The black swan theory or theory of black swan events is a metaphor that describes an event that comes as a surprise, has a major effect, and is often inappropriately rationalized after the fact with the benefit of hindsight. The term is based on a Latin expression which presumed that black swans did not exist. The expression was used until around 1697 when Dutch mariners saw black swans living in Australia. After this, the term was reinterpreted to mean an unforeseen and consequential event.
- From Wikipedia, the free encyclopedia
I believe DeepSeek is the Black Swan we never expected. While DeepSeek isn't new, their latest AI models have put Claude, OpenAI, and Meta on full alert. Not only does DeepSeek perform as well or better than frontier models, it was created a fraction of the cost, in fact as much as 95% less than OpenAI premier models. That alone would be enough to make waves in the industry, to take it further, DeepSeek is completely open, unlike what OpenAI's name implies. To make matters worse for OpenAI, DeepSeek has been very forthcoming about how they did it.
With Trump providing $500 billion dollar to advance AI infrastructure, DeepSeek releasing a superior model that only cost $5.6 million sent shock waves through the industry. Tech stocks went into a tail spin today as a result, with Nvidia (NVDA) dropping almost 17% bleeding almost $500 billion in market capitalization.
The US stock market has been doing amazingly well the last couple of years and this is largely due to Nvidia and other AI & Tech stocks performing extremely well with NVDA being a major driver of this.
OpenAI was founded 10 years ago, has 4,500 employees, and has raised $6.6 billion in capital. DeepSeek was founded less than 2 years ago, has 200 employees, and was developed for less than $10 million. How are these two companies now competitors?
- Adam Kobeissi - The Kobeissi Letter
Everyone assumed closed systems and massive funding was required to keep AI advancing, yet a small foreign company provided them wrote with just a "side project" of theirs. This also puts into question how much demand there really needs to be for high end GPU technology.
While I personally believe this is only a blip on radar, it will re-calibrate expectations for much of the industry and market. There is no questions others will try to accomplish the same thing DeepSeek has, and it will be much easier as they provide all their research. I truly believe this is the Black Swan event we never predicted.
To put in perspective how disrupted this technology is, look at the pricing comparison of OpenAI and DeepSeek.
At nearly 30x less expensive than OpenAI, DeepSeek not only performs as well or better, can be run completely on your own hardware without paying a third party API fees. You can also fine tune DeepSeek to be even more powerful while using smaller models.
An interesting tweet by an OpenAI employee and an epic community response.
This is such a big deal, I am considering building a system to run the full 671 billion parameter version of DeepSeek R1 locally. I am already running the 32B & 70B models.
Independent benchmarks have put it on par or better than OpenAI's top reasoning model in all categories.
This is only the beginning, hold on!
