Chinese Startup Deepsec’s cheap AI is intensifying the investigation of billions of US tech veterans, putting technology to develop technology and analysts say the industry will dominate the much -awaited results of this week from Belavethers.
Deepsek has claimed that it costs only two months and $ 6 million to make AI models using Nvidia’s low-up-up H800 chips. An app operated by the V3 model became the top iPhone download in the US on Monday.
Startup established in 2023 has stated that its AI models either match or top the US rivals at a fraction of cost, challenging the approach that AI requires huge computing power and investment to score AI. Is.
The requirement of such a business has operated a growth of approximately $ 10 trillion in the market price of “luxurious seven” companies as the chat closed the AI ​​Boom in November 2022.
Stacey Rasgan, analyst from Bernstein, said, “Deepsek actually built Openai for $ 5 million?” It seems that the innovations being deployed by Deepsek seems like a stretch to think that the world. Top Tier AI in many other AI labs is completely unknown by researchers. “
The price of Deepsek removes anything from the competition, he said. The shares of AI Chip Pioneer Nvidia fell 16%, Microsoft 3.8%and TSMC’s US stock fell 14%.
Rasgon and other analysts argue that the training cost of Deepsek may be higher for its V3 model as approximately $ 6 million cited by the startup includes the amount spent on computing power, while the construction of the more publicized R1 model The costs for are very rarely known.
Nevertheless, it estimates an $ 250 billion analysts that large American cloud companies will spend on AI Infrastructure this year. In the last one year, that expenditure has been questioned by concerned investors concerned about slow returns.
Along with most American tech veterans to report this week and the next results, analysts and investors are expected to offer more clarity on their strategy.
“(The rise of Deepsak) asks whether the current speed of CAPEX expenses/technology upgradation is necessary. Comment from American hypersscalers will be important this week whether they are aggressive with AI expenses,” CFRA analyst Angelo Zino said.
“They probably insist on the need for more computing power because we shift to agents AI and physical AI,” Zeno said, referring to autonomous AI agents, which require little human intervention for regular tasks, Also robots and self-driving cars.
Price
While the price of using the AI ​​model is falling with rising competition and progress in technology, Bernstein’s Rasgon said that Deepsek is standing out because it gave its model at 40 times lower price than comparable models of OpenAII Is.
Analysts said, “Start a price war for AI services, potentially pressurizing technical companies such as OpenIAI who are already losing billions of dollars every year due to high operating costs of services such as chat.
“If the adoption of the deepsac increases, it can begin a decrease in value from the contestants, who have the same open source products,” Gadjo Sevilla said, the buildingcator’s senior analyst Gadjo Sevila said.
Market leaders such as “Openai (push for profitability) are unlikely to be low pricing in short term. They will be doubled as major differential features on trusts and security, which matter to enterprise users.”
Some experts also suspect that American businesses will be ready to embrace Chinese AI technology, with China-US stress and concerns about privacy and security.
Deepsek has stated that it stores user information in the server in China, which may be a glued point in its American adoption.
However, some investors believe that American technical giants will focus on Deepsek’s successes and are bound to adopt inexpensive AI services technology, which can raise the demand for chips.
“Did Deepsek seek and discovered a more efficient processing model for AI?
“(This) will only make AI opportunity bigger in the future.”
(Except for the headline, the story has not been edited by NDTV employees and is published by a syndicated feed.)