Underdog to Disruptor: DeepSeek-R1’s $10M Revolution That Rattled NVIDIA, OpenAI, and Wall Street | by Writerofeverything | The Capital | Jan, 2025
I know I said I would write about top 11 most expensive NFTs to be ever sold but this was important!
The markets bled and bled because we are all scared of the ‘C’ word! Let’s know why:
An underdog team of Chinese developers created an AI model called DeepSeek-R1 which is apparently a side project, and this fun side project suddenly becomes the talk of global circuits! And why is that?
- Wiped off Billions of dollars from the US Stock market
- If that was not enough it made crypto bleed!
- Dethroned ChatGPT from the Numero-Uno position on the App Store
DeepSeek has been around since November 2023 but the newly discovered bold DeepSeek-R1 claims to be a big competitor of ChatGPT. It is touted to be a reasoning model instead of old worn-out responsive models in the markets today. And the cherry on the top is that it is free of cost which is a straight head-to-head battle with ChatGPT’s 200$/ month reasoning subscription model. For those who are still unaware, a responsive model just responds to problems by scraping data off the net whereas the reasoning model works on the answers using their own step-by-step logic like how we humans do. The scare it has caused with its double-edged sword of reasoning and free model has shattered competition in the US market in one go. Here’s more:
- In September 2022, the US government banned the export of advanced Nvidia AI chips (like the A100 and H100 models) to China. These chips are crucial for training powerful AI models. To overcome this limitation, the DeepSeek-R1 team had to design their model to be highly efficient, allowing it to run on less powerful hardware readily available in China. Their focus on efficiency led to a surprising result: smaller versions of the DeepSeek-R1 model can now run on devices like iPhones, eliminating the need for internet connectivity or powerful computers. Check out this thinking UI which has DeepSeek-R1 models running on the iPhone — https://x.com/private_llm/status/1882029800279490856
- Took 2 months and less than $10M in total to develop and in many cases matches or even outperforms major US-based AI models.
Source: @joshliusg
- On top of all this, it is an open-source model!
Taking all this into consideration, how does that wipe off Billions?
Think of DeepSeek-R1 as a norm-breaker. It came and it broke all stereotypes associated with AI models. You went to NVIDIA for the AI chips, you went to OpenAI Pro because you wanted your AI model to be reasonable and not responsive, you opted for the new iPhone 16 Pro because it had Apple Intelligence. But DeepSeek — R1 comes and becomes the one-stop shop for all! It is free of cost, doesn’t need the NVIDIA chips, and can even run on Mobile devices!
But is that the only reason for the epitome of dismal performance?
Let’s also not forget that there have been other contributing factors like the Bank of Japan’s rate hike decision and to a certain extent Trump’s threatening gestures towards trade partners like Mexico, Canada, and China over the levy of 10–25% tariffs. Crypto is not immune to stock market movements and that sell pressure is finding its way to crypto.
But we all know that come what may crypto will weather this storm just like it has done in the past. You tell me how will this play out in the comments?
?? More dips expected in short term.
?? US tech corps will panic more.
?? Companies will put an insane amount of money into developing better and cost-effective AI models.
??Quiet recovery
Really hope it’s the last one!
Don’t forget the claps, commenting and the follows!
