.jpg?w=1024)
DeepSeek made its name with low-cost AI models. Its next reported bet is the processor needed to run them.
Reuters reported that the Chinese startup is developing its own AI chip, a move that could reduce its reliance on Nvidia and Huawei. Rising AI spending across China is pushing companies to seek homegrown hardware as US restrictions limit access to foreign chips.
Businesses weighing Chinese AI tools now face a bigger question than model performance. Can providers keep prices steady and service levels reliable when small pilots become daily work?
What we know about the chip plan
Reuters says DeepSeek is focusing on the stage after model training, when an AI service responds to prompts and handles user requests. Training new models from scratch is outside the chip’s reported purpose.
Only fragments of the plan are public. The Hangzhou-based company has spoken with potential partners in chip design and manufacturing, as well as memory suppliers, people familiar with the work told the outlet.
Recruitment offers another sign of movement. Engineers with chip-design experience have been hired privately in recent months, though the company did not respond to a request for comment. Success could expand the startup’s role in China’s AI market. Known for its efficient models, the company could gain greater control over the hardware required to keep its services running at scale.
Chinese buyers turn to local chips
DeepSeek’s reported chip plan would enter a market where buyers are already budgeting for more homegrown AI hardware.
Bloomberg Intelligence found that Chinese executives expect local AI chips to account for 46% of AI chip budgets over the next year, up from 30% today. Eight in 10 surveyed executives said spending on computing infrastructure is already over budget, mostly because of AI projects.
AI use can quickly turn chip access into a budget problem. More usage means more computing capacity and higher data center costs. Export controls add more pressure, with US restrictions limiting access to Nvidia’s most advanced chips and Beijing urging companies to use domestic technology.
Service gains would take time
Chinese companies testing DeepSeek-based tools should look beyond model scores. A chatbot used by a few teams is one thing; a service handling customer support queues, coding requests, and companywide document searches is another.
Capacity problems often appear in ordinary ways. Replies slow down. Access gets limited. Usage costs climb.
Most customers will not know who designed the processor behind an AI tool. They will judge the service by whether it stays available during busy periods and whether pricing holds up as more employees use it.
Any gain from DeepSeek’s reported chip work remains uncertain. Moving from design work to production would require access to manufacturing, sufficient memory supply, and years of capital investment.
Until the Chinese AI company shows a working chip or names a production path, companies using its tools have little reason to change plans. Service quality and cost remain more effective measures.
Also read: Alibaba is blocking Claude Code after undocumented China-linked checks raised privacy and enterprise security concerns.