Crypto AI tokens crushed by DeepSeek-fuelled NVDA sell-off

In early trading on Monday, AI-related cryptocurrencies saw sharp price drops amid a wider downturn in U.S. equities, as investors began to digest news of Chinese-based large language model DeepSeek having all the familiar abilities OpenAI’s, Google’s or Meta’s popular large language models, but operating at a fraction of the cost.

Amidst the sell-off, leading AI-related tokens such as Near Protocol (NEAR), Internet Computer (ICP), Render (RENDER), and Artificial Superintelligence Alliance (FET) experienced drops of 8%, 9%, 10% and 11%, respectively, over the past day. This sharp decline in crypto AI tokens comes on the back of broader market trends, including a significant drop in U.S. tech stocks.

The Dow opened today about 369 points, or 0.8%, lower. The SP 500 fell by 2%, and the tech-heavy Nasdaq plunged by 3.6%. For context, the Nasdaq hasn’t closed 4% lower since September 2022. Nvidia stock, in particular, saw a significant downturn, falling 12% in early trading to around $124.

DeepSeek disruption leads to market volatility

Despite the volatility in traditional markets, Redstone COO Marcin Kazmierczak stressed the importance of recognizing that the blockchain and crypto AI ecosystems differ from the stock market in several ways.

"The blockchain and AI spaces are fundamentally different, with each driven by unique use cases and long-term innovation, like the upcoming collaboration between AI Agents through the CLARA framework," Kazmierczak told The Block. "It’s only a matter of time before DeepSeek models start being leveraged in Crypto AI projects alongside the recently announced Operator - AI agent by OpenAI."

This sentiment was echoed by Bitget COO Vugar Use Zade, who acknowledged that technological disruptions, such as DeepSeek’s AI models, often trigger short-term market adjustments.

"The broader impact of such developments intersects with existing macroeconomic pressures, including U.S. interest rate decisions and the FOMC meeting, amplifying volatility in the crypto market," Zade told The Block. "Bitcoin’s recent dip correlates with the combination of these factors, and as history records, tech advancements and their market impacts often stabilize as ecosystems adapt."

The emergence of DeepSeek is particularly noteworthy for its ability to build AI models at a fraction of the cost of established players. With pricing reportedly at just 14 cents per million input tokens—compared to $15 for OpenAI’s GPT-4 model—DeepSeek’s model potentially offers significant cost advantages. This efficiency is sparking concerns among investors about the sustainability of current AI investment strategies, especially as tech giants like Microsoft and Meta have committed billions to their own AI initiatives.

Analysts react to DeepSeek disruption

JPMorgan analyst Sandeep Deshpande suggested that DeepSeek’s success could signal a turning point in the AI investment cycle.

"The news over the past few months has been about the announcements of Microsoft, which is spending $80 billion in 2025, while Meta recently announced investments between $6 billion and $65 billion," Deshpande said, according to CNBC. “Thus, with these considerable sums flowing into AI investments in the U.S., that Deepseek’s highly efficient and lower resource-intensive AI model has shown such significant innovation and success is posing thoughts to investors that the AI investment cycle may be over-hyped and a more efficient future is possible."

However, other analysts, including Atif Malik from Citi , have pointed out that DeepSeek’s efficiency does not necessarily diminish the need for advanced computing hardware like GPUs, which remain essential for fine-tuning and optimizing large-scale AI models.

"While DeepSeek's achievement could be groundbreaking, we question the notion that its feats were done without the use of advanced GPUs to fine tune it and/or build the underlying large language models that the final model is based on through the distillation technique," Malik said in an investment note on Monday. "What is real is that DeepSeek is charging end users significantly less, 14 cents per million input tokens, vs. $15 for OpenAI's o1 model."

The Block's GMCI USA index dropped 17.39 points over the past day, down 6.6% to 245.93.