DeepSeek can be a force of a force for small AI chip firms

Deepseleek voiced the United States with the latest model of the LED EC Ecosystem, shaver hundreds of billions of chip leaders Nvidia the market cover. While the sector leaders are fighting, small AI companies see a scale opportunity with Chinese start.

Firms related to several AIs, the threat of Deepseek’s emergence to CNBC, he said, “Massive” for them.

“Developers are eager to change expensive and closed models with open source models with open source models such as DeepSeek R1

The company competes with NVIDIA graphic refining sections and offers cloud-based services through their own computing groups. Feldman said he created one of the largest spikes in the demand for the release of the R1 model.

“R1 shows that (AI market) growth will not dominate by a single company – hardware and software Moats are not available for open source models,” Feldman said.

The open source refers to the program that the source code is freely available on the Internet to change possible change and redistribution. DeepSeek models are open source unlike opponents such as Openai.

DeepSeek also has the best American technology and competitors, although the R1 is the best valuable and best-valued graphical processing sections Questioned these allegations.

“As in PC and Internet markets, the falling prices help the fuel global adoption. The AI ​​market is a similar world-to-growth,” Feldman said.

NVIDIA takes a dominant position in GPU used for AI education, many opponents see Expanding room In the “Inference” segment, where they promise higher efficiency for low costs.

AI education is highly calculated, but programmed to perform a result to perform a range, added Lee, programmed to perform less powerful chips.

AI chip starts told CNBC that they stated that they had made more demand for computing and calculation, as it was adopted and established by customers.

“(DeepSeek) can be trained in smaller open models to be more skilled or more skilled than larger property models, and in one part of the value of it,” said, AI Chip General Director Matrix.

“The CNBC, which results in a wide range of small-skilled models, told CNBC that the company recently saw the interests of global customers to accelerate instability plans.

Ai Chipmaker’s co-founder and coo, Coo, co-founder, Robert Wachen, Deepseek’s substantiation models have reached the beginning of the company.

“Companies (now) are spending their teachings to the majority groups,” he said.

“DeepSeek-R1, the inference time calculation is now a great model seller (the most modern) in the seller and has proven that there is no cheapness in mind.”

“Deepseek’s performance is based on a number of engineering innovations based on a number of engineering innovations, while improving teaching costs, improving the cost of learning,” report From Bain & Company.

“In the bullish scenario, the improvement of the ongoing efficiency will lead to a cheaper, the larger AI adoption,” he said.

This pattern tells Jevon’s paradox, a requirement that requires reducing the cost of a new technology.

Financial Services and Investment Company Wedbush, last week’s research notes, said they expect to be requested to be required for demand and retail consumers.

Talker CNBC’s “fast money” Last week, the Gunashli madra, preparing a chubby chip for the AI ​​inference, the small demand grew up, the smaller demand suggested to have more rooms to grow smaller players.

“The world will need more verses (a data unit of the AI ​​model) NVIDIA can’t provide enough chips, so we will sell more aggressively to the market,” he said.