They said they did it at a fraction of the cost. They said the whole training thing cost them $6 million. Companies in the US have spent billions of dollars building the same models.

So in its current form, DeepSeek breaks a few big assumptions:

Maybe you don’t need a lot of computation to build and train an AI model
Maybe you don’t need a lot of money to build and train an AI model
OpenAI, Microsoft, Google, Amazon are all paying way too much to build and train their models. Bad news!
Fewer chip sales are also hurting Nvidia’s sales projections. Really bad news!
That explains why tech stocks crashed a few days ago. Because it could set a new precedent for the future.

But that’s one version of the story.

Let’s take another look at the report, but this time with a little more nuance.

OK! So let’s say China has found a way to build and train a basic model for very little money. Great! So what’s stopping other, better-funded players from doing the same thing? Why can’t they take the same approach, reduce costs, and make big profits while improving their models?

Yes, they can, and it’s likely someone will do it in the future.

That’s something investors should think about.

But that doesn’t mean everything in this story is spotless. The researchers’ claims that they used limited computational resources to train a basic model are just that, claims. It’s fair to say that there are academics who have reviewed the research papers and tried to replicate the results. We’ll probably find out for sure very soon whether the model is as cheap as they claim.

But even if we take the claims at face value, there’s an additional distinction here. The $6 million training cost is likely the cost of renting the chips, not the cost of the chips themselves.

If you assume that renting a chip costs $2 per hour, that totals $5.57 million. That’s not the actual cost of the chips. And that doesn’t include the cost of the experiments, previous research, salaries, and other overhead. So no, it didn’t just take $6 million to build. It probably took many times more.

And finally, let me touch on another popular question that has been talked about a lot lately: Will the bubble burst? I'm keeping the answer short, no, because I don't see a bubble right now.

The information, comments and recommendations contained herein are not within the scope of investment consultancy. Investment consultancy services are provided within the framework of the investment consultancy agreement to be signed between brokerage firms, portfolio management companies, banks that do not accept deposits and customers. The comments in this article are only my personal comments and these comments may not be appropriate for your financial situation and risk return. For this reason, investments should not be made based on the information and comments in my articles.

Source: DeepSeek-V3 Technical Report