301 - The Whale
DeepSeek has sparked big waves in the open-source and closed-source AI domains alike, by fully open-sourcing with a commercial use license, and by competing strongly with closed-source models that spent orders of magnitude more on compute. They've released a slew of material right before the Lunar New Year, which is a major time of traditional celebration with family for those unfamiliar "Western" countries.
It isn't a silver bullet of a model, but the consensus among AI practitioners and developers is currently that it competes well overall with closed-source options, while being far more flexible via open-sourcing. It has major advantages to cost efficiency, open-sourced modifiability, and reduced "guardrails" (aka fraud) compared to others. Whatever flaws they may have, they at least don't appear to have committed clear-cut fraud like OpenAI and Anthropic have consistently demonstrated.
Some have pointed to DeepSeek's association with a Hedge Fund, which is assumed to be busy shorting Nvidia and/or similar stocks that promptly began taking a sharp downward trajectory upon the spread of this news. Whether or not they are indeed shorting those stocks, which if they are could easily fund their next 10+ models, this illustrates a much more significant opportunity for such Hedge Funds than anything that DeepSeek can tap.
DeepSeek is still fundamentally based on GPUs, and even demonstrating notably greater GPU efficiency has triggered a major downturn in US tech stocks, particularly Nvidia. At the time of writing Nvidia's stock is down by -10.56% YTD.
The major opportunity is that the vastly more powerful AI technology than simple systems like LLMs and RL, which has been around and under the radar for a few years now, isn't based on GPUs. The company behind such technology can entirely and explicitly tell Nvidia to "Go %^&* themselves", if they so choose, unlike any of these other AI companies. When the market becomes fully aware of that far more potent technology, it may predictably cause far more severe losses to companies like Nvidia than we've recently seen from the news of DeepSeek.
When that time comes, either Hedge Funds will be positioned and ready to profit from it, or it will be a multi-trillion-dollar missed opportunity. Hedge Funds could fund that far more potent technology, if they were smart, giving them a precise ticking clock for when the news breaks. The "Big 7" tech giants have invested heavily in trash technology, and ballooned by trillions of USD in market cap since January 2023, and the so-called "biggest bet in history" may have some equally big winners and losers when the bubble pops.
If nobody predicts when it will pop correctly, then everyone loses, but if one group is smart enough to back viable technology and know or control when the news breaks, then they'll capture that pivotal moment that makes or breaks the future of companies and nations alike.
