Imagine you bought a mobile phone for $2,000, and your friend bought the same mobile for $200 and enjoyed the same quality mobile phone cheaply. You will also be shocked to hear this.
Recently, something similar happened to major American technology companies when DeepSeek, an artificial intelligence application developed by China, woke up OpenAI and its related major tech companies and investors.
According to Bloomberg, American Wall Street suffered a loss of $1 trillion and the reason for this was that DeepSeek launched its new artificial intelligence R1 model.
While other companies, including the American company OpenAI’s chatbot ‘ChatGPT’, invested several hundred billion dollars in the world of artificial intelligence, a Chinese company made a chatbot for just $6 million and made it free for the public.
According to DeepSeek, their chatbot can easily find solutions to extremely difficult questions and its performance is similar to ChatGPT, but the cost is very low and users can use all its features without paying anything.
DeepSeek’s R1 is also an open-source model, meaning other companies can access its code and create alternative artificial intelligence models.
This is why DeepSeek has surpassed ChatGPT in a week to become the top free application on Apple’s App Store.
It is worth noting that DeepSeek came to the fore when US President Donald Trump announced the $500 billion ‘Stargate AI project last week with the participation of major companies in artificial intelligence.
The conventional wisdom in the technology world is that artificial intelligence companies need expensive and sophisticated computer chips to train their systems, which can provide more computing power and carry the burden of technology.
While these companies need the best computing chips, Nvidia, a computing chip manufacturer, seems to be at the forefront of meeting these needs.
However, the arrival of DeepSeek shattered the perception that success in AI required expensive chips and technology, and this had a direct impact on Nvidia’s value.
recently the US stock market saw its worst-ever decline, and Nvidia’s value fell by $600 billion.
The stock market suffered a loss of $1 trillion, and the stock value of big names in the AI industry such as Microsoft, Google, Meta, and others fell.
Not only that, a total of $108 billion was also taken out of the pockets of America’s tech billionaires.
What was Nvidia’s response?
In a statement to Bloomberg, a Nvidia spokesperson said that DeepSeekk is an advancement in the world of AI.
The spokesperson said, “This shows how a company can build new AI models using a test time scaling approach, and that too while taking advantage of widely available models and computer that is completely export controlled.”
It should be noted that Nvidia is the company that suffered the most when DeepSeek came along and saw its shares fall by 17%.
What did Donald Trump say?
A few hours ago, US President Donald Trump warned the country’s technology companies that our industry should wake up after DeepSeek.
Trump said that we need to pay more attention now. This could be a positive thing for us, we could come up with a similar solution at a lower cost instead of spending billions of dollars.
Response from Sam Altman, owner of ChatGPT;
Sam Altman, founder of OpenAI, said that the arrival of a new competitor in the market is an encouragement for us.Sam Altman called DeepSeek a great model, especially because of its low cost.
What did DeepSeek do that Google, Meta, and OpenAI couldn’t?
One thing that DeepSeek’s arrival showed is that necessity is still the mother of invention. DeepSeek’s ‘fractional computer’ intensity compared to American companies is the result of significant progress.
According to research by scholars of Large Language Models (LLM), DeepSeek’s model does not fire on all cylinders all the time, but rather at different stages of the computer cycle. It predicts which part of the model will need to be activated and thus keeps the computational burden to a minimum.
This is completely different from the LLMs of Meta, Google, and OpenAI. Safe computational output is a significant development of DeepSeek, made possible by ‘KV cache compression’. This means that their engineers were able to fit more information into limited memory applications, and therefore the cost of DeepSeek was much lower.
Another report revealed that DeepSeek used Nvidia’s H100 chips, which were allowed to be exported to China from the US until October 2023. After the ban, these cheaper chips were combined with other lighter chips to produce the technology at a lower cost.
It should be noted that US microchip export controls were designed to freeze China’s development of supercomputers that are used to develop nuclear weapons and artificial intelligence systems.
Who invented DeepSeek?
DeepSec, based in Hangzhou, China, was founded in late 2023 by Liang Wenfeng, an entrepreneur who also runs a hedge fund called HighFlyer.
Although Liang is little known outside of China, he has an extensive history in the combination of growing technologies and investments.
In 2015, Liang Wenfeng founded a Chinese quantitative hedge fund called HighFlyer. According to a report in the Wall Street Journal, quantitative or ‘quant’ hedge funds rely on trading algorithms and statistical models to find patterns in the market and automatically buy or sell stocks.
To analyze financial data and support complex operations, Liang set up a research branch under HighFlyer called Fire-Flyer and began assembling graphics processing units (GPUs) to build supercomputers.
Liang Wenfeng stockpiled Nvidia’s H100 chips, and the key point here is that China can still export cheap chips despite the ban on chip imports.
In 2023, Liang decided to redirect the fund’s resources into a new company called DeepSeek, to develop foundational AI models and ultimately crack artificial general intelligence (AGI).
Instead of hiring experienced engineers who knew how to build AI products for consumers, Liang recruited PhD students from top Chinese universities to be part of DeepSeek’s research team, according to a report by Chinese tech news site QBitAI.
DeepSeek faces cyberattacks;
According to the British news agency Reuters, DeepSeek announced on Monday that it was halting registrations on its application, citing large-scale cyberattacks.
