Why DeepSeek didn’t cause an investor frenzy again in 2025
About a year ago, DeepSeek shook up the world of artificial intelligence.
Shares in some of the leading Western tech companies fell as markets went into panic mode in anticipation of a new model from a relatively unheard-of Chinese artificial intelligence lab that challenges the premise of US dominance in space.
Nvidia It lost nearly $600 billion in market value, down 17%. US chip manufacturer broadcom also fell by 17% and ASML It fell 7 percent in one day.
Eleven months have passed and these companies have not only recovered, but continued to grow. Nvidia became the first company to reach a $5 trillion valuation in October. While Broadcom’s shares increased by 49% in 2025, ASML’s shares increased by 36%.
“Fireplace [DeepSeek] (R1) caused a broad, visible repricing as it shifted global beliefs about frontier model cost curves and China’s competitiveness, and did so in a way that directly impacted the semiconductor and hyperscaler narrative,” Gartner senior managing analyst Haritha Khandabattu told CNBC.
Since then, DeepSeek has released seven new model updates. None caused the kind of waves seen in January. So why didn’t the markets react?
shock factor
Founded in 2023, DeepSeek released a free, open-source large language model (LLM) called V3 in late 2024. This model was trained with less powerful chips and at a fraction of the cost of models built by companies like OpenAI and Google, he said.
Weeks later, in January 2025, he published a reasoning model called R1 that met similar criteria or outperformed many of the world’s leading LLMs.
The Chinese AI lab’s announcement in January “really surprised the market,” Alex Platt, senior analyst at investment firm DA Davidson, told CNBC. “Narrative [at the time] “China was 9 to 12 months behind the United States.”
The promise of a model that achieves results similar to state-of-the-art systems but uses less computing has raised concerns that demand for AI infrastructure in the market will be impacted and the revenues of companies like Nvidia will be hit, Brian Colello, senior equity analyst at Morningstar, told CNBC.
“Instead, we see no slowdown in spending in 2025, and looking ahead, we anticipate spending accelerating in 2026 and beyond.”
There are also releases of the type DeepSeek has been making since January; these are all updates to the V3 and R1 models, as opposed to completely new models.
limited transaction
Analysts told CNBC that one reason DeepSeek hasn’t released a new model is likely due to limited computing.
“There was a huge bottleneck in the computer,” Platt said. “You can only do so much algorithmic research and find architectural creativity.”
According to the Financial Times, the artificial intelligence company postponed the launch of the R2 model, originally planned for May, due to difficulties in training on its own developed Huawei chips. reported in August.
The publication said Chinese officials encouraged DeepSeek to use the processors as it aims to reduce reliance on US alternatives in the face of export controls on Nvidia’s most powerful chips. DeepSeek has been approached for comment on the report.
“The amount of computing power China has had access to over the last few years has been constrained largely by U.S. restrictions on chip sales,” Chris Miller, author of “Chip War,” told CNBC.
“If you want to build advanced models, you need access to advanced computing.”
DeepSeek said in a research paper published earlier this month that it acknowledged “certain limitations compared to frontier closed-source models” such as Gemini 3, including computing resources.
Markets were also reassured that the United States would continue its leadership in the field of artificial intelligence with new advanced models emerging from leading laboratories in the West.
OpenAI released GPT-5 in August, Anthropic Claude released Opus 4.5, and Google released Gemini 3 in November.
“Competition between these providers is intense due to rapid model releases and increasing improvement in capabilities,” Gartner Analyst Arun Chandrasekaran told CNBC. “As a result, fears of a sudden commoditization shock have been alleviated.”
But there are signs that DeepSeek is preparing for the release of a more significant model in the coming months. On New Year’s Eve, the company published a paper detailing a more efficient way to develop AI models.
Dan Ives of Wedbush Securities expects more market shocks to come. “We’ll continue to see some of these moments that we’ve seen next year,” he told CNBC.
“There will be another DeepSeek.”
