A $50 billion AI market is up for grabs in China. That’s a significant piece of the global tech pie, and it’s where DeepSeek’s new AI model is making waves, presenting a unique challenge to Nvidia’s dominance.
As a reviewer focused on what truly works in AI toolkits, I pay close attention to the underlying hardware. The chips powering these models are crucial, determining performance, scalability, and ultimately, a toolkit’s real-world utility. DeepSeek’s approach, using chips from both Nvidia and Huawei, tells an interesting story about the evolving AI space, particularly in China.
DeepSeek’s Two-Chip Tactic
Recent reports indicate that DeepSeek’s upcoming AI model was trained on Nvidia’s most advanced AI chip series, Blackwell. This is notable because it shows DeepSeek using what many consider the industry standard for high-performance AI training. Nvidia’s chips are a known quantity for intense computational tasks, and their presence in DeepSeek’s training regimen suggests a push for top-tier capability.
However, the plot thickens. DeepSeek’s new AI model is also trained on Huawei’s chips. This dual-chip strategy is a clear signal. While tapping into Nvidia’s established strength, DeepSeek is also developing solutions with domestic hardware. This isn’t just a technical decision; it’s a strategic one, especially in the context of China’s growing AI market.
Nvidia’s Rising Stock Amidst New Competition
Despite this emerging competition from DeepSeek and its use of Huawei chips, Nvidia’s stock continues to rise. MarketWatch reported on April 24, 2026, that Nvidia’s stock was up 4.32%, even with the DeepSeek news circulating. This reflects strong performance and investor optimism surrounding Nvidia’s overall position in the AI market.
From an investor perspective, it’s easy to see why. Nvidia has been a dominant force, and their Blackwell series represents the latest in AI chip technology. Many firms continue to rely heavily on Nvidia for their AI infrastructure, contributing to the company’s solid financial standing.
The US-China Tech Tensions Echo in AI
This situation with DeepSeek highlights the ongoing US-China tensions that affect global tech markets. The fact that DeepSeek is training models on Huawei’s chips, in addition to Nvidia’s, isn’t just about technical preference. It reflects a broader geopolitical dynamic where countries aim to build self-sufficient tech ecosystems. For those of us evaluating AI toolkits, this means paying attention to the origins and dependencies of the hardware. A toolkit built on domestically sourced chips in China could offer different advantages and disadvantages than one built entirely on imported technology.
China’s AI market is valued at $50 billion, and it’s a fiercely contested arena. DeepSeek’s move to use Huawei chips can be seen as a direct challenge to Nvidia’s long-standing leadership in that specific space. It suggests an effort to foster domestic chip development and reduce reliance on foreign suppliers.
What This Means for AI Toolkits
For users and developers of AI toolkits, this competition is a net positive. It pushes both sides to innovate and refine their offerings. As DeepSeek develops models on Huawei’s architecture, we might see new optimizations and capabilities emerge that are tailored for those specific chips. Conversely, Nvidia will continue to push the boundaries of what their Blackwell series can do.
When I review AI toolkits, I’m always looking for efficiency and effectiveness. The emergence of strong contenders like DeepSeek, utilizing alternative hardware, means more choices and potentially better performance for specific use cases. It forces us to consider not just the model or the software, but the entire hardware-software stack. Will a toolkit perform better on a Huawei-backed system for certain tasks within China? Will Nvidia’s latest chips continue to be the universal standard for others?
The next few years will be telling. The competition between DeepSeek and Nvidia, fueled by both technological progress and geopolitical currents, will shape the future of AI infrastructure and, by extension, the toolkits we all use.
đź•’ Published: