The AI race will consume more money in 2025 than ever before. The major CEOs of Tech are currently launching around the billions of dollars they wish to invest in AI infrastructure (mainly in GPUs). The following has been announced so far:
- OPENAI & SOFTBANK: $ 100 billion
- Microsoft: $ 80 billion
- Meta: 60 to 65 billion dollars
- Bytedance: $ 20 billion
- Google: TBA.
- NVIDIA: TBA.
- Amazon: TBA.
- XAI: TBA.
So there are huge sums of money at stake, and Trump wants to obtain the domination of the United States for a long time. An export ban on AI fleas has already been introduced under Biden, which has been extended to include the “Export control framework for the dissemination of artificial intelligence“. The latter means that only us, the allies, will be able to obtain Chips Ai without restrictions.
And then the Chinese startup In depth comes and returns everything upside down. First of all, the open source Deepseek V3 model was tied with the best models of big names like Anthropic, Xai, Meta, O1.ai and Alibaba, then they added the bomb: Deepseek R1. The reasoning model (these models can reproduce longer reflection processes) is in some cases better in references than Optaai O1, which is the strongest model to date of the best funded IA startup in the world ::
Deepseek reaches the same level as Openai and Google
These (better) Deepseek performances, which were published until January 20, 2025 (a week ago), are obvious not only in the laboratory but also in real life. In the chatbot arena, where real users assess the responses of AI, Deepseek R1 models is now almost equal to the best IA models to date, namely GPT-04 of Openai and Gemini from Google, and Even in front (!) Of Openai O1Claude 3.5 SONNET, META LLAMA 3.1, GROK 2 of XAI and many, many other AI models of the largest technological companies in the industry.
But Deepseek also currently conquers the consumer market. There is, of course, a chatbot which was built on the basis of V3 and R1. The smartphone application is already in second place in productivity graphs in the Apple App Store In the United States, behind Chatgpt, and whoever looks regularly Tiktok will have already noticed the wave of influencers who are completely excited by Deepseek:
The impact of Deepseek R1 on January 20 has the potential to descend in the history of AI as the launch of Chatgpt in November 2022. What is revolutionary about the model, is that, according to the developers (the AI startup belongs to the Chinese hedge fund Huàn Fāng (“Haut-Vol”)), It only costs $ 5.6 million has Develop – a fraction of what other AI models of American companies cost. Their development is generally estimated between $ 100 and $ 1 billion.
At the same time, Deepseek is offered under MIT license, which means that in principle, each developer and each company can download it for commercial purposes. Anyone who wants to get Deepseek via the API can get it at dumping prices. Deepseek V3 costs a tenth to fiftieth from what Openai, Anthropic, Xai or Google charge their API AI. The disadvantage is that whoever buys the computing power of the AI directly from the The Chinese company obtains it with censorship.
In the Silicon Valley, where the Big Tech and the great AI startups are based, all the people involved in the AI have now woken up. “”The AI supremacy race is over, at least for the moment, and the United States has not won. In recent weeks, two Chinese companies have published three impressive articles that destroy any assertion that the United States has a decisive lead,“Said the AI scientist and entrepreneur Gary Marcus.
Deepseek in the application version also has a censorship problem, of course. Before our eyes, a factual response on the situation of the Uighur minority oppressed in China was removed by the Deepseek chatbot. Here is the report on trendy subjects:
Deepseek: This is what live censorship looks like in the Chinese AI chatbot
“A warning drawn through the bow of the current AI hypersaling approach”
“The fact that Deepseek works on basic products and equipment without attachment and is the open source is a warning in the prow of the current approach to IA hypeling”, “ analyzes Steven Sinofsky, a former high -ranking Microsoft leader who is now working for the main VC of Silicon Valley Andreessen Horowitz.
The first in -depth effects are already apparent:
- Meta: The social media giant, which was based on “open source” with Llama, A facility 4 “war rooms” to discover how Deepkseek was able to reduce training costs so much and with what data the models have been formed
- AI stocks: On the Asian scholarships, which already opened its doors on Monday morning, the prices of companies related to AI such as Disco Corp, Advantest (an NVIDIA partner) or SMIC (the main manufacturer of AI of China) have decreased considerably Between 2.6 and 10%
- Deepseek application In the United States, in the United States, Deepseek has already reached number 1 in the productivity graphs of the Apple Apple App Store App while writing this article
When the stock markets opened on Monday, it will be exciting to see how hyper-specials actions will react to the Deepseek effect.