Challenging U.S. AI Dominance: Surge of Chinese Startup DeepSeek Threatens Tech Shares, Competing Against OpenAI

Administrator -

Topline

Challenging U.S. AI Dominance: Surge of Chinese Startup DeepSeek Threatens Tech Shares, Competing Against OpenAI

Early Monday, major U.S. tech and AI stocks took a dive in premarket trading, following the successful launch of Chinese startup DeepSeek's latest AI model. The notable achievement of this model, known as R1, impressively outperformed U.S. competitors like OpenAI, running efficiently on less powerful chips. This sparked questions about the U.S.'s dominance in the AI sector.

Key Facts

  • The tech-centric Nasdaq 100 Futures plummeted by 4.3% to 20,950, while the S&P 500 Futures dropped around 2.5% and Dow Futures almost 0.8%.
  • AI-related tech and silicon stocks were adversely affected by the selloff. Notable losses included Nvidia's share price dropping nearly 13% to $124, while Microsoft, Google, and Meta fell 6.7%, 4.6%, and 5.5% respectively.
  • Other chipmakers, such as AMD and Broadcom, suffered as well; their shares slumped 6.3% and 12.9% in premarket, respectively. ASML, the Dutch-listed shares of the world's most advanced chip-making machine maker, saw a 10.62% decrease within two hours post-European market opening.

DeepSeek and R1

DeepSeek, a Chinese AI startup based in Hangzhou, has been making waves ever since its launch in November 2023. Originating from High-Flyer Quant, a Chinese hedge fund that heavily leveraged AI for trades, DeepSeek has distinguished itself with its AI language model and R1, an open-source reasoning model. R1 reportedly matches OpenAI's o1 model performance in certain benchmarks while requiring less computing power.

DeepSeek's Efficiency and Affordability

DeepSeek's model's efficiency is noteworthy, which allows it to function effectively with less computing power and lower costs. MIT Technology Review suggests that this is likely an unintended outcome of U.S. export controls on high-end AI chips to China, forcing China's startups to focus on efficiency. Billionaire and Silicon Valley venture capitalist Marc Andreessen described R1 as "AI's Sputnik moment" on Reddit. DeepSeek's API pricing is significantly cheaper than OpenAI’s, at $0.55 per million input tokens and $2.19 per million output tokens, compared to $15 and $60 for OpenAI's o1.

DeepSeek's Censorship Policy

DeepSeek, like several Chinese tech platforms, appears to censor content that Beijing deems sensitive. For example, when asked about the Tiananmen Square protests, DeepSeek's chatbot responds with: “Sorry, that's beyond my current scope. Let's talk about something else.”

DeepSeek's App Surge

Interestingly, DeepSeek's iPhone app witnessed a surge, topping the U.S. App Store's download charts for free apps on Sunday, surpassing OpenAI's ChatGPT.

DeepSeek's Innovative Approach

DeepSeek's AI model's efficiency is mainly due to its innovative architecture and training techniques, such as:

  1. Mixture-of-Experts (MoE) Architecture: DeepSeek uses MoE, divides the model into multiple experts that process different parts of the input, enabling the model to decide which expert to use for a particular query and saving energy.
  2. Focused Training Data: DeepSeek focuses on generating training data in domains like mathematics where correctness is unambiguous, reducing compute waste on redundant data.
  3. Efficient Reward Functions: The team developed efficient reward functions to identify which new training examples would improve the model.
  4. Low Training Costs: DeepSeek only spent 2.79 million GPU hours to train its model, significantly less than competitors.

After the successful launch of DeepSeek's AI model R1, which outperformed U.S. competitors with less computing power, NASDAQ futures declined significantly. The Nasdaq 100 Futures dropped by 4.3%, with tech giants like Nvidia seeing a 13% decrease in their share price. As a result, iPhone charts showed a surge in downloads for DeepSeek's app, surpassing OpenAI's ChatGPT. The efficiency of DeepSeek's model, attributed to innovative techniques and less reliance on high-end chips, has sparked discussions about the future of AI stock markets, including NASDAQ futures and U.S. stock futures.

Read also: