Deepseek may be a multiplier force for small chips for small chips

Deepseee has squeezed the US AI ecosystem with its last model, shaving hundreds of billion books Nvidia’s Market cap. While sectoral leaders associate with the fall, they see the opportunity to scale the AI ​​small companies with Chinese launch.

AI Several related companies CNBC said that Deepseeek’s creation is a “massive” option for them, rather than a threat.

“Developers are very enthusiastic to replace Openi’s open models with Open models like R1 …” said Andrew Feldman, General Manager of Artificial Intelligence Chip Cerebras Systems.

The company competes with NVIDI graphic treatment units and provides cloud-based services through computer clusters. Feldman said that releasing R1 models created one of the largest demands of their services.

“R1 (AI Market) growth will prevail in a single company – hardware and software beaks do not exist for open source models,” Feldman added.

The open source source code is freely available on the web. Deepseeek’s models are open sources, unlike competitors like Openai.

Deepsee also claims its R1 reasoning model the best American technology, despite running at lower costs, despite being trained without graphic-cutting-edge processing, despite industry supervisors and competitors These statements questioned.

“As in PC and Internet markets, the prices help to fall on the global fuel adoption. The AI ​​market is located in a secular path of growth,” Feldman said.

NVIDIA has a main position of GPU used for AI training, many contestants see Spreading place In the “inference” segment, where they promise to achieve higher costs.

AI training is a very computing, but inferences can function with less powerful chips programmed to make a tighter range of tasks, LEE added.

Several KNBC’s initial KNBC said the CNBC allowed the support and deepening the inferenced chips and computer clients as Deepsees builds on the open source model.

“(Deepsee) has shown that smaller models may be able to be able to or be able to more than the larger models of proprietary,” said Sid Sheth, AI chip Start-up Director General. Matrix.

“Extensive use of small models, CNBC said, CNBC said that the company has just added interest in global customers to speed up their inference plans.

Robert Wachen, ai chipmaker creator and coorden, said, dozens of companies have launched since Deepse released his reasoning patterns.

“Companies (now) are changing clusters of the inference clusters who train their expenses,” he said.

“Deepseeek-R1 has been shown that the computation time computation today (leading) model and thinking is not cheap – we will need a greater capacity for these models and more computing users.”

“Deepseeek’s performance is based on an engineering series of innovation, which significantly reduce inference costs by” depending on “ inform From Bain & Company.

“In a Bullish scenario, constant improvements would lead to a cheaper inference, they would promote a greater adoption,” he added.

This model explains Jevon’s paradox that reducing the costs of a new technology has increased demand.

The financial services and research release carried out by the Wedbush investment company last week, AI companies and retail consumers continue to promote the use of consumers worldwide.

By talking CNBC’s “Quick Money” Last week, sunny Madra, Groq’s coo, who develops french fries for inference, suggested that the general demand of Air, will have more space to grow smaller players.

“The world needs more tokens (Data units with AI models) NVIDIA can’t give enough chips to all, so it allows us to sell it in the market more aggressively,” Madra said. “