DeepSeek’s delayed V4 model and a new U.S. State Department warning are now shaping the same story. On one hand, the Chinese AI firm is tuning its model to run on local chips from Huawei Technologies. On the other hand, the U.S. is raising alarms about Chinese firms using distillation to copy American AI systems. Together, these moves point to a deeper shift in how global AI is built and controlled.AVGO
Claim 30% Off TipRanks
New trading tool for NVDA bearsChina Moves Closer to a Self-Reliant AI Stack
DeepSeek’s delay of its V4 model was not due to slow progress. Instead, the company spent months reworking its software to better fit Huawei’s Ascend chips. According to Yuyuantantian, a state-linked media account, the goal was to move “beyond basic compatibility toward hardware-specific tuning.”
In fact, it signals a change in strategy. Rather than relying on foreign chips, Chinese firms are now building models designed for local hardware from the start. This aligns with Beijing’s push for more control across the AI supply chain.
At the same time, DeepSeek has released preview versions of its new model and called it its most powerful open-source platform to date. The company aims to compete with firms like OpenAI and Anthropic, but with a focus on lower cost and wider access.
Industry analysts cited in the report said that more models are now trained and run on domestic chips.
The U.S. Responds With New Warnings
Meanwhile, according to a Reuters report, the U.S. is stepping up its response. A State Department cable sent to global posts warns about “concerns over adversaries’ extraction and distillation of U.S. A.I. models.”
Distillation is a method where smaller models learn from larger ones. It can lower costs and speed up development. However, U.S. officials argue that it may also allow firms to copy proprietary systems.
The cable states that such efforts can lead to models that “appear to perform comparably on select benchmarks at a fraction of the cost but do not replicate the full performance of the original system.” It also claims these models may remove safety features built into the original versions.
OpenAI has already told lawmakers that DeepSeek may be targeting U.S. firms to replicate their models. In response, the Chinese Embassy has rejected these claims, calling them “groundless” and a way to slow China’s progress.
As a result, some Western governments have restricted the use of DeepSeek models due to data and security concerns. Still, the company’s tools remain widely used on global platforms.
Two Paths for Global AI
Taken together, these developments show a growing split in the AI market. China is building a system based on local chips and open models, while the U.S. is focusing on high-end performance and tighter control.
This shift could have long-term effects for companies like Nvidia (NVDA), Advanced Micro Devices (AMD), and Taiwan Semiconductor Manufacturing Company (TSM). These firms sit at the center of the global chip supply chain and may face new risks as China builds its own ecosystem.
In the end, the story circles around model scaling and policy shifts. We’re seeing how two major economies are shaping the future of AI in different ways, with clear impacts on cost, access, and control.
We used TipRanks’ Comparison Tool to align notable Chip manufacturers’ stocks. It’s a great tool to gain an in-depth view of each stock and the broader chip industry.
