Technology leaders react to the rapid rise of Deepseek

If you haven’t heard it yet: There is a new AI star in the city: DeepseekThe subsidiary of the Hong Kong-based quantitative analysis company High-Flyer Capital Management has triggered shock waves consistently Silicon Valley and the wide world With the publication of a new open source model for great argument at the beginning of this week, Deepseek R1What corresponds to the most powerful available Openai Model O1 – and at a fraction of the costs for users and the company itself (during training).

While the introduction of Deepseek R1 has already mixed a consistently turbulent, fast-moving and highly competitive market for new AI models, Openaai has had a competition with Anthropic and Google for the most powerful available proprietary models, while meta platforms often With “” close enough “to open source competitors-this time the difference is that the company is based in China behind the hot model, the geopolitical”Enemy”Up to this point, the United States, and its technology sector was widely regarded as that of Silicon Valley.

Therefore, there is no shortage of hand rings and existentialism on the part of the technology freaks from the USA and the Western Block, which suddenly doubt Openai and the general strategy of large technology companies, more money and more computing power (graphics processors, GPUs, normally the powerful gaming chips) Used to train AI models) to the problem of inventing increasingly powerful models.

Nevertheless, some western technology leaders have received a predominantly positive public reaction to the quick rise of Deepseek.

Marc Andreessen, co -inventor of the groundbreaking web browser “Mosaic”, co -founder of the browser company Netscape and current complementary complementary famous Andreessen Horowitz (A16Z) Risk capital society, Posted on X today: “Deepseek R1 is one of the most amazing and impressive breakthroughs that I have ever seen-and as an open source a profound gift to the world (robot emoji, Salut-emoji).”

Yann Lecun, chief AI scientist of the fundamental Ai Research (Fair) department of Meta, posted on his LinkedIn account:

“People who see and think about Deepseek’s performance:
“China exceeds the USA when it comes to AI.”
You read that wrong.
The correct reading is:
“Open source models outperform proprietary models.”

Deepseek has benefited from open research and open source (e.g. Pytorch and Llama from Meta).
They developed new ideas and built them up on the work of other people.
Since their work is published and open source, everyone can benefit from it.
That is the power of open research and open source. “

And even Mark “Zuck” Zuckerberg, founder and CEO of Meta Ai, seemed to want to face the rise of Deepseek with his own means Own contribution on Facebook promises that a new version of Facebook’s open source AI model family Llama would be “the leading model on the latest state of the art” if it will be released at some point this year. As he put it:

“”This will be a crucial year for AI. I assume that Meta Ai will be the leading assistant for more than a billion people in 2025, Llama 4 will be the leading model at the latest state of the art and that we will build a AI engineer that will begin, always To bring more code into our research and development efforts. To make this possible, Meta builds a 2GW+data center that is so large that it would cover a significant part of Manhattan. We will bring about 1 GW computing power online in 2025 and complete the year with more than 1.3 million GPUs. We are planning to invest $ 60 to $ 65 billion in property investments this year and at the same time significantly increase our AI teams, and we will have capital to continue investing in the coming years. This is a huge effort that will drive our core products and business in the coming years, release historical innovations and expand the American technology leadership. Let’s go building!“”

He even shared a graphic that shows the 2-gigawatt data center via Manhattan mentioned in his article:

Even if it works for open source AI, Zuck is obviously not convinced that Deepseek’s approach to optimize efficiency and at the same time use far less GPUs than large laboratories, for meta or for the future of AI the right one is.

But US companies take on records and/or issue record sums for them New AI infrastructure Many experts have found a rapid loss of value (due to hardware/chip and software progress). However, the question remains which future vision will prevail and the dominant AI provider will be for the world. Or may there be a variety of models, each with smaller market shares? Stay tuned because this competition becomes narrower and harder than ever.