Meta CEO evaluates implications of DeepSeek AI model
Mark Zuckerberg, CEO of Meta Platforms Inc., stated on Wednesday that it's still early to assess the impact of DeepSeek's innovative AI model on the company and the broader technology sector.
Zuckerberg expressed skepticism about the idea that Meta's overall AI investments will soon decline, despite DeepSeek's claims that advanced AI can be developed with significantly lower financial and computational costs.
"It's probably too early to really have a strong opinion on what this means for the trajectory around infrastructure and CapEx," Zuckerberg commented, noting the various simultaneous trends at play.
The technology market experienced turbulence when DeepSeek, an AI lab in Hangzhou linked to the Chinese hedge fund High-Flyer, announced that it developed its R1 large language model in just two months for less than $6 million. This revelation caused Nvidia's share price to plummet by 17% on Monday, suggesting that companies may not need as many of Nvidia's graphics processing units for their own AI initiatives.
During a call with analysts discussing Meta's fourth-quarter results, Zuckerberg mentioned DeepSeek, revealing that the company's sales grew by 21% YoY to $48.39 billion, surpassing Wall Street's expectations.
Meta's shares increased nearly 2% on Monday, indicating investor confidence that DeepSeek's cost-effectiveness could enable the social media giant to build powerful AI systems at lower costs. Last week, Meta announced plans to invest between $60 billion and $65 billion in AI efforts this year. The company also projected its total expenses for 2025 to range from $114 billion to $119 billion.
Zuckerberg acknowledged that Meta is still assessing some of DeepSeek's achievements and expressed hope for incorporating certain advancements into its own AI projects.
"That's part of the nature of how this works, whether it's a Chinese competitor or not," he commented.
Although it may become possible for companies to utilize fewer computational resources to train advanced models, Zuckerberg warned that substantial server resources might still be necessary when the software is actively running, which could lead to "a higher level of intelligence and a higher quality of service."
"It's going to be expensive for us to serve all of these people because we are serving a lot of people," he reiterated, noting that Meta reported 3.35 billion daily active users in the fourth quarter.
Meta plans to learn from technologies such as DeepSeek's and will continue to release open-source and free Llama AI models, which may also contribute positively to the industry.
"I continue to think that investing very heavily in CapEx and infra is going to be a strategic advantage over time," Zuckerberg asserted, adding that it is premature to conclude otherwise.
Zuckerberg also remarked that DeepSeek's rise validates Meta's open-source AI strategy, emphasizing the importance of establishing an American open-source standard.
"There's going to be an open-source standard globally," he stated. "For our own national advantage, it's important that it's an American standard."
