Nvidia caught up in software stock sell-off, shares fall for four consecutive days, but one analyst says this makes no sense
The recent plunge in software stocks has spread to the chip sector, but an analyst at Bank of America says this sell-off is completely irrational.
Bank of America analyst Vivek Arya said in a report on Tuesday that the "indiscriminate" volatility surrounding chip stocks "seems contradictory and may be another case of excessive sell-off similar to the DeepSeek incident."
On Wednesday, shares of chipmakers including Nvidia and Broadcom fell more than 3% and 6%, respectively. Nvidia has fallen for four consecutive days. Memory and storage stocks such as Micron Technology and SanDisk, which had previously surged due to AI-driven demand, also saw double-digit declines.
Arya believes that the extent of the sell-off reflects a "contradictory" viewpoint. In his report, he noted that the decline in AI-related chip stocks suggests that investments in AI are "deteriorating to the point where returns will be weak and further growth unfeasible." At the same time, the decline in software stocks suggests that the market believes "AI adoption will be so widespread and productivity-enhancing that software-based businesses and applications are destined to disappear."
"Both of these outcomes cannot happen at the same time," Arya said, adding that this is reminiscent of concerns about DeepSeek, "and these concerns ultimately proved to be unfounded." The AI startup's reasoning model, released in January 2025, disrupted tech stock trading as investors worried that if DeepSeek could develop a competitive AI model at low cost, other companies could do so as well without needing so many expensive and powerful chips.
But Arya said that after that brief market episode, what followed was even more investment in AI and an accelerated growth in AI tokens (the information fragments processed by AI models). He added that in 2025, cloud capital expenditure is expected to grow by 69% year-on-year, more than double the initially forecasted increase of 20% to 30%.
Arya said that while current AI models are impressive, it may take several years for them to prove their value from a productivity perspective. Meanwhile, he believes that AI investment will not slow down, and expects that even after AI models achieve significant productivity breakthroughs, investment will continue because they still need improvement and must maintain user engagement.
Arya said: "While we cannot predict the final form of the software industry, we believe the chip industry has already benefited and will continue to benefit from AI development."
Arya noted that enterprise adoption of AI is still in its early stages, and current investments are also coming from sovereign nations hoping to deploy and control AI initiatives.
Meanwhile, Arya said that valuations in the chip sector have already accounted for cautious sentiment about potential spending slowdowns or declining earnings expectations, "and we believe this scenario is unlikely."
Editor: Li Tong
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
The 5% U.S. Treasury yield storm is coming! The refinancing time bomb countdown begins—who will be the first victim?
The 10-year US Treasury yield has surpassed 5%, reaching a new high since 2007. The longer high interest rates persist, the greater the refinancing pressure will be on real estate companies, commercial real estate, and highly indebted firms, with systemic risks likely to accelerate and emerge within the next 12 to 18 months.
Don't Fight the Profit Cycle! Will U.S. Stocks Break 8,000 Points This Year?
Jefferies predicts that, driven by the dual engines of the AI investment boom and stronger-than-expected corporate earnings, the S&P 500 index is expected to soar to 8,000 points by the end of 2026 and further reach 9,000 points in 2027. AI-driven profit expansion has spread from the "Magnificent Seven" to the entire market, with the S&P 500's EPS forecast to surge by 35% this year, far exceeding market consensus—marking the strongest earnings supercycle since 1995! The only real threat: if US Treasury yields continue to spike, the risk of valuation compression cannot be ignored.
Will “continual learning” AI extend the memory “shortage” until 2031?
Citi believes that as AI enters the era of "continuous learning" beyond simple training and inference stages, demand for HBM, server DDR5, and enterprise SSDs (eSSD) will experience explosive and simultaneous growth starting from 2027. While demand will surge rapidly, the supply side is constrained by HBM production capacity usage and slower technology migration, leading to expansion lagging far behind demand. This supply-demand imbalance is expected to continue until 2031.
Crypto prices
More
