For small manufacturers of DEPSEEK, DEPSeek and was not a threat, but a chance

DeepSeek shocked the world artificial intelligence market led by American companies-only the manufacturer of the NVIDIA and trace of NVIDIA has lost several hundred billions of dollars of capitalization. And while the market leaders are trying to overcome the consequences, small manufacturers see a chance to increase the scale of activity in what happened, CNBC writes.

Image source: Solen Feyissa / unsplash.com

«The developers really want to replace expensive and closed OpenA models with open source models such as Deepseek R1, ”said Andrew Feldman, general director of Cerebras Systems, producing chips for AI. The company acts as a competitor to NVIDIA and offers cloud services in its own clusters. The output of Deepseek R1 provoked one of the largest bursts of demand for the company’s services in its entire history, and, according to its head, showed that the growth of the AI ​​market will not be associated with the dominance of only one company, because open models are not attached to certain equipment or software. Deepseek claims that its reasoning model consumes less computing resources than American analogues and studies without advanced accelerators.

The Chinese startup is able to accelerate the process of deploying new technologies in the field of AI-accelerators, covering both the training of models and their launch. NVIDIA occupies a dominant position in the AI ​​training market, and many of its competitors believe that they have the opportunity to expand their presence in the field of launching already trained models, promising customers a higher efficiency for less money. Training of AI requires significant computing resources, but for the operation of an already trained system there is enough and less powerful equipment, limited by a narrower circle of tasks. And here, the developers of alternative accelerators note the growth of demand, because many customers are ready to solve their problems based on the already trained Deepseek models.

Analysts and industry experts are confident that the Chinese laboratory, which has lowered the bar on training and launching AI systems, will affect the development of the entire industry: if the launch services of already trained models will become cheaper, and the technology of AI will begin to be implemented more actively because the cost reduction leads to an increase in demand – This phenomenon is called Jevon’s paradox. Demand growth was confirmed by representatives of D-MATRIX and ETCED startup accelerators. “Thanks to the wide availability of small models, they served as a catalyst for the [II] output era,” D-Matrix said. “Companies transfer their costs from training clusters to output clusters,” added dozens of corporate customers from the moment of DeepSeek R1 had already addressed from the moment of the release of DeepSeek R1. Finally, it should be remembered that NVIDIA resources are not not true-even a technological giant of its scale will not physically satisfy the entire world demand for AI-accelerators. So, small players really have a chance.