DeepSeek's new AI model fails to wow markets in a rapidly changing industry

The market reaction to the announcement of DeepSeek's long-awaited next-generation artificial intelligence model has been subdued so far, compared to the Chinese startup's global breakthrough last year after the launch of its low-cost AI models.

The release and international reception of the DeepSeek-V3 and R1 models, which the Hangzhou-based company said were trained with significantly less computing power than their American competitors, sparked a sell-off in technology stocks around the world as investors began to question the logic of large investments in AI infrastructure, Reuters reports.

Analysts described the moment as a "black swan," forcing markets to reassess assumptions about costs, competition and China's ability to innovate despite U.S. restrictions on chips.

But the lukewarm response to DeepSeek-V4, unveiled on Friday, shows how quickly those expectations have changed. Markets and industry have now become accustomed to cheap and efficient models developed with limited resources, which reduces the surprise factor.

"This announcement followed a fairly predictable path," said Lian Jie Su, principal analyst at Omdia, noting that advances in model architecture and efficiency have already been widely researched in industry and academia.

The test data confirms this. According to Artificial Analysis, DeepSeek-V4 Pro shows significant improvement over previous versions, but still ranks among the leading open source models, without a clear advantage over the competition, while rivals such as Kimi and Qwen are closing the gap, Reuters reports.

That contrasts with last year, when DeepSeek seemed to leapfrog domestic competitors, spurring rapid adoption in China and strengthening global influence.

Analysts say that the shock at the time was the result of a combination of factors: high valuations of American technology companies, expectations of dominance by a small number of players, and the emergence of a relatively unknown Chinese startup with unexpectedly strong results.

Those conditions no longer exist.

"The expectation that new players will emerge is now factored into valuations," Su said, adding that markets have become more realistic about the possibilities and limitations of artificial intelligence.

At the same time, competition in China has intensified, as more companies develop increasingly advanced models, reducing DeepSeek's relative advantage.

On Monday, stock markets in South Korea and Taiwan hit new records, fueled by general optimism about the AI ​​sector.

Alfredo Montufar-Helu, director at Ankura China Advisors, said the significance of the V4 model lies less in market influence and more in the technological competition between the US and China.

He pointed out that DeepSeek has adapted V4 to work on Huawei chips, at a time when US export restrictions seek to limit the Chinese market's access to advanced American chips crucial for the development of AI models.

"The 'wow' factor was there last year - that's already been factored in," he said. "The key now is whether China can continue to advance in AI, and possibly do so with its own chips - the geopolitical implications of that would be significant."