The stock price has hit record highs, created the longest consecutive rise in history, and the market value has equaled that of Saudi Aramco. Perhaps Zuckerberg himself may not have imagined that Meta would become the "biggest winner" in the U.S. stock market as China's new AI force DeepSeek triggered a global wave...Market data shows that Meta's stock price closed up another 1% on Tuesday, closing at $704.19, closing above the $700 mark for the first time in history.

What is more eye-catching than the new high of the stock price is perhaps that Meta has risen for 12 consecutive trading days, creating the longest continuous rising record in the history of the stock. This is obviously particularly rare in the context of the recent turmoil in the stock prices of U.S. technology stocks.


According to statistics, during this period, Meta's market value has increased by approximately US$240 billion, and its current total market value has reached nearly US$1.8 trillion, almost tying the Saudi energy giant Saudi Aramco.


Many industry insiders say that the recent continued rise in Meta’s stock price seems to have caught a ride on the craze triggered by DeepSeek, a new Chinese AI force. The biggest similarity between the two is that Meta's Llama series of artificial intelligence models, like DeepSeek, are open source.

CFRA Research analyst Angelo Zino said that Meta is the only company among the "Big Seven" that focuses on the open source model and looks very good. "Ultimately, the DeepSeek model should speed up the development of artificial intelligence products and hopefully lead to a faster return on investment for Meta."

However, what is quite interesting is that in essence, although they are both open source models, there are still huge differences between Llama and DeepSeek.

The most representative one is the cost between the two. In fact, just after DeepSeek released two large models, DeepSeek-V3 and DeepSeek-R1, in late last month and triggered heated discussions across the US Internet, the first "earthquake" in Silicon Valley reported by the media was panic within Meta, and engineers began to try to replicate DeepSeek's results overnight.

On the American anonymous workplace forum TeamBlind, a Meta company employee posted at the time that since DeepSeek released DeepSeek-V3, Meta's Llama4 has fallen behind in various tests. What's worse is that "the Chinese company only spent US$5.5 million on this (pre-training costs)." It is reported that Meta has currently established four specialized research groups to study the working principle of DeepSeek and improve its large model Llama based on this.

Therefore, even if it has the label of being an open source model, it still seems to remain to be seen whether Meta can benefit from the wave of DeepSeek for a long time.

Meta CEO Mark Zuckerberg said last week that this is a "very important year" for Meta. He expects the company's AI assistant to become the most widely used in the industry. The Menlo Park, Calif.-based company also plans to spend up to $65 billion in AI-related investments in 2025.

Meta announced its latest Q4 performance report at the end of last month. The financial report shows that Meta’s sales and profits in the quarter hit record highs, reflecting the application of AI technology to enhance the digital advertising business. However, Meta's latest performance guidance is not outstanding-revenue in the first quarter of 2025 is expected to be US$39.5 billion to US$41.8 billion, while the average of analysts' previous expectations was at the high end of the guidance range (US$41.67 billion). At the same time, Meta also failed to provide revenue guidance for this year.