After the inauguration of President Trump in the USA, the tech sector of the country seemed pretty much unstoppable. A number of tech billionaires were present not only at his inaugural ceremony but also seated ahead of some of Trump’s own cabinet picks. These included Elon Musk, CEO of Tesla, X and SpaceX, Sam Altman, CEO of OpenAI, Jeff Bezos, Chairman of Amazon, Mark Zuckerberg CEO of Meta and Tim Cook CEO of Apple amongst many others.

deepseek us tech bubble

The scene was straight out of a tech fairytale. The biggest tech names in the world were all grouped together under one roof, giving the impression that tech in the country had nowhere to go but up.

This belief has since been busted. And by a simple, humble Chinese AI service that has caused more harm to the US stock market than any other since its inception.

Anyone with an Internet connection would have heard of the phenomenon that is DeepSeek. It is an AI model that has caused disruption in the US market, like a true Chinese product. However, this case is especially worth noticing because no other event has managed to wipe out over USD 1 trillion off the US market. Chipset manufacturer Nvidia was hit the worst and lost over USD 600 billion, the worst hit in US market history, while Google’s parent company, Alphabet, lost USD 100 billion and Microsoft USD 7 billion.

All it took was a single large language model to cause all this destruction.

But what is so special about DeepSeek, and how did the platform manage to bring the mighty US tech market to its knees? Diving a little into the history of the platform might give us an answer.

Born out of a hedge fund founded by engineers from Zhejiang University, DeepSeek was initially supposed to create an architecture to build artificial general intelligence (AGI). Headed by CEO Liang Wenfeng, the goal of DeepSeek was to “explore the essence of AGI.” Keeping this in mind, Liang started buying thousands of Nvidia chips in 2021, evidently as a hobby to develop AI models. Instead of building their model from scratch, DeepSeek enjoyed a late movers’ advantage in the AI sector and used existing open-source AI models as a starting point. In December, OpenAI’s CEO Sam Altman posted on X, “It is (relatively) easy to copy something that you know works. It is extremely hard to do something new, risky, and difficult when you don’t know if it will work.” Many considered this as a dig at DeepSeek.

deepseek nvidia
Image: Forbes

But no amount of digs and disses can change what DeepSeek has managed to do.

Before DeepSeek became the sensation it is today, it was believed that developing large language models was no easy feat and cost billions of dollars every single year. DeepSeek entered the AI market and flipped that concept on its head. Its R1 model cost a total of USD 5.6 million to get it up and running. This raised eyebrows, questions, and doubts over billions of dollars that the American tech was pouring into developing and running existing AI models.

It is not just DeepSeek’s cost that is a fraction of the price of its rivals; its running costs are incredibly economical, too. To give you an idea, DeepSeek’s R-1 model takes USD 0.55 per million tokens (input) to run, while the OpenAI model o1 takes USD 7.50 for the same process. And even though DeepSeek is monumentally cheaper, it actually performs much better. In a series of tests, the platform outperformed OpenAI in areas like mathematics, language, and coding. In addition to all of this, DeepSeek is an open-source platform, unlike Open AI.

All of these factors combined mean DeepSeek is not only more accessible and affordable but will also act as a base for developers to work on.

deepseek opinion

Even if it is not “original,” it is still incredible to see how a brand has managed to overthrow all that people thought they knew about AI models. This is especially impressive given that DeepSeek had to work with much less powerful H800 chips Nvidia specifically created for the Chinese market after Nvidia’s most advanced H100 chips were banned from being exported to China in September 2022 because of US sanctions. The US later banned the export of H800 chips in October of last year. Ironically, this basically led to DeepSeek innovating and tweaking their AI models and creating a platform that not only requires exponentially less funds to develop but is also easy on the pocket of users.

This is not an isolated occurrence either by a small hedge fund that managed to crack a code by chance. There are other documented Chinese companies that have managed to cut AI models’ development costs dramatically. As we write this, Chinese conglomerate Alibaba has also entered the AI race and claims that its AI model performs much better than all existing AI platforms, including DeepSeek.

OpenAI’s CEO Sam Altman recently tweeted, “Deepseek’s R1 is an impressive model, particularly around what they’re able to deliver for the price. We will obviously deliver much better models, and also, it’s legitimately invigorating to have a new competitor! we will pull up some releases.” While this is a very optimistic way to approach a problem that threatens the monopoly that Altman’s OpenAI had created in the world of AI, it very obviously looks like a very difficult problem to deal with.

This also overthrows the idea of AI’s future that many had in mind and the certainty of the US being the frontrunner in the AI race. All of a sudden, a panic has set in the US tech space, taking away the very stranglehold it had over the AI segment. The democratization of AI also means the ethical boundaries around the subject will be pushed, and while the dangers of AI misuse and its environmental implications were already hovering over us before DeepSeek made AI easy and accessible for everyone, it will now also move up a notch. Italy has already blocked the service on Apple and Google’s app stores, and this stems from the similar fear that the US has against TikTok – user data is stored in servers in China, which many fear can be misused and the Chinese government can push its agenda through these platforms.

Are more bans on the horizon? We do not know.

What we do know is that the future of AI is not confined to just the tech bros of the US anymore. In a supreme irony, AI might have just democratized thanks to the efforts of a company in a Communist nation.

Was this article helpful?
YesNo