduce dependence on U.S. technology. The V4 preview marks the first publicly announced DeepSeek model built primarily for Huawei hardware.
Key Players and Stakeholders
- DeepSeek (Hangzhou-based AI startup) – developer of V4.
- Huawei Technologies – supplier of Ascend AI chips; announced full support for V4.
- Nvidia – provider of earlier GPU resources; subject of U.S. export restrictions.
- U.S. White House – accused China of large-scale IP theft in AI.
- Chinese Embassy in Washington – rejected the accusations.
- Anthropic and OpenAI – alleged improper “distilled” use of their proprietary models.
- Zhipu AI and MiniMax – domestic rivals whose shares fell after V4’s announcement.
- Alibaba and Tencent – reported as potential investors in DeepSeek’s upcoming funding round.
- High-Flyer Capital Management – owner of DeepSeek.
Data & Statistics
- V4’s pro version ranks second only to Google’s Gemini-Pro-3.1 among open-source competitors.
- Vals AI testing shows V4 leads open-source models in code-generation tasks.
- Zhipu AI shares dropped 9 % and MiniMax shares fell 7 % on the news.
- DeepSeek is targeting a valuation above $20 billion for its next financing round.
- Discussions are ongoing with Alibaba and Tencent for equity stakes.
Why It Matters: Implications for AI Self-Sufficiency and Competition
The V4 preview demonstrates China’s capacity to produce high-performance AI models on domestically sourced hardware, potentially narrowing the gap with leading U.S. systems. Success could bolster China’s strategic goal of semiconductor independence and influence the dynamics of the upcoming U.S.–China summit on technology and trade.
Official Statements & Responses
- Huawei: “The entire Ascend supernode product line now supports the DeepSeek V4 series models.”
- Chinese Embassy, Washington: Opposes “the baseless allegations,” adding that Beijing “attaches great importance to the protection of intellectual property rights.”
- DeepSeek: Reported that V4’s pro version outperforms other open-source models on world-knowledge benchmarks, though no direct quote was provided.
Criticism & Opposition
The White House has accused China of stealing intellectual property from U.S. AI labs on an industrial scale. Washington also alleges that DeepSeek violated export controls by acquiring cutting-edge Nvidia chips. Anthropic and OpenAI claim DeepSeek improperly “distilled” their proprietary models, raising concerns about compliance with U.S. sanctions.
Conflicting Reports & Gaps
DeepSeek did not specify which processors were used to train V4, leaving open whether any Nvidia chips were subject to export bans. The company also omitted a timeline for the final product launch, creating uncertainty about when the model will be commercially available.
Verbatim Quotes
- “the baseless allegations,” — Chinese Embassy, Washington
- “attaches great importance to the protection of intellectual property rights.” — Chinese Embassy, Washington
- “The entire Ascend supernode product line now supports the DeepSeek V4 series models,” — Huawei spokesperson
- “DeepSeek moment,” — description used by The New York Times
- “distilled” — term cited by Anthropic and OpenAI regarding DeepSeek’s use of proprietary models
What’s Next
DeepSeek plans to finalize V4 without a public schedule, while pursuing a fundraising round that could exceed a $20 billion valuation. Potential equity participation from Alibaba and Tencent may reshape the Chinese AI investment landscape. The upcoming U.S.–China summit will likely address the broader dispute over AI chip exports and intellectual-property protections.
