KUALA LUMPUR (Jan 31): BMI, a Fitch Solutions unit, deemed the sell-off in global technology stocks due to Chinese artificial intelligence (AI) platform DeepSeek as "somewhat exaggerated", as it hasn't significantly disrupted the market.

"...and a correction to the share prices of some of the most AI-centric companies should soon follow," said the research house on Thursday.

The market sell-off, BMI said, has signalled a reassessment by investors, suggesting that the entire capital expenditure of the largest hyperscalers has shifted due to the emergence of more economical, open-source large language models.

“The market had priced in a favourable scenario, assuming continued dominance by Nvidia, with investments from hyperscalers deemed justified.

"The sell-off may have been exacerbated by the self-fulfilling prophecy of markets surrounding margin calls, leading to divestments from major tech names to maintain sufficient margin levels,” BMI said.

Assuming the DeepSeek development data are correct, it would become easier and cheaper for companies to implement AI using its approach, benefiting those who execute this effectively, BMI said.

Despite no disruption to the market, there is concern that Nvidia's attractiveness has diminished with this new kid on the block, BMI noted.

“This refers to the barriers to entry, which have seemingly disappeared with a cheaper computing alternative. The efficiency gains underscore the market's fragility, particularly regarding the high valuations of the Mag-7 (Magnificent Seven) stocks — Google, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla — coupled with stable or declining interest rates,” it said.

Hence, a significant repricing of major US stocks would not be surprising, BMI flagged.

Nonetheless, big tech executives along with Europe's largest tech firm ASML have welcomed DeepSeek in a positive light with respect to the semiconductor industry, according to BMI.

Meanwhile, Meta and Microsoft have acknowledged DeepSeek's disruption, but both are committed to continuing their AI investment.

“Meta indicated that future iterations of its open-source Llama models aim to be the most widely used AI models,” BMI said.

“Meta will likely seek to displace OpenAI and Anthropic by leveraging open-source models and profiting from the applications built upon them. Lower inference costs would be ideal for large tech companies with AI embedded in their P&L (profit and loss)," it said.

Still, many American companies remain sceptical of Chinese-based AI even if the chips used are export-compliant, noted BMI.

“Nonetheless, the DeepSeek development challenges the narrative that one party's gain is another's loss, suggesting that services need not be built upon proprietary models,” it added.