Deepseek AI And Its Impact on Global Stock Market

Paul Krugman, a renowned American economist, noted in his recent Bluesky post: "Tech stocks were partying like it was 1999; soon we may be hungover like it's 2000," referencing the stock market dot-com crash of 2000.

Over the last year and a half, AI has been all hype—Nvidia, a chip maker whose advanced chips are used for AI processing, went from $5 per share in early 2020 to $150 per share in late 2024. On Monday, January 27th, it lost 17% of its value. While the changes were not as rapid for all U.S. tech giants, the losses were widespread.

So, what happened?

DeepSeek AI: A Game-Changer in the AI Market

The market reaction was triggered by the release of R1, a large language model (LLM) developed by DeepSeek, a Chinese company. And the model is good—mostly comparable to OpenAI’s ChatGPT. That alone wouldn’t mean much, but DeepSeek managed to achieve this performance at a fraction of the cost and without using the most advanced chips (the U.S. restricts the export of GPUs and other advanced chips to China).

This explains the fall in Nvidia’s market value—until now, investors believed it held an unquestionably dominant position in the chip market. Now, this belief is gone.

The DeepSeek AI breakthrough has sparked a broader analysis of the AI industry, leading to a reassessment of the future of technology and its economic implications.

Rethinking the Economics of AI

More pressing are the questions about the economics of this technology. Analysts were treating AI LLMs like any other software, where economies of scale rule, and the dominant companies’ market position allows them to derive outrageous profits from their customers.

However, with LLMs, this assumption doesn’t seem to hold true, as replicating other companies’ technology has just been proven possible and cheap. The AI industry analysis suggests that market might be much more competitive than previously thought, and with fierce competition, profits are likely to be kept in check.

DeepSeek vs AI Giants: A New Competitive Dynamic

Replicating what OpenAI has done is one thing, but delivering frontier performance is another. OpenAI’s dominance in this aspect has not been questioned, but if replicating any new advancements becomes easy, investors may think twice before channelling large sums of money into AI. This could potentially slow down progress.

Another aspect of AI economics under scrutiny is the amount of data needed to train models. So far, the common belief has been “the more data, the better,” but DeepSeek’s model has delivered strong performance with much less training data. This question remains unresolved, and there is a good chance that it will struggle with niche problems and specialist queries.

To summarize, markets are going through a reassessment phase, trying to determine how the AI market will function, and which economic principles will apply. This might be a painful realization for investors who went big on AI. But for the average user, this is (probably) a good thing, as more competition tends to make products cheaper.