The sudden arrival of the groundbreaking AI model from DeepSeek, a Chinese AI startup, was angry from the Silicon Valley elite and brought out EN hope and praise. DeepSeek models, which provide abilities comparable to AI leaders, such as Openai, Meta (Meta), and Anthropic, have been developed at only a small part of the cost of the US counterpart.
The DeepSeek’s V3 model, which was released in December and has a new reasoning model earlier this month, was created only with 2,000 expensive graphic processing units (GPU) of NVIDIA (NVDA). Hedge fund high flyer last year. The company’s self -claiming ability to create advanced AI models for inexpensive SP proud investors betting on the American AI Giants. NVIDIA’s shares fell 17 % today (January 27). The stock of AMD, Alphabet (Googl), and Microsoft (MSFT) has also declined.
Some technicians are wondering if DeepSeek has more AI chips than they have allowed, so that China will be more from gaining new technology advantage. Some advocate severe GPU export management. Venture capitalist Mark and Leisen called DeepSeek’s break -through the modern “SPUTNIK moment.” Microsoft’s Satya Nadella praised the company’s efficiency, and Meta’s AI scientist Yann Lecun praised the open source.
One of the certain things is that Silicon Valley speaks through the development of DeepSeek. Let’s take a look at what a famous technician is talking about the amazing success of the startup.
MARC Andreessen, a co -founder of Andreessen Horowitz
Mark andrisen, a general partner of Venture Capital Company Andrisen Holowitz, explained the Deep Shek R1 reasoning model released on January 20. The Soviet Union’s launch of satellite in 1957 spurred the United States to strengthen space efforts. Deepseek-R1 said, “I’m one of the most surprising and most impressive breakthroughs I’ve ever seen,” and Andreessen praised the company’s decision to open a model as a “deep gift to the world.” Is added.
Yann Lecun, Meta Chief AI Scientist
According to Yann Lecun, DeepSeek’s performance does not necessarily talk about China’s AI ability, which is more than US AI ability, but rather emphasizes the power of open source. “The correct reading is that” open source models are more than their own models, “a computer scientist stated in a recent post on meta threads.
Lecun further called Deepseek to sell AI shares in response to “terrible and unfair things.” Many of the technology companies pay attention to being used to execute them, not always used to train models.
Anthropic CEO, DARIO AMODEI
For some technical leaders, DeepSeek’s breakthrough is a sign that the United States should continue to crack down on chip export control. DARIO Amodei recently told CNBC that Chinese AI companies have more GPUs than expected. “It has been reported that there are at least 50,000 clusters,” said Amodei.
According to human CEOs, export control needs to focus on preventing chips from growing to millions. “If the United States could not lead this technology, we would be in a very bad place in geopolitical,” Amday told CNBC.
Salesforce’s CEO, Marc Benioff
For Marc Benioff, DeepSeek’s new popularity proves that data, not calculations or models, will bring differences to future high -tech companies. “The real treasure of AI has become a product, not a user interface or model,” said Benioff. “The true value is oxygen that supplies fuel to the potential of AI.”
Microsoft’s CEO, Satya Nadella
Earlier this month, while talking to CNBC on Davos, Satiya Nadera praised China’s startup progress. “I think we should remove the development from China very seriously,” Nadela said, explaining the Deep Shek model as “super -sealed.”
CEO followed up today’s comments with X -posts that seem to be linked to DeepSeek to “Jevons Paradox”. “As AI is more efficient and accessible, it will be clear that the use will increase rapidly and it will be a product that is not enough,” he said.
